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 1752324910.1371/journal.pmed.0020118EditorialMedical EducationMedical EducationMedical journalsEngaging Students in PLoS Medicine
EditorialThe PLoS Medicine Editors 4 2005 26 4 2005 2 4 e118Copyright: © 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.
==== Body
Medicine might not be the oldest profession, but the practice and the teaching of it certainly stretch far back into antiquity. In the developed world, medical schools are now thriving enterprises whose main aim remains to turn out sufficient numbers of well-trained medical practitioners. But alongside this primary purpose, medical schools have become centers of excellence for research also, mostly basic scientific research. Certainly the medical schools of the University of Cambridge and University of California at San Francisco (UCSF)—the two schools closest to the PLoS editorial offices—have large research programs.
Getting the balance right between understanding the research that will drive future medical discoveries, and providing training in clinical practice is obviously crucial in training tomorrow's doctors. However, follow-up studies of those who graduate suggest that medical schools may not be producing doctors who are happy with the profession, or who fit all the needs of medical services today. In the United Kingdom, two striking observations came out of recent surveys. One is the difficulty in recruiting doctors who want to become general practitioners, the primary care doctors who form the backbone of the UK health service. Whereas in the 1970s and 1980s 40%–50% of the qualifiers intended to enter general practice, in the 1990s the proportion was only 20%–26%. (BMJ 326: 194–195). But even more disturbing is the substantial proportion of doctors in the same surveys who indicated that they regularly consider leaving medicine altogether. Contrast these results with those of a survey from Uganda, where the survival of physicians was the most urgent concern. From a cohort of 77 doctors who graduated from the University of Makere, Uganda, in 1983, 22 had died over the following 20 years (BMJ 329: 600–601). The most important cause of death was presumed to be AIDS. But all of the surviving doctors were in some form of medical work.
What these figures show perhaps more than anything is the difficulties of generalizing about what are the top priorities for medical students across the globe. To provide a place for all students to debate the issues that matter to them, PLoS Medicine is launching the Student Forum, a new quarterly section written by medical students. An international team of student advisers has been elected that will help shape the content of this new section. The Student Forum recognizes the vital role that medical students have in guiding the future of the medical profession.
The first Student Forum essay brings together students from three continents: Brian Palmer of the American Medical Student Association, Amanda Wong of the European Medical Students' Association, and Mohit Singla of the Indian Medical Students' Organization. They lay out some of their concerns, such as the need for medical education to distance itself from the pharmaceutical industry and the importance of students worldwide having unfettered access to the medical literature. The students also ask questions about their training, commenting that the “reductionistic logic of medical education offers only false comfort, and only to those who content themselves with pathways and receptors.”
Perhaps surprisingly, this opinion is not far away from that of an established academic clinician and editorial board member of PLoS Medicine, Jonathan Rees. In another essay in this issue, Rees similarly questions the overemphasis on basic research in medical schools, posing the question, “why are our institutions not fit for the purpose of improving patients' health?”
Not an easy question to answer, but one that might also be relevant to the question of what a medical journal, especially one that wants to engage students, should publish. Should we follow Rees's advice on what the composition of a medical school curriculum should be and have no more than 20% of our articles devoted to basic scientific research as applied to medicine? In this issue the balance of our research papers comes close to this suggestion, with one paper on the differentiation of insulin-producing cells from human neural progenitors, and other papers throughout the journal on topics as diverse as climate and dengue fever, treatment of malaria, HIV genotypes, prescription practices, counterfeit drugs, Buruli ulcer, and health and human rights. But, perhaps like the medical schools, we are not paying enough attention to other topics Rees points out as important—operations research, decision-making, informatics, and economics.
In his lightning trip through medicine, Blood and Guts: A Short History of Medicine (W. W. Norton, 2003), the late medical historian Roy Porter wrote that “the biomedical model can be myopic, searching ever more microscopically for disease but often omitting the wider picture of populations, environments and health.” In his description of the differing opinions on the role in 19th century medical schools of the new practice of laboratory medicine, it becomes clear that the debate over what should be taught and what matters to medical students might not be new, but it is clearly as important as ever.
PLoS Medicine is launching a new regular student section run in association with an international panel of medical student advisors
| 17523249 | PMC1087211 | CC0 | 2021-01-05 10:39:45 | no | PLoS Med. 2005 Apr 26; 2(4):e118 | utf-8 | PLoS Med | 2,005 | 10.1371/journal.pmed.0020118 | oa_comm |
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 1583973910.1371/journal.pmed.0020080Research ArticleCardiology/Cardiac SurgeryEpidemiology/Public HealthHealth EconomicsHealth PolicyDrugs and adverse drug reactionsEvidence Based PracticeGuideline adherenceHealth EconomicsHypertensionFirst-Line First? Trends in Thiazide Prescribing for Hypertensive Seniors Trends in Thiazide PrescribingMorgan Steve
1
*Bassett Kenneth L
1
Wright James M
2
Yan Lixiang
1
1Centre for Health Services and Policy Research, University of British ColumbiaVancouverCanada2Pharmacology and Therapeutics Department, University of British ColumbiaVancouverCanadaKeech Anthony Academic EditorUniversity of SydneyAustralia
Competing Interests: The authors have declared that no competing interests exist.
Author Contributions: SM, KB, and JW designed the study. SM and LY analyzed the data. All authors contributed to writing the paper.
*To whom correspondence should be addressed. E-mail: [email protected] 2005 26 4 2005 2 4 e8020 7 2004 12 1 2005 Copyright: © 2005 Morgan.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.
Trends in the Prescribing of Thiazides for Hypertension
Background
Evidence of reduced cardiovascular morbidity and mortality as well as cost support thiazide diuretics as the first-line choice for treatment of hypertension. The purpose of this study was to determine the proportion of senior hypertensives that received thiazide diuretics as first-line treatment, and to determine if cardiovascular and other potentially relevant comorbidities predict the choice of first-line therapy.
Methods and Findings
British Columbia PharmaCare data were used to determine the cohort of seniors (residents aged 65 or older) who received their first reimbursed hypertension drug during the period 1993 to 2000. These individual records were linked to medical and hospital claims data using the British Columbia Linked Health Database to find the subset that had diagnoses indicating the presence of hypertension as well as cardiovascular and other relevant comorbidities. Rates of first-line thiazide prescribing as proportion of all first-line treatment were analysed, accounting for patient age, sex, overall clinical complexity, and potentially relevant comorbidities. For the period 1993 to 2000, 82,824 seniors who had diagnoses of hypertension were identified as new users of hypertension drugs. The overall rate at which thiazides were used as first-line treatment varied from 38% among senior hypertensives without any potentially relevant comorbidity to 9% among hypertensives with previous acute myocardial infarction. The rate of first-line thiazide diuretic prescribing for patients with and without potentially relevant comorbidities increased over the study period. Women were more likely than men, and older patients were more likely than younger, to receive first-line thiazide therapy.
Conclusions
Findings indicate that first-line prescribing practices for hypertension are not consistent with the evidence from randomized control trials measuring morbidity and mortality. The health and financial cost of not selecting the most effective and least costly therapeutic options are significant.
Prescription practices for hypertension were not consistent with existing evidence--newer and more expensive drugs were prescribed preferentially despite evidence that thiazides are equally or more effective
==== Body
Introduction
Hypertension significantly increases the risk of serious cardiovascular morbidity and mortality and is the leading primary diagnosis for patient visits to physicians' offices in Canada [1]. Pharmaceuticals used in the treatment of hypertension constitute the leading therapeutic category of prescription drugs in Canada, accounting for 20% of total prescription drug sales in Canada [2]. Decisions concerning the management of hypertension will therefore have significant impact on population health and health care costs. First-line treatment of hypertension with thiazide diuretics has been shown to significantly reduce serious cardiovascular morbidity (stroke and myocardial infarction [MI]) and mortality in randomised controlled trials, with benefits at least as great as first-line treatment with other classes of antihypertensive drug [3,4,5,6]. Thiazide diuretics are also less costly than other antihypertensive drugs. Therefore, thiazide diuretics are the most cost-effective first-line therapeutic option for the majority of patients.
The purpose of this study was to determine whether prescribing practices are in accordance with this evidence. Using population-based research datasets at the University of British Columbia's Centre for Health Services and Policy Research, we calculated trends in first-line prescribing of antihypertensive drugs for seniors (residents aged 65 or older) in the province of British Columbia over the period of 1993 to 2000. We analyse the likelihood of first-line thiazide prescribing as a function of patient age, sex, overall clinical complexity, and potentially relevant comorbidities.
Methods
Administrative data from public medical, hospital, and pharmaceutical insurance programs were analyzed to determine trends in first-line hypertension drug use. All residents of British Columbia are covered under a comprehensive public health insurance plan for medical and hospital services. Public insurance for prescription drugs is restricted to selected populations, but includes universal and comprehensive drug coverage for all seniors [7]. Administrative claims data from British Columbia PharmaCare, the public drug plan for all seniors, were used to track prescription drug use for this demographic cohort. Data from the public health insurance plan for medical and hospital services were used to identify diagnoses of hypertension and potentially influential comorbidities for those seniors who filled prescriptions for drugs commonly used in the management of hypertension.
The study cohort included all community-dwelling seniors who had evidence of first-time hypertension drug use and hypertension diagnosed in administrative databases. The classes of hypertension drugs were defined by the World Health Organization Anatomical Therapeutic Chemical classification system, and included angiotensin-converting enzyme (ACE) inhibitors, angiotensin-II receptor blockers, beta-blockers, calcium-channel blockers, alpha-antagonists, thiazide diuretics, nonthiazide diuretics, and other antihypertensives (e.g., reserpine). First-time hypertension drug use was defined as the receipt of any hypertension drug following at least one year of eligibility for PharmaCare coverage during which no prescriptions for any hypertension drugs were filled. Eligibility for coverage was measured starting in January 1992. A patient was considered a first-time user only once; those who stopped drug treatment and then filled prescriptions for antihypertensive drugs a year or more later were only included in the analysis of first-line use based on their initial course of treatment.
The study cohort was limited to the subset of first-time users of antihypertensive drugs who had a diagnosis of hypertension in administrative data records. This included patients for whom there was at least one diagnosis of hypertension (an International Classification of Diseases-9 code in the 401.x family) in Medical Services Plan and Hospital Separations databases spanning two years prior to the date of first-line hypertension treatment and one year following first-line treatment. Each Medical Services Plan record pertains to a fee-for-service billing and contains one diagnosis for the physician visit or related service. This is assumed the primary diagnosis for the visit or service. Hospital Separations records describe patient stays at the point of discharge and contain up to 16 diagnoses, all of which were searched. In order to avoid bias as a result of under-reporting the diagnosis of hypertension, diagnoses of hypertension were sought from records spanning two years prior to and one year following the first prescription for hypertension. Diagnoses occurring after drug use were included in the study because, for approximately one-third of users of hypertension drugs, the diagnosis of hypertension was recorded as the primary diagnosis during follow-up after the initiation of drug treatment.
Medical and hospital data spanning two years prior to the date of first-time hypertension treatment were searched for International Classification of Diseases-9 diagnostic codes pertaining to potentially relevant comorbidities. Identified using “Expanded Diagnostic Clusters” [8], the diagnoses sought included acute MI, aneurysm, angina, congestive heart failure, cardiac arrhythmia, cardiovascular valve disorder, gout, hyperlipidaemia, ischaemic heart disease, migraine headache, diabetes mellitus, and chronic renal failure. As has been done elsewhere [9], these comorbidities were selected because clinical guidelines, expert opinion, or marketing information suggested that their presence might influence the likelihood of a physician prescribing thiazide therapy; they are not necessarily substantiated by randomized controlled trial evidence of morbidity or mortality implications. A list of these diagnoses and their possible effects on treatment choices is provided in Table 1. Additional diagnoses, such as mild kidney disease, are also likely to be relevant, but these are not captured in administrative data.
Table 1 List of Potentially Confounding Conditions
Patients' first-line therapy was determined by the categories of drugs that they received on the date they first filled a hypertension prescription. In addition to patients who started treatment with the purchase of a thiazide alone or a combination product that contained thiazides, patients whose first hypertension treatment was a thiazide diuretic along with another antihypertensive drug were also considered to be receiving first-line thiazide therapy.
Statistical Analysis
The analysis is based on data describing the entire population of users fitting the protocol definitions. Summary statistics describe average cost per day of treatment for each category of drug and first-line treatment choices over time. Multivariate logistic regression determined the probability that a patient would receive a thiazide diuretic as a first-line drug treatment. Independent variables for the multivariate analysis included patient age, sex, and year of first prescription. As a measure of overall clinical complexity, the regression included the number of different types of prescription drug purchased within the year prior to the patient's first antihypertensive drug purchase. Binary variables indicating prior diagnosis of comorbidities were included as independent variables. The analysis also included a binary variable that identified patients who filled no more antihypertensive prescriptions during the year following their first antihypertensive drug purchase. Wald chi-square tests were calculated to determine statistical significance.
Results
A total of 82,824 British Columbia residents over age 65 received an antihypertensive drug for their first time between 1993 and 2000. Over the period of study, the share of British Columbia seniors who were first-time users of antihypertensive drugs in a given year increased from 1.8% to 2.2%. The median age of patient at the date of first-time hypertension use was 72; 55% of newly treated seniors were female.
The average cost per days-supply of hypertension treatment received by British Columbia seniors varied by type of drug prescribed and was affected by policies implemented over the period of study (Figure 1). Available from 1996 onward, angiotensin-II receptor blockers were the most expensive treatment options by 2002, costing Can$1.16 per day of treatment. While the average cost per day of treatment with calcium-channel blockers had been as high as Can$1.32, it declined to Can$1.04 following the 1997 implementation of a reference-based pricing policy [10]. The daily cost of ACE inhibitors fell from Can$1.00 to Can$0.87 per day, also due to reference pricing [11]. Owing to increased generic availability, the cost per day of therapy on beta-blockers fell from Can$0.59 to Can$0.34 over the period. Over the entire period, thiazide diuretics were the least expensive treatment option, costing less than Can$0.01 per day of therapy.
Figure 1 Average Cost per Day of Therapy Supplied to British Columbia Seniors by Category of Hypertension Drug
Thiazide diuretics were prescribed as a first-line therapy for less than a third (29%) of newly treated patients. Over time, the trend in the share of newly treated patients that received thiazide diuretics as a first-line therapy was similar for those with and without diagnoses indicating potentially influential comorbidities. Figure 2 illustrates these time trends. The shares of patients with and without comorbidity that received first-line thiazide treatment increased from 25% and 16%, respectively, in 1995 to 42% and 23% in 1997. These shares remained relatively stable thereafter, ending at 42% and 22% in 2000. First-line thiazide diuretic prescribing as a share of newly treated patients did not exceed 45% at any point from 1993 to 2000.
Figure 2 Thiazide Diuretics as Percentage of First-Line Treatments for Hypertensive Patients with and without Potentially Confounding Comorbidities
The proportion of newly treated patients receiving thiazides as first-line therapy varied according to the presence of comorbidity diagnoses (Table 2). Nearly half (39,764) of these newly treated senior hypertensives had no medical or hospital records indicating a potentially confounding comorbidity; 38% of these “uncomplicated” patients received thiazides as first-line treatment. The 22% of newly treated patients that had multiple comorbidities received thiazides at a rate of approximately 1 in 7 (14%). Those with evidence of previous acute myocardial infarction received thiazides least frequently (9%). As shown in Table 3, multivariate logistical regression analysis indicated that first-line treatment was affected by a time trend (rates of thiazide use increased over the period, p < 0.001), gender (women more likely to receive thiazides, p < 0.001), and age (older patients more likely to receive thiazides, p < 0.001). The measure of overall clinical complexity (number of different drugs used) did not have a significant impact. After adjusting for these factors, all potentially influential concurrent diagnoses other than migraine headache were associated with a statistically significant reduction in the likelihood of first-line thiazide diuretic use. Patients with diagnoses of acute MI, angina, or chronic renal failure were less than half as likely to receive thiazide diuretics as those without such diagnoses.
Table 2 Total Newly Treated Patients and Rate of First-Line Thiazide Use by Comorbidity
Table 3 Logistic Regression Results—Odds Ratios
Study Limitations
Despite the advantage that it is population based, this research study is limited by the use of administrative claims data. Claims data may result in under-reporting of actual conditions [12,13]. In an effort to compensate for this, the definitions of comorbidities were based on one or more diagnoses over a two-year period. A further limitation of claims-based data is that they track only those prescriptions that are dispensed from pharmacies. This analysis therefore does not track first-line use by way of professional samples given to patients. Because thiazide diuretics are no longer patented, none of their manufactures provide professional samples in British Columbia. Thus, the results may over-report the true first-line use of thiazides.
Discussion
The overall share of British Columbia senior hypertensives that received thiazide diuretics as first-line treatment ranged from 21% to 34% between 1993 and 2000. Thiazide use was affected by the presence of diagnoses for certain comorbidities; however, the rates of first-line thiazide use were relatively low (below 45%) for patients both with and without any evidence of conditions that might influence treatment choice. We believe that such rates of first-line thiazide use among newly treated hypertensive seniors are low, given the evidence for the morbidity and mortality benefit associated with the first-line use of this class of antihypertensive drugs [4,5,6,14] and the substantial cost differences between products.
Because the cost per day of treatment on alternative treatments is upwards of 100 times that of thiazide diuretics, the financial consequences of first-line treatment decisions are significant, even when limited to the group of “uncomplicated” cases. For the subset of patients with no comorbidities, the additional cost to British Columbia PharmaCare in 2000 stemming from decisions to treat newly hypertensive seniors with nonthiazides was approximately Can$1 million per year. This additional cost will accumulate for each year that these incident cases continue to receive treatment. It is therefore necessary for professional associations and those who pay for (and benefit from) pharmaceutical benefits to work together to ensure that prescribing practices more accurately reflect research evidence.
One might conjecture that low rates of first-line thiazide use result from physicians' concerns regarding the presence of comorbidity. In our study, approximately half of the newly treated hypertensive seniors had diagnoses associated with specific comorbidities that may influence prescribers' beliefs about the appropriateness of thiazide diuretics. In all instances except migraine headaches, the presence of a comorbidity diagnosis predicted a decreased incidence of first-line thiazide use. Some of the associations are rational and consistent with outcomes evidence. Following acute MI, for example, beta-blockers and ACE-inhibitors are first-line choices because they have been shown in randomized controlled trials to reduce morbidity and mortality in addition to having a blood pressure lowering effect [15,16]. It is encouraging that in accordance with the evidence in this population, we found that only 9% of patients received thiazide diuretics as first-line therapy. Some of the associations that we found between comorbidity and first-line antihypertensive choice were not substantiated by evidence of morbidity and mortality implications. Hyperlipidaemia, for example, was associated with a statistically significant decrease in thiazide use versus patients without this diagnosis. This is likely explained based on the reported small increase in cholesterol associated with thiazides [17], but is not rational based on the morbidity and mortality evidence from randomized controlled trials [5,6,17].
Clinical practice guidelines starting in the early 1980s recognized the benefits of treatment with thiazide diuretics by recommending them as the first-line treatment of choice for virtually all hypertension patients [18,19,20,21,22,23,24,25]. Newer guidelines, while acknowledging thiazide diuretics as a first-line choice, increasingly recommended other classes of antihypertensive drugs as “alternate” first-line choices or recommended other classes for specific patient populations [26,27,28]. An example of this is the recommendation against use of thiazide diuretics for patients with diabetes mellitus. It was argued that thiazide diuretics might not achieve the same morbidity and mortality benefits because they increase blood glucose in some patients. However, this effect of thiazides on glucose has not diminished the morbidity and mortality benefit seen in randomized controlled trials: patients with diabetes achieve similar morbidity and mortality benefit with thiazides to non-diabetic patients [29]. Similarly, recent clinical practice guidelines recommend ACE-inhibitors as the preferred first-line therapy for hypertensive patients with mild kidney disease [28]. Proponents of ACE-inhibitors argue that hypertensive patients with kidney disease benefit both from the general antihypertensive effect of ACE-inhibitors and a specific 'kidney sparing' effect associated with this class of drugs. This theoretical benefit has not been substantiated in randomized controlled trials [30].
Given that hypertension treatment is one of the fastest growing drivers of seniors' drug expenditure in British Columbia [31], a critical challenge facing clinicians and policy makers is to use lower-cost therapeutic options whenever they produce equivalent or superior morbidity and mortality benefit to patients. This can be a particularly difficult task when set against a backdrop of marketing claims that draw associations between marketed products and superior performance for using surrogate markers for sub-populations. One mechanism for improving the cost-effectiveness of product selection decision is education. Our study results show that first-line thiazide therapy increased towards the end of 1995 and through 1996, and then remained at a relatively stable rate of approximately 30%. The increase in the rate of first-line thiazide prescribing coincided with the publication of two Therapeutics Letters, produced by the Therapeutics Initiative at the University of British Columbia, on the topic of first-line hypertension treatment [4,14]. The observed impact of these letters, which were mailed to all prescribing physicians in British Columbia in the summer and fall of 1995, has been reported elsewhere [9,32]. Except for this increase, the rate of first-line thiazide therapy did not change substantially over the study period despite a growing body of available evidence indicating the morbidity and mortality benefit of first-line thiazide therapy [5,6,33,34]. Policymakers might consider engaging in regularly reinforced educational interventions to better inform prescribers regarding the appropriate cases for selecting lower and higher cost medicines. Alternatively, or perhaps in conjunction with educational efforts, innovative forms of incentive pricing or benefit-sharing may be used to provide both prescribers and patients the incentive to start with the low cost therapies [35]. Either way, it is incumbent upon the medical profession to weigh carefully the relative costs and benefits of treatments for particular patients; for, every additional dollar spent on drugs without any benefit is a dollar that is unavailable for other medical or pharmaceutical services.
This analysis predates publication of the largest randomized antihypertensive trial designed to answer the question of which first-line antihypertensive is best, the ALLHAT trial [30]. ALLHAT compared four first-line antihypertensive classes (a thiazide-like diuretic [THZ], an alpha blocker, an ACE inhibitor, and a calcium channel blocker) in 33,357 patients. The alpha-blocker caused increased cardiovascular morbidity, and that arm was terminated early [36]. At the end of the trial, total mortality, coronary heart disease, and end-stage renal disease were similar for the remaining three arms. The first-line calcium channel blocker caused more heart failure compared to the THZ and ACE inhibitor. First-line ACE inhibitor caused more strokes compared to THZ. The overall conclusion from ALLHAT was that thiazide diuretics are the preferred first-line therapy for hypertension [30].
Patient Summary
Background
High blood pressure (hypertension) is common and increases the chances of developing heart disease. Different types of drugs are available for treating blood pressure, and they have been compared in many different studies. When doctors recommend a particular drug for a particular patient, they base their decision on the latest studies, on whether the patient has any other medical problems that might affect drug choice, and on the price differences between different drugs. If there are two drugs that have been shown to work equally well for a particular patient, the doctor should prescribe the cheaper one.
Why Was This Study Done?
The researchers wanted to see whether doctors follow the latest guidelines when they prescribe drugs for treating high blood pressure (antihypertensive drugs).
What Did the Researchers Do?
They looked at which drugs doctors had given to people aged over 65 years with high blood pressure in British Columbia, Canada, from 1993 to 2000. They only looked at prescriptions for people who had never taken blood-pressure medication before, a total of 82,824 patients. The drug of choice for most of these patients is a thiazide. Thiazides belong to a class of drugs called diuretics, sometimes referred to as water pills. These drugs have been around for a long time and in most patients they are at least as effective as and much cheaper than any of the newer drugs.
What Did They Find?
The prescription of thiazides in these patients increased from 22% to 33% during the study period, but it was consistently lower than it should have been based on the existing evidence.
What Does This Mean?
Doctors in British Columbia do not prescribe thiazides for seniors with high blood pressure as much as they should. As a consequence, some patients don't get the best treatment and the costs are considerably higher than they need to be. The reasons are not clear, but the findings suggest that doctors might not be up-to-date with the latest study results and guidelines. There have been more studies since 2002 and more efforts to educate doctors, and it would be interesting to know whether the current prescription practices better match the latest results.
More Information Online
The website of the US National Heart, Lung, and Blood Institute (http://www.nhlbi.nih.gov) has information about antihypertensive drugs and their characteristics, including facts on a recent big study (called the ALLHAT study) that compared antihypertensive drugs: http://www.nhlbi.nih.gov/health/allhat/facts.htm
Information pages from the Blood Pressure Association, a UK-based charity: http://www.bpassoc.org.uk/information/information.htm
Medline Plus information pages on blood pressure: http://www.nlm.nih.gov/medlineplus/highbloodpressure.html
This research is supported in part by the Canadian Institutes of Health Research (CIHR). SM is currently supported by a Canadian Institutes of Health Research New Investigator and a Scholar Award from the Michael Smith Foundation for Health Research. The Centre for Health Services and Policy Research and the University of British Columbia Therapeutics Initiative are funded in part by the British Columbia Ministry of Health Services. Funding agencies were not involved in the study design, analysis, decision to publish, or preparation of the manuscript. The views presented here are solely those of the authors.
Citation: Morgan S, Bassett KL, Wright JM, Yan L (2005) First-line first? Trends in thiazide prescribing for hypertensive seniors. PLoS Med 2(4): e80.
Abbreviations
ACEangiotensin-converting enzyme
MImyocardial infarction
THZthiazide-like diuretic
==== Refs
References
IMS 2002 Canadian pharmaceutical industry review: Diagnoses and therapies. Available: http://www.imshealthcanada.com/htmen/pdf/IMSR02_T31–32.pdf
2003 Accessed 28 January 2004
Morgan SG Drug spending in Canada: Recent trends and causes Medical Care 2004 42 635 642 15213487
Therapeutics Initiative Therapeutics Letter 47: The Answer: Thiazides first-line for hypertension. Available: http://www.ti.ubc.ca/pages/letter39.htm
2003 Accessed 22 March 2004
Therapeutics Initiative Drugs of choice in the treatment of hypertension (1). Available: http://www.ti.ubc.ca/pages/letter7.html
1995 Accessed 14 Aug 2003
Wright JM Lee CH Chambers GK Systematic review of antihypertensive therapies: does the evidence assist in choosing a first-line drug? CMAJ 1999 161 25 32 10420860
Wright JM Choosing a first-line drug in the management of elevated blood pressure: What is the evidence? 1: Thiazide diuretics CMAJ 2000 163 57 60 10920733
Morgan SG Barer ML Agnew JD Whither seniors' PharmaCare: Lessons from (and for) Canada Health Aff (Millwood) 2003 22 49 59 12757272
Johns Hopkins University ACG Software Documentation and Users Manual 2001 Baltimore Johns Hopkins University Press 294
Maclure M Dormuth C Naumann T McCormack J Rangno R Influences of educational interventions and adverse news about calcium-channel blockers on first-line prescribing of antihypertensive drugs to elderly people in British Columbia Lancet 1998 352 943 948 9752816
Schneeweiss S Soumerai SB Maclure M Dormuth C Walker AM Clinical and economic consequences of reference pricing for dihydropyridine calcium channel blockers Clin Pharmacol Ther 2003 74 388 400 14534526
Schneeweiss S Dormuth C Grootendorst P Soumerai SB Maclure M Net health plan savings from reference pricing for angiotensin-converting enzyme inhibitors in elderly British Columbia residents Med Care 2004 42 653 660 15213490
Robinson JR Young TK Roos LL Gelskey DE Estimating the burden of disease. Comparing administrative data and self-reports Med Care 1997 35 932 947 9298082
Muhajarine N Mustard C Roos LL Young TK Gelskey DE Comparison of survey and physician claims data for detecting hypertension J Clin Epidemiol 1997 50 711 718 9250269
Therapeutics Initiative Drugs of choice in the treatment of hypertension (2). Available: http://www.ti.ubc.ca/pages/letter8.html
1995 Accessed 14 August 2003
Wright JM Choosing a first-line drug in the management of elevated blood pressure: what is the evidence? 3: Angiotensin-converting-enzyme inhibitors CMAJ 2000 163 293 296 10951729
Wright JM Choosing a first-line drug in the management of elevated blood pressure: what is the evidence? 2: Beta-blockers CMAJ 2000 163 188 192 10934984
SHEP Cooperative Research Group Prevention of stroke by antihypertensive drug treatment in older persons with isolated systolic hypertension. Final results of the Systolic Hypertension in the Elderly Program (SHEP) JAMA 1991 265 3255 3264 2046107
Logan AG Report of the Canadian Hypertension Society's consensus conference on the management of mild hypertension CMAJ 1984 131 1053 1057 6149806
Ogilvie RI Burgess ED Cusson JR Feldman RD Leiter LA Report of the Canadian Hypertension Society Consensus Conference: 3. Pharmacologic treatment of essential hypertension CMAJ 1993 149 575 584 8364814
Myers MG Carruthers SG Leenen FH Haynes RB Recommendations from the Canadian Hypertension Society Consensus Conference on the Pharmacologic Treatment of Hypertension CMAJ 1989 140 1141 1146 2565758
Reeves RA Fodor JG Gryfe CI Patterson C Spence JD Report of the Canadian Hypertension Society Consensus Conference: 4. Hypertension in the elderly CMAJ 1993 149 815 820 8374844
Joint National Committee The 1980 report of the Joint National Committee on detection, evaluation, and treatment of high blood pressure Arch Intern Med 1980 140 1280 1285 6775608
Joint National Committee The 1984 Report of the Joint National Committee on detection, evaluation, and treatment of high blood pressure Arch Intern Med 1984 144 1045 1057 6143542
Joint National Committee The 1988 report of the Joint National Committee on detection, evaluation, and treatment of high blood pressure Arch Intern Med 1988 148 1023 1038 3365073
Joint National Committee The fifth report of the Joint National Committee on detection, evaluation, and treatment of high blood pressure (JNC V) Arch Intern Med 1993 153 154 183 8422206
Joint National Committee The sixth report of the Joint National Committee on prevention, detection, evaluation, and treatment of high blood pressure Arch Intern Med 1997 157 2413 2446 9385294
Chobanian AV Bakris GL Black HR Cushman WC Green LA The seventh report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure: The JNC 7 report JAMA 2003 289 2560 2572 12748199
McAlister FA Levine M Zarnke KB Campbell N Lewanczuk R The 2000 Canadian recommendations for the management of hypertension: Part one–therapy Can J Cardiol 2001 17 543 559 11381277
Curb JD Pressel SL Cutler JA Savage PJ Applegate WB Effect of diuretic-based antihypertensive treatment on cardiovascular disease risk in older diabetic patients with isolated systolic hypertension. Systolic Hypertension in the Elderly Program Cooperative Research Group JAMA 1996 276 1886 1892 8968014
Major outcomes in high-risk hypertensive patients randomized to angiotensin-converting enzyme inhibitor or calcium channel blocker vs diuretic: The Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT) JAMA 2002 288 2981 2997 12479763
Morgan SG Agnew JD Barer ML Seniors' prescription drug cost inflation and cost containment: Evidence from British Columbia Health Policy 2004 68 299 307 15113641
Dormuth CR Maclure M Bassett K Jauca C Whiteside C Effect of periodic letters on evidence-based drug therapy on prescribing behaviour: A randomized trial CMAJ 2004 171 1057 1061 15505268
Randomized double-blind comparison of a calcium antagonist and a diuretic in elderly hypertensives. National Intervention Cooperative Study in Elderly Hypertensives Study Group Hypertension 1999 34 1129 1133 10567194
Brown MJ Palmer CR Castaigne A de Leeuw PW Mancia G Morbidity and mortality in patients randomised to double-blind treatment with a long-acting calcium-channel blocker or diuretic in the International Nifedipine GITS study: Intervention as a Goal in Hypertension Treatment (INSIGHT) Lancet 2000 356 366 372 10972368
Schneeweiss S Maclure M Dormuth C Avorn J Pharmaceutical cost containment with reference-based pricing: Time for refinements CMAJ 2002 167 1250 1251 12451080
The ALLHAT Officers and Coordinators for the ALLHAT Collaborative Research Group Major cardiovascular events in hypertensive patients randomized to doxazosin vs chlorthalidone. The antihypertensive and lipid-lowering treatment to prevent heart attack trial (ALLHAT) JAMA 2000 283 1967 1975 10789664
| 15839739 | PMC1087212 | CC BY | 2021-01-05 10:39:49 | no | PLoS Med. 2005 Apr 26; 2(4):e80 | utf-8 | PLoS Med | 2,005 | 10.1371/journal.pmed.0020080 | oa_comm |
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 1598651210.1371/journal.pmed.0020099Student ForumOtherMedical EducationMedical EducationMedical journalsInternational healthMedicine in Developing CountriesHearing the Voice of Medical Students Worldwide Student ForumPalmer Brian A *Wong Amanda Singla Mohit Brian A. Palmer is the national president of the American Medical Student Association (AMSA), Reston, Virginia, United States of America. Amanda Wong is a past president of the European Medical Students' Association (EMSA), Brussels, Belgium. Mohit Singla is the caretaker president of the Indian Medical Students' Organization (IMSO), Hoshiarpur, India, and is the treasurer of the International Federation of Medical Students' Associations (IFMSA), Ferney-Voltaire, France.
Competing Interests: The authors declare that they have no competing interests.
*To whom correspondence should be addressed. E-mail: [email protected] 2005 26 4 2005 2 4 e99Copyright: © 2005 Palmer 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 Student Forum, a new section of PLoS Medicine, is a space where medical students from across the world can exchange ideas about the critical issues affecting health and health care from their unique perspective
==== Body
A World of Contrasts
Medical students today enter a profession defined by stark contrasts. In developed countries, public health improvements have nearly doubled life expectancy in the last century, and sophisticated technology and innovative research hold the promise of longer and higher quality life. At the same time, life-threatening infectious diseases such as tuberculosis and malaria continue to affect billions of people, and a rapidly escalating HIV/AIDS pandemic now kills more than 8,300 people each day [1]. New drugs are being developed at a record pace, but only about 1% of drugs that reach the market are aimed at treating neglected diseases, such as kala-azar, Chagas' disease, and sleeping sickness, even though these account for over 10% of the global disease burden [2]. Both within and between countries, the gap between wealthy and poor has continued to grow, leading to widening disparities in health outcomes.
Despite these contrasts, physicians-in-training around the globe—in rich and poor countries—have a common goal: we seek the skills and knowledge to improve the health of individuals and populations as we devote ourselves to a career of service. The transformative quest to become a physician imbues the learner with insights from within and outside the profession of medicine. As medical students, we hold ourselves accountable to our patients, and we critically evaluate the state of our profession as we see it with new eyes.
This issue of PLoS Medicine heralds the start of the Student Forum section, a space where medical students from across the world can exchange ideas about the critical issues affecting health and health care from their unique perspective. It is appropriate, then, in this inaugural Student Forum, that we consider some of the key issues of information exchange in the context of global disease and medical education. How do we access information? How is that information shaped and influenced? How can we shape the content of the debate?
Open Access to the Medical Literature: Critical for Our Future
As physicians-in-training, we appreciate that the medicine we will practice will differ from that practiced by our predecessors. Mastery of access to information now trumps mastery of facts, and we enter an era of medical practice that will develop synergistically with biomedical research. Our generation of physicians will face the daunting task of translating an exploding body of biomedical and clinical research into medical practice. This task increasingly relies not only on the ability to access all the available medical literature, but on the ability to mine and synthesize it. PLoS represents one of a number of possible approaches toward ensuring free and open access to the world's literature [3], and students should be at the forefront of this movement.
As the democratization of access brings developed and developing countries together, the very content of the research published should follow suit. Too often the thoughts and opinions of those in the rich world have dominated research agendas and the content of medical journals. For example, less than 8% of articles published in the six leading tropical medicine journals in 2000–2002 were generated exclusively by scientists from developing countries [4]. Bringing the entire world of learners into a global conversation should be accompanied by a shift in editorial priorities to ensure that medical journals inform practice in a way that offers the greatest benefits to those suffering the greatest burdens.
The Student Forum: A space where medical students across the globe can exchange ideas
The Influence of the Pharmaceutical Industry
Journal editors and publishers pride themselves on publishing high-quality research subjected to rigorous peer review. Indeed, this model has contributed to significant advances in the standards for clinical investigation and rightly drives much of clinical practice. However, without the least bit of acknowledged irony, medical journals have begun publishing articles about how pharmaceutical industry advertising and marketing drives clinical practice in nonrational ways while at the same time publishing the very advertisements they critique.
The American Medical Student Association (AMSA), which has entered into a formal partnership with PLoS Medicine, accepts no pharmaceutical industry funding and encourages its members to eschew the industry largess through its “PharmFree” initiative (www.amsa.org/prof/pharmfree.cfm). Similar initiatives are afoot in medical student organizations around the world as we consider appropriate relationships with industry. Surely it is time that journals examine the content of their pages and stop serving as vehicles through which physician practice is influenced by industry. PLoS Medicine is the first major international general medical journal to refuse to advertise medical drugs or devices [5]. We hope that other major journals will follow suit.
The Need for Student Ideals and Ideas
Throughout the world, students represent the leading edge of social change. Indeed Paulo Freire, one of the most influential educational thinkers of the late 20th century, describes education as the “practice of freedom,” as learners continuously work to integrate action and reflection in a spiral of deepening connections. This process of integration is rooted in a deep-seated longing for the possible, informed by the lived experience of the painful. Nowhere is this more true than in medicine, where the suffering of our patients underpins our scientific inquiry, our clinical development, and our personal growth as healers.
As we students learn medicine, we begin to ask hard questions about the proximal—and the distal—causes of our patients' suffering. The reductionistic logic of medical education offers only false comfort, and only to those who content themselves with pathways and receptors. Learners want to understand what led to the derangement of the pathway, of the receptor. Genes? Environmental factors? Economics? In asking, we bring a sense of possibility, of hope, to the process. What could be? What tools do we need to address these broader causes? And we hope that thinking about these questions from a student's view may offer our senior colleagues a chance to reflect on their own work as physicians and their own place in our world.
We hope this space emerges as a vibrant source of student energy and passion, grounded in evidence, accountable to the patients who will entrust their care to us, and reflecting the hope we have for the future of medicine.
Toward the Future
As we come together from three continents to inaugurate the Student Forum in PLoS Medicine, we are committed to ensuring that this section of the journal remains a source of vigorous debate and thoughtful analysis. The section will initially be published quarterly, and will contain essays of up to 1,000 words that address any health-related issue from a medical student perspective (see Box 1).
Box 1. Contributing to the Student Forum
PLoS Medicine welcomes essays of up to 1,000 words, which have not been published elsewhere, from medical students for the Student Forum.
Before submitting an essay, please send a pre-submission enquiry to E-mail: [email protected], explaining in 100 words what the essay will be about.
The journal is also keen to forge alliances with medical student associations in Africa, Latin America, and Australia, and to elect a representative from each of these associations to join the PLoS Medicine student advisory group. Interested associations should contact the PLoS Medicine magazine editor (E-mail: [email protected]).
The PLoS Medicine medical student advisory group, which currently has representatives from the American Medical Student Association, the Indian Medical Students' Organization, and the European Medical Students' Association, will be involved in selecting and editing essays for publication.
The Student Forum will be freely available for all to read and critique, free from inappropriate financial interests, and committed to the drive toward justice that student ideals embody. We look forward to helping shape this forum in the best traditions of free exchange, and we look forward to collaboration with our senior colleagues to hold us to these standards.
Citation: Palmer BA, Wong A, Singla M (2005) Hearing the voice of medical students worldwide. PLoS Med 2(4): e99.
==== Refs
References
Avert.org World HIV and AIDS statistics 2005 Horsham (United Kingdom) Avert. org Available: http://www.avert.org/worldstats.htm . Accessed 5 February 2005
Trouiller P Olliaro P Torreele E Orbinski J Laing R Drug development for neglected diseases: A deficient market and a public-health policy failure Lancet 2002 359 2188 2194 12090998
Engber D Gass A Yamey G Who should own medical knowledge? Student BMJ 2004 12 310 Available: http://www.studentbmj.com/issues/0904/editorials/310.html . Accessed 5 February 2005
Obuaya C Reporting of research and health issues relevant to resource poor countries in high impact medical journals Eur Sci Editing 2002 28 72 77
The PLoS Medicine Editors Prescription for a healthy journal PLoS Med 2004 1 e22 17523248
| 15986512 | PMC1087213 | CC BY | 2021-01-05 10:39:46 | no | PLoS Med. 2005 Apr 26; 2(4):e99 | utf-8 | PLoS Med | 2,005 | 10.1371/journal.pmed.0020099 | oa_comm |
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 1583975010.1371/journal.pmed.0020105PerspectivesInfectious DiseasesHealth PolicyMalariaMedicine in Developing CountriesHealth PolicyInternational healthThe Benefits of Artemisinin Combination Therapy for Malaria Extend Beyond the Individual Patient PerspectivesGarner Paul *Graves Patricia M Paul Garner is Professor of Community Health, Liverpool School of Tropical Medicine, Liverpool, United Kingdom. Patricia M. Graves is Guest Researcher, Division of Parasitic Diseases, Centers for Disease Control and Prevention, Atlanta, Georgia, United States of America.
Competing Interests: The authors declare that they have no competing interests.
*To whom correspondence should be addressed. E-mail: [email protected] 2005 26 4 2005 2 4 e105Copyright: © 2005 Garner and Graves.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.
Reduction of Malaria Transmission to Anopheles Mosquitoes with a Six-Dose Regimen of Co-Artemether
Garner and Graves discuss the implications of a new study in PLoS Medicine that found that artemisinin combination treatment reduces infectiousness
==== Body
The traditional, low-cost mainstay drugs for malaria, chloroquine (CQ) and sulphadoxine-pyrimethamine (SP), have a very limited lifetime left in terms of their clinical usefulness. These drugs have been relatively ineffective in Asia for two decades, and rising drug resistance levels have now also rendered them ineffective in many sub-Saharan African countries [1]. Artemisinin drugs, such as artesunate and artemether, derived from the Chinese herb Artemisia annua, are rapidly being adopted as standard treatments in Africa, in the hope that effective treatment will assist in reversing the apparently increasing death rates in African children [2].
The Impact of Artemisinin Derivatives on Gametocytes
In the most severe form of human malaria, Plasmodium falciparum, adequate treatment of a malaria attack does not necessarily prevent the infected person from transmitting the disease to others. Mosquitoes become infected by ingesting mature sexual stages, known as gametocytes, which take around ten days to develop. Mature gametocytes present at the time of treatment may be unaffected, and parasites already committed to gametogenesis at the time of treatment will continue their development. Therefore, infectious gametocytes may be present in the blood for many days after the patient has been treated and feels better. Gametocytes responsible for such “post-treatment transmission” are more likely to carry and spread drug-resistant alleles [3].
ACT reduces infectiousness to mosquitoes
(Photo: Centers for Disease Control and Prevention/ James D. Gathany)
Trials showed that adding three days of the artemisinin derivative artesunate to existing antimalarial treatments led to a marked reduction in the number of gametocytes in the month following the treatment [4]. Whether this translates to an impact on transmission is less clear: in an area of low transmission in western Thailand, the introduction of artesunate in 1994 coincided with a sustained decrease in incidence of malaria [5]. However, infectiousness of patients to mosquitoes in this study was not measured directly, and the association of drug introduction and incidence reduction may have been coincidental.
A randomized controlled study by Sutherland and colleagues in this month's PLoS Medicine [6] shows conclusively that a six-dose course of artemether-lumefantrine (co-artemether) given to children with P. falciparum malaria in Gambia reduces gametocyte prevalence, duration of gametocyte carriage, and infectiousness to mosquitoes, compared to dual treatment with CQ and SP (CQ/SP). Presumably owing to lower gametocyte density, none of the artemether-lumefantrine-treated gametocyte carriers were infectious on day 7 after treatment, compared to 37% of the CQ/SP-treated children. The results also show that artemether-lumefantrine has a specific action against developing gametocytes in addition to its action on asexual stages. This effect of artemether-lumefantrine gives it a significant advantage over CQ/SP, although in terms of parasitological and clinical failure no difference was found between the two regimens.
Scaling Up Artemisinin-Based Combination Treatment
Whilst there is little argument that artemisinin-based combination treatments (ACTs) such as artemether-lumefantrine are effective in treating malaria, they are a lot more expensive than the treatments currently being used. The advocates of policy change are lobbying hard for the global financing mechanisms that will likely be necessary in order to subsidize the world's supply of ACTs [7]. The fervor of the advocates is intense [8], but national policies—particularly the introduction of interventions that are in short supply and cost more to produce—are unlikely to be implemented at the speed that a clinician treating an individual patient would desire. It is more likely that countries will explore the variety of strategic options open to them, which include ACTs as first-line treatment, as well as non-artemisinin combination treatments, such as amodiaquine with SP, and will examine the potential benefits and costs [9].
With ACTs, the potential public health impact of reducing transmission is a factor to include in the evaluation of benefits versus costs, but there is little research on this impact in areas that are highly endemic for malaria. The finding of Sutherland and colleagues that ACTs also greatly reduce infectiousness will contribute to appropriate and informed decision-making for sustainable changes in treatment policy at the country level.
Citation: Garner P, Graves PM (2005) The benefits of artemisinin combination therapy for malaria extend beyond the individual patient. PLoS Med 2(4): e105.
Abbreviations
ACTartemisinin-based combination treatment
CQchloroquine
SPsulphadoxine-pyrimethamine
==== Refs
References
White NJ Nosten F Looareesuwan S Watkins WM Marsh K Averting a malaria disaster Lancet 1999 353 1965 1967 10371589
Koerenromp EL Williams BG Gouws E Dye C Snow RW Measurement of trends in childhood malaria mortality in Africa: An assessment of progress toward targets based on verbal autopsy Lancet Infect Dis 2003 3 349 358 12781507
Sutherland CJ Alloueche A Curtis J Drakeley CJ Ord R Gambian children successfully treated with chloroquine can harbor and transmit Plasmodium falciparum gametocytes carrying resistance genes Am J Trop Med Hyg 2002 67 578 585 12518847
International Artemisinin Study Group Artesunate combination for treatment of malaria: Meta-analysis Lancet 2004 363 9 17 14723987
Nosten F van Vugt M Price R Luxemburger C Thway KL Effects of artesunate-mefloquine combination on incidence of Plasmodium falciparum malaria and mefloquine resistance in western Thailand: A prospective study Lancet 2002 356 297 302
Sutherland CJ Ord R Dunyo S Jawara M Drakeley CJ Reduction of malaria transmission to Anopheles mosquitoes with a six-dose regimen of co-artemether PLoS Med 2005 2 e92 15839740
Board on Global Health Committee on the Economics of Antimalarial Drugs Arrow KJ Panosian CB Gelband H Saving lives, buying time: Economics of malaria drugs in an age of resistance 2004 Washington (DC) Institute of Medicine of the National Academies Available: http://www.nap.edu/books/0309092183/html/ . Accessed 3 March 2005
Attaran A Rescuing malaria treatment, or not? Lancet 2004 364 1922 1923 15566995
Duffy PE Mutabingwa TK Drug combinations for malaria: Time to ACT? Lancet 2004 363 3 4 14723982
| 15839750 | PMC1087214 | CC BY | 2021-01-05 10:39:45 | no | PLoS Med. 2005 Apr 26; 2(4):e105 | utf-8 | PLoS Med | 2,005 | 10.1371/journal.pmed.0020105 | oa_comm |
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 1583972810.1371/journal.pmed.0020111EssayOtherMedical EducationAcademic MedicineMedical EducationThe Problem with Academic Medicine: Engineering Our Way into and out of the Mess EssayRees Jonathan Jonathan Rees is a professor in the Department of Dermatology at the University of Edinburgh, United Kingdom. E-mail: [email protected]
Competing Interests: JR is paid by the University of Edinburgh, and currently receives grant support from the Wellcome Trust and other sources. In the past he has received payments from companies for travel, consulting, and working as an expert on a submission to the United Kingdom Committee on Safety of Medicines and the United States Food and Drug Administration.
4 2005 26 4 2005 2 4 e111Copyright: © 2005 Jonathan Rees.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.Medical schools have come to resemble schools of molecular biology, while medical research has become focused on molecules rather than patients
==== Body
“My advice is to go for the messes—that's where the action is.”—Steven Weinberg [1]
Academic medicine has tried to avoid the mess that has characterised much of clinical practice over the last quarter century. Instead its success has come from using techniques developed in biochemistry and molecular genetics, and applying them to disease. It has been a highly productive approach. For example, identification of the clinical syndrome AIDS, isolation of the causative agent, and development of diagnostic testing and effective therapy all happened in the space of a few years. Alongside such successes has come a redefinition of academic medicine. Medical research has become biomedical research; discovery flows from bench to bedside and if it stutters, an injection of funds for more effective translation is required [2]. A plethora of organisms are sequenced, and all the genes in the mouse are to be “knocked out” [3]—all to serve the goal of improving human health, so we are told.
Even the idea of clinical disciplines appears slightly passé, their traditional names a throwback to a less scientific age—dermatology is now dermatological science, and neurology and psychiatry are now clinical neuroscience [4]. And if you are tempted at the end of the clinic—before you return to the sanctuary of the lab—to wonder about the mess of clinical medicine, to ask how it all fits together, and to question the real outcomes of what you do, you run the risk of being put down. The put down goes like this: we do science; we do not question whether “Coke is better than Pepsi or approach A is better than approach B” [5]; we leave such choices to the consumer and the HMOs to decide upon. Back to the mice, and future tenure.
Clinical Medicine: The Missing Subject
A striking aspect of the modern medical research university is the relative absence of clinical medicine from its portfolio of activities. Just look at what our successors are being encouraged to study. Most students will spend three or more years studying subjects such as cell biology, genetics, stem cell biology, or biochemistry, with a small remnant learning some clinical epidemiology and trials methodology. Now take yourself back to the clinic, spend a week seeing patients, and ask what academic establishment is needed to improve health care? Well, of course, some cell biology and the like are needed, but I suggest such disciplines should account for only 20% of our effort.
A striking aspect of the modern medical research university is the relative absence of clinical medicine from its portfolio of activities
(Illustration: Sapna Khandwala, Public Library of Science)
The rest of our effort should be directed towards operations research; the basis of expertise and decision-making processes; informatics, from the mundane issue of paper versus digital clinical records to the representation of clinical information in ways that are helpful for both doctor and patient; health-care policy, determining why effective therapies are so often not available, and why some diseases are researched more than others; and psychology, determining why patients' worldviews are so often different from ours and whether cognitive limitations (by either health professionals or patients) can account for some of these differences. Underpinning all of this should be the basic medical sciences: economics, statistics, and physiology.
Academic Kudos Versus Improving Health
So why are our institutions not fit for the purpose of improving patients' health? Herbert Simon, the polymath and Nobel laureate in economics, observed many years ago that medical schools resembled schools of molecular biology rather than of medicine [6]. He drew parallels with what had happened to business schools. The art and science of design, be it of companies or health care, or even the type of design that we call engineering, lost out to the kudos of pure science. Producing an economics paper densely laden with mathematical symbols, with its patently mistaken assumptions about rational man, was a more secure way to gain tenure than studying the mess of how real people make decisions.
The same problem also applies to medicine. Flushed with success from the golden age of medical discovery, a marriage was brokered between a simple and mistaken model of medical discovery in which there was a unidirectional flow of information from bench to clinic, and the safety and control afforded by the technical facility of modern biology. Once in place, the forces needed to sustain this ill-judged union were not difficult to recognise.
Most medical researchers rarely, if ever, see patients. Most who argue for the necessity of the pyramid of discovery, with biochemistry and genetics at the base and clinical research at the apex, work at the base themselves [2]. However, many of the major discoveries that have had a direct impact on clinical practice arose from clinical disciplines rather than from generic biomedical approaches—consider hip replacements, cataract surgery, the importance of Helicobacter pylori, phototherapy, in vitro fertilisation, and minimally invasive surgery [4]. In the biomedical model, these successes are brushed aside as being of historical interest only. From finding genes to gene therapy and stem cell therapy, both the public and research community itself are fed a biased view of medical advances. Cancer cured (in mice) again.
Medicine: A Science of the Artificial
A number of arguments suggest that the conventional wisdom that has underpinned the institutions of academic medicine is breaking down. First, the pace of major clinical advance, despite increases in funding, is slowing [2]. Second, there is increased external scrutiny (from many sources) of medicine, not least because of the large increases in the cost of health care. Third, and a useful indicator that something is amiss, a career in academic medicine is increasingly unattractive—the new generation of students clearly see the academic towers as ill suited to solving the problems that they experience in the clinic.
The Two Epistemologies That Underpin Clinical Advance
There have been two intellectual movements in the last half century influencing medical practice. The first—the enormous evolution of technological facility in biochemistry and genetics—needs little comment as its influence is all around. The second, still inchoate, is the attempt to elucidate the laws that govern clinical practice. It is to the great credit of the clinical epidemiology movement that it focussed attention once again on how medicine was practiced. No longer was medicine just the application of theories developed in the laboratory, but there was the glimmering of an intellectually coherent discipline centred on diagnosis and intervention. When seen alongside the now classical work of Wennberg, showing the remarkable variation in the delivery of health care across populations [7], medicine could be seen in a new light. No longer was medicine a mere application of “basic” science, of lessons from physiology, but, instead, medicine was an artefact of engineering—health and medical care was something we build, something we design. Just as informatics and computing are not solely about building faster chips, but about how information is ordered and represented, and how machines and humans interact; the same is true for medicine. To use Herbert Simon's phrase, medicine is a “science of the artificial”, artificial as in artificial intelligence: an artefact or something created for a purpose [6].
The Codification of Medicine: Bringing Order to the Mess
The principal intellectual question of the next quarter century for academic medicine is to what extent medicine is capable of codification. How and to what degree can we codify—produce rules—of how decisions in medicine should be made. The march of industrial progress for the last 200 years has been the replacement of implicit knowledge and experience by formal teaching, rule-based behaviour, specialisation, and the division of labour. How will the mess that is clinical practice be ordered by this new epistemology?
How do the results of clinical trials inform the therapy of an individual patient? Despite the views of many, how to interpret numerical data as evidence to guide action in a particular case has troubled the greatest mathematicians for centuries [8]. How do we assimilate information of different kinds? What is the relative proportion of local factors versus summary measures in the treatment of a patient? The answers are not clear. Trials are expensive, infrequent, and remote from much of clinical care. Often the debate seems to resemble some medieval battle, with artillery and army being towed into place to siege a recalcitrant city—think, for example, of the intense debate surrounding the value of mammography [9]. By contrast, clinical practice seems to resemble more guerrilla warfare, where the lie of the local terrain trumps any fickle general predictions. The set of statistical techniques developed for use in agriculture by the great R. A. Fisher and others hardly seems appropriate for the questions we wish to address [10]. Populist ideas about evidence, and the weight of evidence—the pseudoscience that adorns so many guidelines—is almost too embarrassing to mention.
I mentioned the issue of cost with good reason. We have gotten used to pretending that cost is not an issue when new treatments are being researched. Medicine is as much engineering as science, and what needs to be rewarded in medical research is the ability to innovate within a financial envelope. Being able to invent solutions at an affordable price is a design constraint: one that needs to be seen as a challenge. The solution “at any price” is lazy: cheap PCs are what made the World Wide Web so revolutionary.
Conclusion: Coke Versus Pepsi Revisited
Codification reminds us that medicine makes heavy demands on its practitioners, demands that in all too many cases can be usefully mitigated with appropriate information systems. Here the question is how we can design systems that record clinical information in a way that facilitates direct clinical care, and feeds back and informs clinical discovery. The danger is to believe that such an advance is not an intellectual problem, but just a mere choice of which brand of software to use for clinical records. There is no greater example of academic medicine's hubris than imagining that designing medical practice is like choosing between “Coke or Pepsi” [5]. It is like saying Tim Berners Lee's invention of the World Wide Web was on a par with inventing another typeface.
This is the first in a series of articles in PLoS Medicine looking critically at the future of academic medicine.
Citation: Rees J (2005) The problem with academic medicine: Engineering our way into and out of the mess. PLoS Med 2(4): e111.
The author is on the editorial board of PLoS Medicine.
==== Refs
References
Weinberg S Scientist: Four golden lessons Nature 2003 426 389 14647361
Rees J The fundamentals of clinical discovery Perspect Biol Med 2004 47 597 607 15467181
Austin CP Battey JF Bradley A Bucan M Capecchi M The knockout mouse project Nat Genet 2004 36 921 924 15340423
Rees J Complex disease and the new clinical sciences Science 2002 296 698 700 11976444
Mullan F Twenty-seven fingers without a palm is not a hand: A conversation with Elias Zerhouni 2004 Health Aff Millwood Suppl Web Exclusives W4.1 W4.6
Simon HA The sciences of the artificial 1969 Cambridge (Massachusetts) MIT Press 215
Wennberg J Gittelsohn Small area variations in health care delivery Science 1973 182 1102 1108 4750608
Gigerenzer G Swijtink Z Porter T Daston L Beatty J The empire of chance: How probability changed science and everyday life 1989 Cambridge Cambridge University Press 358
Goodman SN The mammography dilemma: A crisis for evidence-based medicine? Ann Intern Med 2002 137 363 365 12204023
Royall R Statistical evidence: A likelihood paradigm 1997 Boca Raton (Florida) Chapman & Hall/CRC 191
| 15839728 | PMC1087215 | CC BY | 2021-01-05 10:39:46 | no | PLoS Med. 2005 Apr 26; 2(4):e111 | utf-8 | PLoS Med | 2,005 | 10.1371/journal.pmed.0020111 | oa_comm |
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 10.1371/journal.pmed.0020115SynopsisDiabetes/Endocrinology/MetabolismDiabetesEndocrinologyProducing Insulin from Neural Cells Synopsis4 2005 26 4 2005 2 4 e115Copyright: © 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.
Differentiation of Insulin-Producing Cells from Human Neural Progenitor Cells
==== Body
Before insulin was discovered and purified, doctors could only watch as their patients slowly died of type 1 diabetes mellitus. In January 1922, the prognosis was changed dramatically when a teenager with diabetes in a Toronto hospital became the first recorded recipient of an injection of insulin. Although this preparation was far from perfect, as Frederick Banting said in his Nobel lecture of 1925 (Nobel laureates did not have to wait so long as they do now for recognition), “There was a marked reduction in blood sugar and the urine was rendered sugar free.” Suddenly diabetes was a potentially curable disease. But insulin's very success brought trouble, as demand far exceeded the supply.
Now recombinant insulin is available in plentiful supply, but the control of type 1 diabetes remains far from perfect, and researchers have looked towards an ideal of transplanting insulin-producing cells instead. The best protocol for this, also pioneered in Canada, the Edmonton protocol, uses ß-cells isolated from human cadavers and has shown some remarkable results with around 84% of patients remaining insulin-free after one year and 89% of patients still producing insulin after three years. However, the isolation of ß-cells is laborious and limited by donor availability.
Insulin-producing neurospheres
The next logical step, then, is to look for renewable sources of insulin-producing cells. The emerging science of human stem cell research makes this step possible, and in a paper in this month's PLoS Medicine Seung Kim and colleagues from Stanford University suggest that researchers should look beyond just pancreatic precursors and embryonic stem cells to other cell-type precursors for ideas about ß-cell replacement.
The rationale for the approach in this study comes from observations that although the pancreatic islet cells are the principal source of insulin in humans, in some invertebrate species, such as Drosophila, most circulating insulin is produced by brain neurons. Intriguingly, the gene encoding insulin is also transcribed by some vertebrate neurons—although it is not clear whether they then produce or secrete insulin protein.
Kim and colleagues took human neural progenitor cells derived from brain, and exposed them to a series of signals that are known to drive pancreatic islet development. They were able to produce clusters of insulin-producing cells that were responsive to glucose in vitro. After transplantation into immunocompromised mice, circulating human insulin and C-peptide derived from the proinsulin precursor were detected when the mice were given glucose.
Of course, results in mice do not mean that such treatments would automatically work in humans, and before any such therapies become available there are many hurdles to overcome. Some of the most important include the long-term stability and safety of the cells (although the cells remained differentiated in the mice and did not form tumors, such a risk would need to be very thoroughly investigated because of the chronic nature of diabetes), and how to scale up such a process to produce the much larger numbers of cells needed for human treatment. Nonetheless, work like this study from Kim's group might point to where the future of diabetes treatment lies.
| 0 | PMC1087216 | CC0 | 2021-01-05 10:39:44 | no | PLoS Med. 2005 Apr 26; 2(4):e115 | utf-8 | PLoS Med | 2,005 | 10.1371/journal.pmed.0020115 | oa_comm |
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 1583974610.1371/journal.pmed.0020119Correspondence and Other CommunicationsImmunologyInfectious DiseasesMicrobiologyScience PolicyAllergy/ImmunologyEpidemiology/Public HealthHIV/AIDSSexual HealthGlobal healthHIV Infection/AIDSImmunology and allergyMedicine in Developing CountriesDeveloping an HIV Vaccine: The Role of Efficacy Studies in Nonhuman Primates CorrespondenceÜberla Klaus
1
1Ruhr-University BochumBochumGermanyE-mail: [email protected]
Competing Interests: I am actively engaged in HIV/AIDS vaccine development and receive research grants from the European Community, the German Research Foundation, and the Wilhelm-Sander Foundation.
4 2005 26 4 2005 2 4 e119Copyright: © 2005 Klaus Überla.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 Global HIV/AIDS Vaccine Enterprise: Scientific Strategic Plan
==== Body
Given the scientific hurdles encountered in HIV/AIDS vaccine development, a global effort is needed, and the scientific strategic plan proposed by the Global HIV/AIDS Vaccine Enterprise [1] is a milestone. The plan also provides an important starting point for continued discussion.
Since only a limited number of HIV/AIDS vaccines can be tested for efficacy in phase-III studies, evidence-based criteria for selection of candidate HIV/AIDS vaccines for these trials have to be defined. If an immune correlate of protection were available, small-scale immunogenicity studies in humans would provide required parameters. However, the specific immune responses needed for a successful HIV/AIDS vaccine remain unknown. The Global HIV/AIDS Vaccine Enterprise prioritizes research on vaccines eliciting neutralizing antibodies and T cell responses [1]. Standardization of laboratory assays measuring these parameters is proposed to allow comparison of different vaccines.
The main reasons to assume that T cell responses and neutralizing antibodies are important for HIV/AIDS vaccine efficacy are the following: (i) T cell responses and neutralizing antibodies are known to have a role in preventing infection or disease by other viruses, (ii) there is an inverse correlation of T cell responses with viral load in patients with HIV, (iii) depletion of CD8+ T cells in nonhuman primate models of AIDS increases viral load, and (iv) passive immunization can provide protection in some of the nonhuman primate models [2]. However, extensive studies during the last 15 years using various T cell assays have failed to provide an accepted immune correlate of protection in nonhuman primates. Although more sophisticated assays and larger groups of animals might reveal such a correlate, protection could also be mediated by a mechanism not yet defined and therefore not monitored.
While the Vaccine Enterprise proposes to search for correlates of immune protection in nonhuman primate models of AIDS, efficacy studies in these models are not mentioned [1]. The limitations of these models have been extensively discussed [3], and there is no proof that these models predict efficacy in humans. However, if one accepts that nonhuman primate models are valid for determination of immune correlates of protection, it seems reasonable to also assume that a more effective vaccine approach in an appropriate nonhuman primate model would also be more effective in humans. Thus, by comparing efficacies of different vaccines in these models, it should be possible to select the most promising vaccine approaches for clinical evaluation. In the past, the nonhuman primate models have been little informative with respect to relative vaccine efficacy: the results from different laboratories were difficult to compare because of differences in, among other things, the monkey species, the inoculation route and dose, the pathogenicity of the challenge virus, the homology between vaccine and challenge virus, and the read-out assays used. This problem was recognized almost ten years ago already, but no agreement has yet been reached on which models most closely resemble HIV transmission and infection in humans. Different aspects of vaccine research require different animal models. In addition, worldwide use of one or two selected models would lead to a shortage in the monkey species needed for these models. Therefore, it is unlikely that an agreement can be reached at all. Rather than going through great and costly efforts to standardize the animal models and laboratory assays, a standardized “state of the art” vaccine approach could be included as a control group in each study. The immunogenicity and efficacy of all novel vaccine candidates could then be determined relative to the immunogenicity and efficacy of the vaccine standard.
A number of issues would need to be discussed with respect to a generally acceptable vaccine standard. Ideally, results from a human efficacy study with the standardized vaccine approach should be available in near future. The standardized vaccine approach should be based on one of the most promising vaccines available at present. Due to the diversity of immunodeficiency viruses used in various nonhuman primate models of AIDS, the vaccine standard cannot be a single vaccine, but must be a standardized vaccine approach. For example, the vaccine approach could be defined by subcutaneous immunization of monkeys with a defined dose of a defined viral vector expressing codon-optimized gag and env genes at 24 and eight weeks before challenge. The degree of homology of the encoded Gag and Env proteins of the standard vaccine and the challenge virus should be the same as the homology between the antigens of the novel vaccine to be tested and those of the challenge virus. Thus, depending on the novel vaccines to be tested, different vaccine standards of the standardized vaccine approach are needed.
Including defined vaccine standards in future efficacy studies in nonhuman primate models would greatly facilitate preclinical evaluation of novel vaccine candidates and provide evidence-based criteria for their selection for clinical studies. Once an agreement on a standard vaccine approach has been reached, the approach could be implemented quickly, since efficacy studies in nonhuman primates are well established. Given the urgent need for an HIV/AIDS vaccine, we cannot afford to ignore the only animal models that might well predict efficacy in humans. Exploitation of the potential of carefully controlled comparative efficacy studies in nonhuman primates should be considered by the Global HIV/AIDS Vaccine Enterprise. It remains to be discussed whether inclusion of a vaccine standard in clinical studies might also solve some of the standardization problems encountered.
Citation: überla K (2005) Developing an HIV vaccine: The role of efficacy studies in nonhuman primates. PLoS Med 2(4): e119.
==== Refs
References
Coordinating Committee of the Global HIV/AIDS Vaccine Enterprise The Global HIV/AIDS Vaccine Enterprise: Scientific and strategic plan PLoS Med 2005 2 e25 10.1371/journal.pmed.0020025 15740411
Pantaleo G Koup RA Correlates of immune protection in HIV-1 infection: What we know, what we don't know, what we should know Nat Med 2004 10 806 810 15286782
Staprans SI Feinberg MB The roles of nonhuman primates in the preclinical evaluation of candidate AIDS vaccines Expert Rev Vaccines 2004 3 S5 S32 15285703
| 15839746 | PMC1087217 | CC BY | 2021-01-05 11:13:38 | no | PLoS Med. 2005 Apr 26; 2(4):e119 | utf-8 | PLoS Med | 2,005 | 10.1371/journal.pmed.0020119 | oa_comm |
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 1583974910.1371/journal.pmed.0020087Health in ActionDermatologyDermatologyMedical Informaticstelederm.org: Freely Available Online Consultations in Dermatology Health in ActionSoyer H. Peter *Hofmann-Wellenhof Rainer Massone Cesare Gabler Gerald Dong Huiting Ozdemir Fezal Argenziano Giuseppe H. Peter Soyer, Rainer Hofman-Wellenhof, and Cesare Massone are in the Department of Dermatology at the Medical University of Graz in Austria. Gerald Gabler is in the Department of IT and Telecommunications at Graz University Clinics and General Hospital in Austria. Huiting Dong is in the Department of Dermatology at the University of Zhengzhou in China. Fezal Ozdemir is in the Department of Dermatology at Ege University in Izmir, Turkey. Giuseppe Argenziano is in the Department of Dermatology at Second University of Naples in Italy.
*To whom correspondence should be addressed. E-mail: [email protected]
Competing Interests: HPS is Medical Director of the company E-derm-consult. The Web application on which telederm.org operates has been developed and provided by E-derm-consult. telederm.org is a nonprofit project. The other authors declare that they have no competing interests.
4 2005 26 4 2005 2 4 e87Copyright: © 2005 Soyer 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.Studies have shown that skin disorders can be reliably diagnosed by online consultations. These consultations are provided free of charge by a new nonprofit organization of skin doctors worldwide
==== Body
“e-Health matters. It can improve access to healthcare and boost the quality and effectiveness of the service offered. e-Health describes the application of information and communications technologies across the whole range of functions that affect the health sector.”
This is the introductory statement of a recent communication from the European Commission that produced the report “e-Health—Making Healthcare Better for European Citizens: An Action Plan for a European e-Health Area” [1]. Easy access to expert medical information and counseling independent of social, economic, ethnic, and regional factors is regarded as a major goal of medical policy today.
The visual nature of dermatology makes this discipline an obvious candidate for telemedicine techniques, and the feasibility and reliability of teledermatology is already well-established [2,3]. In recent years the teledermatology group of the Department of Dermatology at the Medical University of Graz, Austria, has been involved in several national and international teledermatology projects, with particular emphasis on clinical dermatology, clinicopathologic correlation, and evaluation of pigmented skin lesions [4,5]. Among the activities of the teledermatology group was the organization of the Third European Symposium on Teledermatology in Graz in November 2002 (http://telederm.uni-graz.at). Furthermore, together with colleagues from the Department of Dermatology in Jena, Germany, the Graz teledermatology group founded the International Society of Teledermatology (http://www.teledermatology-society.org), whose goal is to promote the worldwide exchange of knowledge and expertise in all aspects of dermatology.
telederm.org: The Project and the Application
telederm.org was initiated in 2002 by H. Peter Soyer, Rainer Hofmann-Wellenhof, and Gerald Gabler with the vision of providing and sharing user-generated dermatologic knowledge on a worldwide level. The basic goal of the project was to create a user-friendly platform for providing teleconsultation services and for discussing challenging and unusual cases in clinical dermatology, dermatopathology, and dermoscopy, with special emphasis on diagnosis, differential diagnosis, and treatment. This online dermatology community is moderated according to the general rules of online communities. (We have previously published a more detailed description of our first experiences with the telederm.org project [6].)
The current program (http://www.telederm.org) is available in English, German, Italian, Chinese, and Turkish. A secure connection is available and registration is required. Only physicians and health-care providers are able to subscribe; they can register as either clients or experts. Subscriptions are free, and are controlled and subsequently activated by the moderator. Each user is also given a personal username and password.
The main way users interact with the program is through submitting and responding to “requests”, in which a user solicits a consultation, either from a particular expert or from an online forum, on a particular patient. A maximum of three JPEG images can be uploaded for each request. Requests are integrated with a form for patient clinical data; patient-identifying data are neither required nor allowed. Requests can be sent directly to a special Web site section called “discussion view” that represents the online discussion forum. In this way, consultations are visible to all users, who can either read the consultations as bystanders (users who do not participate in normal forum discourse, but who watch the posted cases) or actively submit online opinions (active users). All users are free to enter the discussion and give their personal opinions; this forum is moderated, and the official language is English. The moderator of the forum currently is Cesare Massone, a colleague with a special background in dermatopathology and clinicopathologic correlation.
Almost 400 health-care professionals from 45 different countries subscribed to the services offered by telederm.org between its start in April 2002 and October 2004. Out of over 900 global consultations, 103 were proposed by users in the online discussion forum (Figures 1 and 2). The remaining consultations were direct, free consultations between clients and experts that were not presented to the discussion forum.
Figure 1 Discussion Forum: Image Viewer Allows Magnification of an Unusual Elbow Lesion
Figure 2 Overview of the Online Discussion following a User's Request Sent to the ForumEvery user can read the contribution of other colleagues and participate in the forum
The telederm.org content is determined by the free-access philosophy, which means that it is free to all users, who create the subject matter themselves. The unique aspect of the telederm.org philosophy is that every colleague can easily seek diagnostic advice for free from a pool of other colleagues willing to share their knowledge, some of whom may have more experience in a particular clinical setting. This structure encourages reciprocal exchange: simply by using the multihub facilities of the Internet, a colleague seeking help may later have the opportunity to provide expert advice. The role of the moderator is crucial to introduce and explain the concept of free-access teleconsultation to the users and to control the content of requests.
Are Web Consultations as Good as Face-to-Face Consultations?
When dealing with skin diseases the question arises whether a Web consultation is as good as a face-to-face consultation or whether there are some important limitations. Store-and-forward systems (SAFs) have been found to be accurate and effective, both when images are sent by E-mail and when they are sent through a Web-based system [7,8].
In 2004, Oztas and colleagues studied the reliability of teledermatology diagnoses made using a Web-based system [7]. Clinical photographs and information relating to 125 patients were placed on a Web server. Three dermatologists made the most likely diagnosis via a Web interface. Their diagnoses were compared with a reference diagnosis made in a face-to-face consultation with a fourth dermatologist; when appropriate the diagnosis was confirmed histopathologically. The teledermatologists were correct in 57% of cases when viewing the images alone, and their diagnostic accuracy improved to 70% when additional clinical information was available. The rate of agreement between the teledermatologists ranged from 44% to 70% (kappa = 0.22-0.32). When clinical information was provided, 77% of patients were correctly diagnosed by at least two dermatologists. The authors concluded that Web-based SAF teledermatology is feasible and is comparable with conventional SAF systems (i.e., E-mail systems) [7].
In 1999, Piccolo and colleagues showed that based on a study of 66 pigmented skin lesions the diagnostic concordance between face-to-face diagnosis and telediagnosis using conventional SAF systems was 91% [8]. In 2000, the same team in a second study on pigmented skin lesions showed that the level of experience of teledermatologists represented a major limitation of the SAF approach [5].
The well-known technical limitations of teledermatology, such as image quality and resolution, nowadays have been basically solved. More basic drawbacks such as the use of inadequate software and hardware certainly depend on the local situation and may differ markedly from country to country. In our experience, however, the most important factor in the accuracy of a given teleconsultation remains the intellectual human component. The telederm.org approach has been designed to discuss cases frankly in an online discussion forum using the asynchronous SAF system in order to overcome “areas of weakness” of a single observer and to receive different diagnostic and therapeutic thoughts from several expert Web viewers on a given case.
The Future of the Service
Nearly 800 direct consultations have been answered and more than 100 cases have been posted in the discussion view. Moreover, the worldwide community of users is steadily growing, and is almost at 400 registered colleagues, despite the fact that until now no professional marketing activities have been initiated to promote telederm.org. As a way of evaluating the quality of the teledermatology service that telederm.org provides, an easy-to-use E-mail rating system has been implemented for direct consultations.
There are other Web sites offering free dermatology information, at least partially provided by users. For example, dermatlas.org (http://dermatlas.med.jhmi.edu/derm/index.cfm) is one of the largest dermatologic atlases supported by users' images. The Dermconsult Web site (http://www.dermconsult.com.au/) is an Australian, private, discretionary, educational Web site, login and password protected, which includes a virtual clinical meeting where dermatologists can post interesting cases with clinical data and images. The Virtual Grand Rounds in Dermatology (http://www.vgrd.org/index.html) also regularly posts cases in clinical dermatology open to users' opinions.
Conclusion
We have found that skin disorders can be telediagnosed by various experts worldwide, stimulating an exchange of knowledge and expertise. Building a connected world in dermatology by promoting free-access teleconsulting is one way to harness the opportunities opened up by the Internet, although concerns over security and privacy of health-care information remain.
We thank Walter H. C. Burgdorf, M.D., for his critical review and editing assistance.
Citation: Soyer HP, Hofmann-Wellenhof R, Massone C, Gabler G, Dong H, et al. (2005) telederm.org: Freely available online consultations in dermatology. PLoS Med 2(4): e87.
Abbreviation
SAFstore-and-forward system
==== Refs
References
European Commission e-Health—Making healthcare better for European citizens: An action plan for a European e-Health Area 2004 Available: http://europa.eu.int/information_society/doc/qualif/health/COM_2004_0356_F_EN_ACTE.pdf . Accessed 9 February 2005
Eedy DJ Wootton R Teledermatology: A review Br J Dermatol 2001 144 696 707 11298526
McColl I Dermatology education on the Web J Telemed Telecare 2003 9 Suppl 2 S33 S35
Argenziano G Soyer HP Chimenti S Talamini R Corona R Dermoscopy of pigmented skin lesions: Results of a consensus meeting via the Internet J Am Acad Dermatol 2003 48 679 693 12734496
Piccolo D Smolle J Argenziano G Wolf IH Braun R Teledermoscopy—Results of a multicentric study on 43 pigmented skin lesions J Telemed Telecare 2000 6 132 137 10912329
Massone C Hofmann-Wellenhof R Gabler G Dong H Kaddu S Global Teledermatology: A specific web application for dermatological consultation Internet Health 2004 3 e3 Available: http://www.internet-health.org/ih200431e03.html . Accessed 9 February 2005
Oztas MO Calikoglu E Baz K Birol A Onder M Reliability of Web-based teledermatology consultations J Telemed Telecare 2004 10 25 28
Piccolo D Smolle J Wolf IH Peris K Hofmann-Wellenhof R ‘Face-to-face’ versus telediagnosis of pigmented skin tumors—A teledermoscopic study Arch Dermatol 1999 135 1467 1471 10606051
| 15839749 | PMC1087218 | CC BY | 2021-01-05 10:39:46 | no | PLoS Med. 2005 Apr 26; 2(4):e87 | utf-8 | PLoS Med | 2,005 | 10.1371/journal.pmed.0020087 | oa_comm |
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 1583975110.1371/journal.pmed.0020106Research ArticleInfectious DiseasesInfectious DiseasesMedicine in Developing CountriesNonstationary Influence of El Niño on the Synchronous Dengue Epidemics in Thailand Influence of El Niño on Dengue EpidemicsCazelles Bernard
1
2
*Chavez Mario
3
McMichael Anthony J
4
Hales Simon
4
1CNRS UMR 7625, Ecole Normale SupérieureParisFrance2IRD UR GEODESBondyFrance3LENA-CNRS UPR 640, CHU Pitié-SalpêtrièreParisFrance4National Centre for Epidemiology and Population Health, Australian National UniversityCanberra, Australian Capital TerritoryAustraliaPascual Mercedes Academic EditorUniversity of MichiganUnited States of America
Competing Interests: The authors have declared that no competing interests exist.
Author Contributions: BC and SH designed the study. MC and BC analyzed the data. BC, AJM, and SH contributed to the writing of the paper.
*To whom correspondence should be addressed. E-mail: [email protected] 2005 26 4 2005 2 4 e10610 11 2004 1 3 2005 Copyright: © 2005 Cazelles 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 Complicated Art of Tracking Dengue
Background
Several factors, including environmental and climatic factors, influence the transmission of vector-borne diseases. Nevertheless, the identification and relative importance of climatic factors for vector-borne diseases remain controversial. Dengue is the world's most important viral vector-borne disease, and the controversy about climatic effects also applies in this case. Here we address the role of climate variability in shaping the interannual pattern of dengue epidemics.
Methods and Findings
We have analysed monthly data for Thailand from 1983 to 1997 using wavelet approaches that can describe nonstationary phenomena and that also allow the quantification of nonstationary associations between time series. We report a strong association between monthly dengue incidence in Thailand and the dynamics of El Niño for the 2–3-y periodic mode. This association is nonstationary, seen only from 1986 to 1992, and appears to have a major influence on the synchrony of dengue epidemics in Thailand.
Conclusion
The underlying mechanism for the synchronisation of dengue epidemics may resemble that of a pacemaker, in which intrinsic disease dynamics interact with climate variations driven by El Niño to propagate travelling waves of infection. When association with El Niño is strong in the 2–3-y periodic mode, one observes high synchrony of dengue epidemics over Thailand. When this association is absent, the seasonal dynamics become dominant and the synchrony initiated in Bangkok collapses.
In Thailand, dengue transmission patterns are complex, with El Nino showing an association with some epidemics only; other reasons will need to be found for the initiation of other outbreaks
==== Body
Introduction
Dengue is a peri-urban disease in the tropics and subtropics, transmitted principally by a single species of mosquito, Aedes aegypti. It has been estimated that 50 to 100 million people each year suffer from dengue and that two-fifths of the human population are at risk. The geographic distributions of dengue and of the potentially fatal form, dengue haemorrhagic fever (DHF), have expanded dramatically in recent decades [1]. The re-emergence of dengue has been connected to societal changes such as population growth, urbanisation, and international travel as well as environmental changes [2,3,4,5,6].
The relationship between climate, human behaviour, and infectious disease is complex, making it difficult to disentangle the different causal mechanisms [3,4,5,6,7,8,9]. It is well established that climate is an important determinant of vector-borne disease epidemics [3,4,5,6,7]. Climate directly influences the biology of the vectors and thereby their abundance and their distribution. Significant correlations have been reported between annual dengue incidence and estimates of Aedes aegypti populations at a national scale, using climate-based models [10]. Meteorological conditions can also directly or indirectly affect pathogen biology and epidemiological factors. Nevertheless, there is relatively sparse evidence of these climatic influences at interannual scales [7]. There is, however, evidence of a relationship between the timing of dengue epidemics and El Niño in the Pacific Islands [11,12] and in some other countries [13].
It is clear that several factors can influence the dynamics of vector-borne diseases, including environmental and climate factors, host–pathogen interactions, and population immunological factors [14]. It has been suggested that the effects of climate are unlikely to contribute to the timing of dengue epidemics in Thailand [14]. Using monthly data for Thailand from 1983 to 1997, Cummings et al. [15] identified travelling waves of dengue, initiated in the capital city, Bangkok, but did not investigate the potential influence of climate.
Dengue incidence data show complex nonlinear dynamics, with strong seasonality, multiyear oscillations, and nonstationarity (changes in dominant periodic components over time). These features of the data mean that conventional statistical methods may be inadequate. Using wavelet analysis, we provide here evidence for a nonstationary association between El Niño and the dynamics of dengue in Thailand. Moreover, we emphasise that this nonstationary association has some important implications for the characteristics of the synchronous dynamics of dengue epidemics in Thailand.
We analysed monthly data of DHF in the 72 provinces of Thailand from 1983 to 1997 [15] in relation to climate variables. Fourier analysis has traditionally been used to analyse the relationships between oscillating time series, but this method is not always appropriate when dealing with complex environmental time series. In particular, this approach can neither take into account the often observed changes in the periodic behaviour of such series, nor quantify the potential association between such series [16,17,18,19,20]. In contrast to Fourier analysis, wavelet analysis has been devised to analyse signals with changing spectra and allows the estimation of the spectral characteristics of a time series as a function of time. Wavelet analysis of a time series provides information on the evolution of the periodic components over time [16,17] and allows the quantification of nonstationary association between two time series [18]. To analyse our datasets we computed the following: (i) wavelet decomposition and wavelet power spectra, which determine the significant oscillating modes; (ii) wavelet coherence patterns, which describe local associations in both time and frequency domains; (iii) phase angles, which indicate the sign of the association, either in phase or out of phase [21]; and (iv) the evolution of the periodic components of each series in the most significant mode of oscillation. We have analysed the time series from the 72 provinces individually, but we show here only results for Bangkok versus the rest of Thailand combined. Similar results are observed with the time series from individual provinces (not shown).
Methods
The Data
The numbers of DHF cases used in this study are the monthly reports of DHF in 72 provinces of Thailand (see http://www.jhsph.edu/cir/dengue.html or [15]). We analysed two incidence time series from this dataset: the incidence in Bangkok, the capital city, and the averaged incidence for the rest of Thailand. The climatic data are climatic indexes that describe El Niño oscillations: the Nino 3 index and the Southern Oscillation Index (http://www.cgd.ucar.edu/cas/catalog/climind). We have also quantified the association with rainfall and temperature for the corresponding time periods and geographic areas [22]. For the wavelet analyses, the incidence time series were square root transformed and all the series were normalised before comparison.
The Wavelet Approach
Among the various approaches developed to study nonstationary data, wavelet analysis is probably the most efficient. In particular, this method gives us the possibility of investigating and quantifying the temporal evolution of time series with different rhythmic components (see [19] and [20] in a populational context). Wavelets constitute a family of functions derived from a single function, the “mother wavelet”, ψa,τ(t)
, that can be expressed as the function of two parameters, one for the time position τ, and the other for the scale of the wavelets a, related to the frequency. More explicitly, wavelets are defined as
In analysis of “natural signals”, the so-called Morlet wavelet is often applied [16,17]. The Morlet wavelet is defined as
The wavelet transform of a time series x(t) with respect to a chosen mother wavelet is performed as follows:
where the asterisk denotes the complex conjugate form. The wavelet coefficient Wx(a, τ) represents the contribution of the scale a to the signal when time is at different position τ. Computation of the wavelet transform is done along the signal x(t) simply by increasing the parameter τ over a range of scales a until all coherent structures within the signal can be identified.
With the wavelet approach, we can estimate the repartition of variance between scale a and different time location τ. This is known as the wavelet power spectrum: S
x(f,τ) = |W
x(f,τ)|2
. An important point is that the wavelet scale a is inversely proportional to the central frequency of the wavelet, f
0. In fact f ≈ 1/a when f
0 = 2π for the Morlet wavelet. Then scale a can be replaced by the frequency f or the period p; this thus greatly simplifies interpretation of the wavelet analyses. Using the inverse wavelet transform, the original signal can be recovered by integrating the wavelet transform over all scales and locations. This integration can be done over a given periodic band, p
1 to p
2. This allows us to filter the raw signal to obtain its oscillating components in the chosen periodic range.
To quantify statistical relationships between two time series, wavelet coherence can be computed [18]:
where the angle brackets around terms indicate smoothing in both time and frequency, Wx(f, τ) is the wavelet transform of series x(t), Wy(f, τ) is the wavelet transform of series y(t), and
is the cross-wavelet transform. The wavelet coherence provides local information about where two nonstationary signals, x(t) and y(t), are linearly correlated at a particular frequency (or period). Rx,y(f, τ) is equal to one when there is a perfect linear relationship at a particular time and frequency between the two signals.
In complement to wavelet analysis, we can use phase analysis to characterise the association between signals [21]. The phase difference provides information on the sign of the relationship (i.e., in phase or out of phase). As the Morlet wavelet is a complex wavelet, we can write Wx(f, τ) in terms of its modulus, |Wx(f, τ)|, and phase,
Similarly with the cross-wavelet transform Wx,y(f, τ) one can compute the phase difference:
and also the instantaneous time lag ΔT(τ) between the time series x(t) and y(t). This time lag is computed as
with F(τ) the instantaneous frequency defined in a given frequency (or periodic) band:
We performed all analyses using original algorithms developed in Matlab (version 6.5, The MathWorks, Natick, Massachusetts, United States). These original algorithms incorporate both cross analyses and adapted statistical procedures (B. Cazelles, M. Chavez, D. Berteaux, F. Ménard, J. O. Vik, et al., unpublished data).
Results
The oscillations of the dengue incidence time series are dominated by the annual mode of oscillation, and the El Niño is dominated by the 4–6-y components. Nevertheless, these time series have a statistically significant common mode of oscillation around a period of 2–3 y (see Figure S1). Different temporal associations between dengue and El Niño are seen in Bangkok and in the rest of Thailand (Figure 1; see also Figure S2). In each case, the wavelet analysis shows a main region of high and significant coherence for the 2–3-y periodic mode, between 1986–1992 (Figure 1B and 1C). In Bangkok, increases in dengue incidence precede changes in El Niño by several months, while for the rest of Thailand average monthly dengue incidence is perfectly in phase with El Niño (Figure 1D).
Figure 1 Association between Dengue in Bangkok and in the Rest of Thailand with El Niño Based on Wavelet Analysis
(A) Bangkok dengue incidence (blue line), Thailand dengue incidence (red line), and Nino 3 index (black dashed line). The incidence series are square root transformed, and all series are normalised.
(B) Wavelet coherence between dengue in Bangkok and Nino 3, computed using the Morlet wavelet function. The colours code for power values from dark blue for low coherence to dark red for high coherence. The nested white dashed lines show the α = 5% and α = 10% significance levels computed based on 1,000 bootstrapped series. The cone of influence indicates the region not influenced by edge effects.
(C) Wavelet coherence between dengue incidence in the rest of Thailand and Nino 3. Colours as in (B).
(D) Phases of time series (colours as in [A]) computed in the 2–3-y periodic band.
The delay between dengue incidence in Bangkok and in the rest of Thailand led us to analyse the synchrony in these data using a wavelet approach (Figure 2). This analysis shows three main regions of high and significant coherence (Figure 2A). The first one is for the 2–3-y periodic band for the time period 1985–1991, the second is for the 1-y bands for 1983–1984 and for 1992–1996, and the last is for the 5-y band after 1988. This last region must be interpreted cautiously because of the short length of the time series. We also analysed the phases (not shown here) and evolution of periodic components in the 2–3-y and the 1-y bands for dengue in Bangkok and in the rest of Thailand (Figure 2B and 2C). The two incidence series are phase locked with a mean delay of 3 mo in the 2–3-y band, but only within the period of high coherence with El Niño oscillations: 1984–1992. During this time period, the major part of the variance of the dengue time series is for this 2–3-y oscillating mode (see Figure S1). For 1983–1985 and 1991–1997, dengue incidence in Bangkok follows the incidence in the remainder of Thailand with an average delay of 1 mo (Figure 2C). In these years, as the 2–3-y mode is not dominant (see Figure S1), phase locking is seen only in the 1-y (seasonal) band.
Figure 2 Synchronisation between Dengue Incidence in Bangkok and in the Rest of Thailand
The incidence series are square root transformed, and all series are normalised.
(A) Wavelet coherence computed based on the Morlet wavelet function between dengue incidence in Bangkok and in the rest of Thailand; colours as in Figure 1B. The white dashed lines show the α = 5% significance level computed based on 1,000 bootstrapped series.
(B) Oscillating components computed with the wavelet transform in the 2–3-y period band (colours as in Figure 1A).
(C) Oscillating components computed with the wavelet transform in the 0.8–1.2-y period band (colours as in Figure 1A).
In (B) and (C) the black line shows the time evolution of the instantaneous time delay in months (ΔT) between the oscillating components of the two incidence time series.
This analysis confirms that there is synchrony between DHF incidence in Bangkok and the remainder of Thailand in the 2–3-y periodic band, as recently reported [15]. However, the present findings show that this synchrony is transient and appears to be influenced by El Niño.
In an effort to further understand this relationship, we analysed spatially averaged estimates of rainfall and temperature by month for Bangkok and for the rest of Thailand. We first focused on the link between El Niño and local climatic variables and found significant coherences in the 2–3-y periodic band only around the time period 1985–1992 in Bangkok, whereas this link is more constant throughout the study period for the remainder of Thailand (see also Figure S3). There is a highly significant coherence between the yearly components of DHF and rainfall (Figure 3A–3D). For this seasonal mode, DHF incidence and rainfall are phase locked in most of the country (Figure 3E). However, in Bangkok, the seasonal pattern of DHF incidence usually follows the seasonal peak of rainfall after a short lag time (Figure 3B). In Bangkok, in the time period 1986–1991, this association is replaced by a strong coherence in the 2–3-y band (Figure 3A). This coherence is also present for the rest of Thailand (Figure 3D), and for this mode, during the period of strong coherence, the dynamics are out of phase (Figure 3C–3F). A similar but weaker pattern of associations was observed for temperature (see also Figure S4).
Figure 3 Association between Precipitation and Dengue Incidence
For precipitation, gridded data [22] spatially averaged over rectangular areas representing Bangkok and the rest of Thailand using the IRI climate data library (http://ingrid.ldgo.columbia.edu/SOURCES/UEA/CRU/New/CRU05/monthly/) are used. The incidence series are square root transformed, and all series are normalised. The left part of the figure concerns Bangkok and the right part the rest of Thailand. On phase graphs, colours are as in Figure 1, and the dotted lines are for the phase difference between the considered series.
(A) and (D) Wavelet coherence (see Figure 1B).
(B) and (E) Phase evolutions of the considered series computed with the wavelet transform in the 0.8–1.2-y period band.
(C) and (F) Phase evolutions computed in the 2–3-y period band.
Discussion
These results provide several pieces of evidence for a complex, nonstationary relationship between El Niño, climatic variables, and DHF incidence. We have demonstrated a significant association between El Niño, climate variables, and DHF incidence for Bangkok and for the rest of Thailand. Our findings suggest that relationships between DHF and climate have a major influence on the previously reported synchrony of DHF epidemics [15].
In Bangkok, the association between DHF and climate occurs in two mutually exclusive modes, a yearly mode and a 2–3-y mode. The observed association between DHF and El Niño in the 2–3-y periodic mode coincides with the occurrence of high synchrony of DHF throughout Thailand initiated in the capital city, Bangkok. If the association in the yearly periodic mode becomes dominant, the synchrony of DHF dynamics initiated in Bangkok collapses and both the dynamics and the synchrony are dominated by the seasonal components (see Figure 2). In the rest of Thailand, the 2–3-y mode is never completely dominant and the seasonal mode persists throughout the dataset.
The complexity of the link between dengue dynamics and climate is emphasised by the positive correlations in the seasonal mode and negative correlations in the 2–3-y periodic mode. The results are consistent with the observation that, in most countries, dengue is most prevalent in the wet season, yet on an interannual scale, dengue epidemics have also been associated with drought [13]. In countries with high rainfall, drought can cause normally fast-flowing rivers to recede into a series of stagnant pools, ideal for mosquito breeding. On the other hand, in the Pacific Islands, dengue epidemics tend to occur during La Niña events, which are associated with conditions warmer and wetter than normal in most islands [11]. Dengue and climate might be driven by temperature, rather than rainfall.
Dengue in Bangkok seems to precede the oscillations of the Nino 3 index. This may reflect the timing of relationships between El Niño and climate. Another potential explanation could be a nonlinear or a threshold effect between large-scale phenomena and local dengue dynamics, as previously suggested for cholera [23]. The oscillations of the epidemics would be produced by local climatic phenomena generated before the maxima of the large-scale phenomena.
Alternatively, dengue epidemics might start in a nearby country where the effect of El Niño is more pronounced. Movement of infected vectors or travellers between countries could lead to propagation of the disease in synchrony with El Niño [12].
Whether the underlying climatic influence is local or regional, our findings suggest a biologically plausible mechanism for the recently reported synchronous dynamics of DHF in Thailand in the years 1985–1991. We hypothesise that under certain conditions, interannual variation in local or regional climate linked to El Niño may act as a pacemaker, modulating both the temporal dynamics and the spatial synchrony of DHF in a travelling wave.
These findings do not exclude an important role for other factors, such as intrinsic disease dynamics, in explaining patterns of dengue incidence in Thailand [24]. A previous study reported no apparent relationship between dengue and interannual climate in Bangkok between 1966 and 1998 [14]. However, in this work the authors [14] used spectral density analysis, which is not sensitive to nonstationary effects. Conventional statistical methods may fail to reveal a strong relationship between climate and a health outcome when discontinuous associations are present. The association between dengue and climate reported here is strong but transient. Nonstationarity can make it difficult to demonstrate even strong climate–health relationships. This has been reported by Rodó et al. [23] in the case of cholera epidemics. They have shown that the association between El Niño and cholera prevalence in Bangladesh is strong but transient. In the earlier part of the century, periodic components of cholera and El Niño were not associated, whereas late in the century (1980–2001) the relationship between these components was strong.
There is considerable interest in the role played by climate variability as a factor driving diseases [2,3,4,5,6,7,23,24,25]. Wavelet analyses can reveal transient population synchrony as well as long-term climate–health relationships. Future studies should use this approach to examine relationships between climate and dengue fever on regional and global scales, and attempt to identify the geographical location of the hypothesised pacemaker.
Supporting Information
Figure S1 Wavelet Transform of the Dengue Incidence and Nino 3 Time Series
The incidence series are square root transformed, and all series are normalised. The dashed lines, white or black, show the α = 5% significant levels computed based on 1,000 bootstrapped series. On the scalograms in (A), (C), and (E), the cone of influence, which indicates the region not influenced by edge effects, is also shown.
(A) Wavelet power spectrum (Sx(f, τ)) of dengue incidence in Bangkok. The colours code for power values from dark blue for low values to dark red for high values.
(B) The average wavelet spectrum of the time series.
(C and D) As in (A and B) but for the time series of dengue incidence in the rest of Thailand.
(E and F) As in (A and B) but for the time series of the Nino 3 index.
(358 KB EPS).
Click here for additional data file.
Figure S2 Association between Dengue in Bangkok and in the Rest of Thailand with El Niño Based on Wavelet Analysis of the Southern Oscillation Index
(72 KB EPS).
Click here for additional data file.
Figure S3 Associations between Climate and El Niño
For El Niño, the Nino 3 index is employed, and for climatic variables, gridded data [22] spatially averaged over rectangular areas representing Bangkok and the rest of Thailand using the IRI climate data library (http://ingrid.ldgo.columbia.edu/SOURCES/.UEA/.CRU/.New/.CRU05/.monthly/) are used. The left part of the figure concerns Bangkok and the right part the rest of Thailand. The seasonal components of the time series have been removed by filtering with a low-pass filter and a cutoff at 15 mo, and all series are normalised. (A–D) are related to rainfall and (E–H) to temperature.
(A), (C), (E), and (G) Wavelet coherence (see Figure 1.).
(B), (D), (F), and (H) Phase evolutions of the considered series computed with the wavelet transform in the 2–3-y period band. On phase graphs, colours are as in Figure 1, and the dotted lines are for the phase difference between the considered series.
(499 KB EPS).
Click here for additional data file.
Figure S4 Association between Temperature and Dengue Incidence
For temperature, gridded data [22] spatially averaged over rectangular areas representing Bangkok and the rest of Thailand using the IRI climate data library (http://ingrid.ldgo.columbia.edu/SOURCES/.UEA/.CRU/.New/.CRU05/.monthly/) are used. The incidence series are square root transformed, and all series are normalised. The left part of the figure concerns Bangkok and the right part the rest of Thailand. On phase graphs, colours are as in Figure 1, and the dotted lines are for the phase difference between the considered series.
(A) and (D) Wavelet coherence (see Figure 1B).
(B) and (E) Phase evolutions of the considered series computed with the wavelet transform in the 0.8–1.2-y period band.
(C) and (F) Phase evolutions computed in the 2–3-y period band.
(270 KB EPS).
Click here for additional data file.
Patient Summary
Background
Many things interact to determine when epidemics of disease occur. Some of these factors are due to the disease-causing agent itself or what carries it; other factors include climate, both local and over a larger region. Dengue fever, caused by a virus and transmitted by a mosquito, has a very complex pattern of epidemics.
What Did the Researchers Do?
They examined the pattern of dengue outbreaks, specifically the most serious form of dengue, dengue hemorrhagic fever, in the 72 provinces of Thailand between 1983 and 1997 and also looked at climate patterns, especially those caused by El Niño.
They found that though El Niño was associated with some specific disease outbreaks between 1986 and 1992, it was not associated with all of them, and for the remaining outbreaks other, more local factors were likely to be more important.
What Do These Findings Mean?
They provide more information about how dengue epidemics start and spread. They may be useful for those who plan public health measures in affected countries.
Where Can I Get More Information?
The United States Centers for Disease Control and Prevention has a Web page on dengue: http://www.cdc.gov/ncidod/dvbid/dengue/
The World Health Organization also provides information: http://www.who.int/mediacentre/factsheets/fs117/en/
MedlinePlus has a Web page aimed specifically at patients with dengue: http://www.nlm.nih.gov/medlineplus/ency/article/001374.htm
We are grateful to Professor D. S. Burke, who made Thailand dengue data available. BC is partially supported by two Gestion et Impacts du Changement Climatique grants (CHOLCLIM and MATECLID) from the French Ministry of Ecology and Sustainable Resources. The funders of this work had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Citation: Cazelles B, Chavez M, McMichael AJ, Hales S (2005) Nonstationary influence of El Niño on the synchronous dengue epidemics in Thailand. PLoS Med 2(4): e106.
Abbreviations
DHFdengue haemorrhagic fever
==== Refs
References
World Health Ov WHO report on global surveillance of epidemic-prone infectious diseases. WHO/CDS/CSR/ISR/2000.1. Geneva: World Health Organization. Available: http://www.who.int/csr/resources/publications/surveillance/en/WHO_Report_Infectious_Diseases.pdf
2000 Accessed 8 March 2005
Guzmàn MG Kouri G Dengue: An update Lancet Infect Dis 2002 2 33 42 11892494
Epstein PR Diaz HF Elias S Grabherr G Graham NE Biological and physical signs of climate change: Focus on mosquito-borne diseases Bull Am Meteorol Soc 1998 79 409 417
Patz JA Public health risk assessment linked to climatic and ecological change Hum Ecol Risk Assess 2001 7 1317 1327
Patz JA Reisen WK Climate-change and vector-borne diseases Trends Immunol 2001 22 171 172 11274908
Hales S de Wet N Maindonald J Woodward A Potential effect of population and climate changes on global distribution of dengue fever: An empirical model Lancet 2002 360 830 834 12243917
Kovats S Campbel-Lendrum DH McMichael AJ Woodward A Cox JS Early effects of climate change: Do they include changes in vector-borne disease? Philos Trans R Soc Lond B Biol Sci 2001 356 1057 1068 11516383
Hales S Kovats S Woodward A What El Niño can tell us about human health and global climate change Glob Change Hum Health 2002 1 66 77
Patz JA McGeehin MA Bernard SM Ebi KL Epstein PR The potential health impacts of climate variability and change for the United States: Executive summary of the report of the health sector of the U.S. national assessment Environ Health Perspect 2000 108 367 376 10753097
Hopp M Foley J Worldwide fluctuations in dengue fever cases related to climate variability Climate Res 2003 25 85 94
Hales S Weinstein P Woodward A Dengue fever epidemics in the South Pacific region: Driven by El Niño Southern Oscillation? Lancet 1996 348 1664 1665
Hales S Weinstein P Souares Y Woodward A El Niño and the dynamics of vector-borne disease transmission Environ Health Perspect 1999 107 99 102 9924003
Gagnon A Bush A Smoyer-Tomic K Dengue epidemics and the El Niño Southern Oscillation Climate Res 2001 19 35 43
Hay SI Myers MF Burke DS Vaughn DW Endy Y Etiology of interepidemic periods of mosquito-borne disease Proc Natl Acad Sci U S A 2000 97 9335 9339 10922081
Cummings DAT Irizarry RA Huang NE Endy TP Nisalak A Travelling waves in the occurrence of dengue haemorrhagic fever in Thailand Nature 2004 427 344 347 14737166
Lau KM Weng H Climatic signal detection using wavelet transform: How to make a time series sing Bull Am Meteorol Soc 1995 76 2391 2402
Torrence C Compo GP A practical guide to wavelet analysis Bull Am Meteorol Soc 1998 79 61 78
Liu PC Foufoula-Georgiou E Kumar P Wavelet spectrum analysis and ocean wind waves Wavelets in geophysics 1994 New York Academic Press 151 166
Grenfell BT Bjørnstad ON Kappey J Travelling waves and spatial hierarchies in measles epidemics Nature 2001 414 716 723 11742391
Klvana I Berteaux D Cazelles B Porcupine feeding scars and climatic data show ecosystem effects of the solar cycle Am Nat 2004 164 283 297 15478085
Cazelles B Stone L Detection of imperfect population synchrony in an uncertain world J Anim Ecol 2003 72 953 968
New MG Hulme M Jones PD Representing twentieth-century space-time climate variability. Part II: Development of 1901–1996 monthly grids of terrestrial surface climate J Climate 2000 13 2217 2238
Rodó X Pascual M Fuchs G Faruque SG ENSO and cholera: A nonstationary link related to climate change? Proc Natl Acad Sci U S A 2002 99 12901 12906 12228724
Koelle K Pascual M Disentangling extrinsic from intrinsic factors in disease dynamics: A nonlinear time series approach with an application to cholera Am Nat 2004 163 901 913 15266387
Pascual M Bouma MJ Dobson AP Cholera and climate: revisiting the quantitative evidence Microbes and Infection 2002 4 237 245 11880057
| 15839751 | PMC1087219 | CC BY | 2021-01-05 10:39:47 | no | PLoS Med. 2005 Apr 26; 2(4):e106 | utf-8 | PLoS Med | 2,005 | 10.1371/journal.pmed.0020106 | oa_comm |
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 1583975210.1371/journal.pmed.0020112Research ArticleInfectious DiseasesHIV/AIDSInfectious DiseasesHIV Infection/AIDSImpact of HIV-1 Subtype and Antiretroviral Therapy on Protease and Reverse Transcriptase Genotype: Results of a Global Collaboration Subtype and Drug Impact on GenotypeKantor Rami
1
*¤Katzenstein David A
1
Efron Brad
2
Carvalho Ana Patricia
3
Wynhoven Brian
4
Cane Patricia
5
Clarke John
6
Sirivichayakul Sunee
7
Soares Marcelo A
8
Snoeck Joke
9
Pillay Candice
10
Rudich Hagit
11
Rodrigues Rosangela
12
Holguin Africa
13
Ariyoshi Koya
14
Bouzas Maria Belen
15
Cahn Pedro
15
Sugiura Wataru
14
Soriano Vincent
13
Brigido Luis F
12
Grossman Zehava
11
Morris Lynn
10
Vandamme Anne-Mieke
9
Tanuri Amilcar
8
Phanuphak Praphan
7
Weber Jonathan N
6
Pillay Deenan
16
Harrigan P. Richard
4
Camacho Ricardo
3
Schapiro Jonathan M
1
Shafer Robert W
1
1Division of Infectious Disease and Center for AIDS Research, Stanford UniversityStanford, CaliforniaUnited States of America2Department of Statistics and Division of Biostatistics, Stanford UniversityStanford, CaliforniaUnited States of America3Hospital Egas MonizLisbonPortugal4BC Centre for Excellence in HIV/AIDS, VancouverBritish ColumbiaCanada5Health Protection AgencyPorton DownUnited Kingdom6Wright Fleming Institute, Imperial CollegeSt. Mary's Hospital, LondonUnited Kingdom7Chulalongkorn UniversityBangkokThailand8Universidade Federal do Rio de JaneiroBrazil9Rega Institute for Medical ResearchLeuvenBelgium10National Institute of Communicable DiseasesJohannesburgSouth Africa11Central Virology, Public Health LaboratoriesMinistry of Health, Tel-HashomerIsrael12Instituto Adolfo LutzSao PauloBrazil13Hospital Carlos IIIMadridSpain14National Institute of Infectious DiseasesTokyoJapan15Fundación HuespedBuenos AiresArgentina16University College London and Health Protection AgencyLondonUnited KingdomHo David D. Academic EditorThe Rockefeller UniversityUnited States of America
Competing Interests: The authors have declared that no competing interests exist.
Author Contributions: See Acknowledgments.
*To whom correspondence should be addressed. E-mail: rkantor@ brown.edu¤Current affiliation: Division of Infectious Diseases, Brown University, Providence, Rhode Island, United States of America
4 2005 26 4 2005 2 4 e1128 10 2004 7 3 2005 Copyright: © 2005 Kantor 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.
Monitoring HIV-1 Resistance across the Globe
Background
The genetic differences among HIV-1 subtypes may be critical to clinical management and drug resistance surveillance as antiretroviral treatment is expanded to regions of the world where diverse non-subtype-B viruses predominate.
Methods and Findings
To assess the impact of HIV-1 subtype and antiretroviral treatment on the distribution of mutations in protease and reverse transcriptase, a binomial response model using subtype and treatment as explanatory variables was used to analyze a large compiled dataset of non-subtype-B HIV-1 sequences. Non-subtype-B sequences from 3,686 persons with well characterized antiretroviral treatment histories were analyzed in comparison to subtype B sequences from 4,769 persons. The non-subtype-B sequences included 461 with subtype A, 1,185 with C, 331 with D, 245 with F, 293 with G, 513 with CRF01_AE, and 618 with CRF02_AG. Each of the 55 known subtype B drug-resistance mutations occurred in at least one non-B isolate, and 44 (80%) of these mutations were significantly associated with antiretroviral treatment in at least one non-B subtype. Conversely, of 67 mutations found to be associated with antiretroviral therapy in at least one non-B subtype, 61 were also associated with antiretroviral therapy in subtype B isolates.
Conclusion
Global surveillance and genotypic assessment of drug resistance should focus primarily on the known subtype B drug-resistance mutations.
Most research has focused on HIV-1 subtype B, although most infections worldwide are non-B. However, this paper suggests that the same drug resistance mutations occur across subtypes
==== Body
Introduction
The HIV-1 pandemic resulted from the cross-species transmission of a lentivirus, most likely of chimpanzee origin, that began spreading among humans during the first half of the previous century [1,2,3]. The progeny of this zoonotic infection—designated HIV-1 group M (main) viruses—make up the vast majority of HIV-1 infections. During their spread among humans, group M viruses have developed an extraordinary degree of genetic diversity, and most can be segregated into nine pure subtypes and several commonly circulating recombinant forms [4].
HIV-1 subtype B is the predominant subtype in North America, Western Europe, and Australia. The antiretroviral drugs used to treat HIV were developed using biophysical, biochemical, and in vitro studies of subtype B isolates, and most data on the genetic mechanisms of HIV-1 drug resistance are from subtype B viruses. However, HIV-1 subtype B viruses account for only approximately 12% of the global HIV pandemic [5], and as therapy is introduced into developing countries, the number of persons with non-B viruses initiating therapy will increase dramatically.
HIV-1 subtypes differ from one another by 10%–12% of their nucleotides and 5%–6% of their amino acids in protease and reverse transcriptase (RT) [6]. Intersubtype nucleotide differences influence the spectrum of amino acid substitutions resulting from point mutations, and intersubtype amino acid differences influence the biochemical and biophysical microenvironment within the protease and RT [7,8]. These differences among subtypes therefore could influence the spectrum of mutations that develop during selective drug pressure.
An increasing number of observational studies, in vitro and in vivo, suggest that the currently available protease and RT inhibitors are as active against non-B viruses as they are against subtype B viruses [9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26]. However, fewer data are available on the genetic mechanisms of drug resistance in non-B viruses, and some in vitro and in vivo observations suggest that the various subtypes may respond differently to certain antiretroviral drugs [27,28,29,30,31,32,33,34,35].
Identifying the relevant drug-resistance mutations among non-B subtypes will be important for monitoring the evolution and transmission of drug resistance, for determining initial treatment strategies for persons infected with non-B viruses, and for interpreting genetic resistance among patients who fail antiretroviral therapy.
In this study, we characterize protease and RT mutations in non-B HIV-1 subtypes from persons receiving antiretroviral therapy, and attempt to answer the following two questions. (i) Do the mutations that cause drug resistance in subtype B viruses also develop in non-B viruses exposed to antiretroviral drugs? (ii) Do novel mutations emerge in non-subtype-B viruses during antiretroviral drug failure that are not recognized in subtype B viruses?
Methods
HIV-1 Sequences and Antiretroviral Treatments
Sequences of HIV-1 protease (positions 1–99) and RT (positions 1–240) from persons whose antiretroviral treatment history was known were collected from the published literature and from 14 laboratories in 12 countries. Persons were considered untreated if they had never been exposed to antiretroviral drugs, and treated if they were receiving RT inhibitors (RTIs) and/or protease inhibitors (PIs) at the time the isolate was obtained. Sequences from treated persons were included for analysis only if they were obtained from persons whose entire treatment histories were known. If multiple isolates from the same person were sequenced, only the latest isolate was included for analysis. Only sequences determined using dideoxyterminator sequencing were included in the analysis. In all, 99% of sequences were determined using direct PCR (population-based sequencing), and 1% of sequences represented the consensus sequence of multiple clones.
Samples obtained from patients were submitted to clinical and research laboratories for resistance testing in the course of evaluation and care of HIV infection. Data analyzed included published and presented data obtained under protocols approved by national and local institutional review boards or ethical review panels in each country. Sequence, demographic, and treatment data, unlinked from all personal identifiers, were analyzed at Stanford University under a protocol approved by the Stanford University Panel on Human Subjects.
Subtype Assignment
Similarity plotting and bootscanning using a window size of 400 nucleotides and a step size of 40 nucleotides were performed using reference sequences for each of the nine pure subtypes (A, B, C, D, F, G, H, J, and K) and two recombinant forms (CRF01_AE and CRF02_AG) [36]. Isolates that contained a combination of more than one subtype were excluded from analysis, except when subtypes A and G were detected in a pattern consistent with CRF02_AG. Because CRF01_AE pol sequences do not contain recombinant breakpoints, subtype assignment was based on the fact that pol CRF01_AE and pure A sequences are divergent. This approach had an accuracy of 96% when applied to the protease and RT genes of 137 well characterized subtype A, CRF01_AE, and CRF02_AG isolates with known subtypes based on pol and gag and/or env, with most errors resulting from the misclassification of subtype A protease sequences as CRF01_AE (data not shown).
Reference sequences used were U455 (subtype A), CM240 (CRF01_AE), IbNG (CRF02_AG), HXB2 (subtype B), C2220 (subtype C), NDK (subtype D), 93BR020 (subtype F), SE6165 (subtype G), 90CR056 (subtype H), SE9173c (subtype J), 97EQTB11C (subtype K), YBF30 (Group N), and ANT70C (Group O). A total of 223 protease and 307 RT sequences of indeterminate subtype were excluded from the analysis.
Mutation Definitions
Each sequence was translated and compared to the consensus B protease and RT sequences in the Los Alamos HIV Sequence database (http://hiv-web.lanl.gov) using the HIVSeq program [37]. Mutations were defined as differences from the wild-type consensus B sequence. Known subtype B drug-resistance mutations were defined as follows: 18 nucleoside RTI (NRTI)–resistance positions at 41, 44, 62, 65, 67, 69, 70, 74, 75, 77, 115, 116, 118, 151, 184, 210, 215, and 219; 15 non-nucleoside RTI (NNRTI)–resistance positions at 98, 100, 101, 103, 106, 108, 179, 181, 188, 190, 225, 227, 230, 236, and 238; and 22 protease inhibitor (PI)–resistance positions at 10, 20, 24, 30, 32, 33, 36, 46, 47, 48, 50, 53, 54, 63, 71, 73, 77, 82, 84, 88, 90, and 93 [38,39]. Mutations also included differences from consensus B that were present as part of a nucleotide mixture.
Polymorphisms were defined as mutations that occurred in more than 1% of sequences from untreated persons. Subtype-specific polymorphisms were defined as mutations that were significantly more prevalent in each non-B subtype than in subtype B viruses from untreated persons. Subtype-specific treatment-related mutations were defined as mutations that were significantly more prevalent in HIV-1 isolates from treated persons than in isolates from untreated persons infected with the same subtype.
Quality Control
Phylogenetic analysis to detect duplicate sequences identified 23 pairs of identical sequences and 1,039 pairs of sequences that differed from one another by no more than 1% of their nucleotides. To reduce the likelihood that isolates from different persons with similar mutations resulted from duplicate reporting, laboratory contamination, or HIV-1 transmission, only one sequence from each of these 1,062 sequence pairs was included in the analyses in this study. Although all extant HIV-1 isolates are epidemiologically linked through chains of transmission, protease or RT sequences that diverge by 1% or less appear to be more likely to result from direct transmission than those that diverge by more than 1% [40].
To distinguish mutations developing in multiple individuals from mutations that developed in a smaller number of founder viruses, we reconstructed the ancestral sequences at each node of the phylogenetic trees for each subtype and counted the number of times each mutation was predicted to have developed within a subtype. Mutations for which founder viruses accounted for 75% or more of occurrences (i.e., the number of branches on which the mutation has developed divided by the number of sequences with the mutation is less than 75%) were not considered treatment-related mutations. For this analysis, phylogenetic trees were created using the neighbor-joining method using the HKY85 model with gamma distribution within PAUP* version 4.0b10 for each subtype and each gene. Ancestral sequences were reconstructed using MESQUITE version 1.02 (http://www.mesquiteproject.org).
To further reduce the influence of transmitted drug resistance on this analysis and to exclude the possibility that some untreated persons were classified incorrectly, sequences from untreated persons containing two or more non-polymorphic resistance mutations were excluded from the analysis. This approach was predicated on the strong likelihood that the presence of two mutations at highly conserved sites in untreated persons does not reflect natural variation in protease or RT but rather is most consistent with previous selective drug pressure. Based on this criterion, 23 protease and 25 RT sequences from 47 persons were excluded from the analysis: 22 CRF01_AE sequences, 15 subtype C sequences, six CRF02_AG sequences, five subtype G sequences, four subtype A sequences, three subtype F sequences, and one subtype D sequence. The mutations in the excluded sequences consisted almost entirely of the NRTI-resistance mutations at positions 41, 67, 70, 210, 215, and 219; the NNRTI-resistance mutations at positions 103, 181, and 190; and the PI-resistance mutations at positions 48, 82, and 90. An analysis that included these 56 sequences did not affect any of the significant findings in the study because these mutations were so much more common in treated than in untreated persons in multiple different subtypes (data not shown).
Statistical Analysis
Frequencies of mutations at each RT and protease codon were analyzed by a binomial response model employing a cube root transformation (similar to a logistic transform, with higher accuracy for extreme values) to identify significant differences in polymorphisms and treatment-related mutations between subtypes. Mutation frequencies for treated and untreated persons were compared for each subtype (Figure 1). This analysis defined three parameters: (i) a subtype parameter (θ1), comparing codons between untreated persons infected with B and each of the non-B subtypes; (ii) a treatment parameter (θ2), comparing codons between treated and untreated sequences of the same subtype; and (iii) an interaction parameter (θ3), comparing the effect of treatment between subtype B and each of the non-B subtypes.
Figure 1 Binomial Response Model Used to Evaluate Subtype and Treatment Effects on Genotypic Evolution for Each Protease and RT Position
A separate model was created for each non-B subtype. The frequencies of mutations at each position in four patient groups (untreated subtype B, treated subtype B, untreated non-B, and treated non-B) were converted to Y scores using a cube root transformation (similar to a logistic transform). Subtype effect was evaluated by calculating θ1, the score differences between non-B and B subtypes in untreated persons. The treatment effect was evaluated by calculating θ2, the score differences between treated and untreated persons within the same subtype. The subtype–treatment interaction was evaluated by calculating θ3, the difference of differences in the 2 × 2 table, or the difference in treatment effects between non-B and B subtypes.
To increase the statistical power of our analysis, we made two simplifications. First, we did not distinguish between distinct substitutions at the same position; all differences from consensus B were considered mutations. Second, viruses were categorized only according to the classes of drugs (PI, NRTI, NNRTI) to which they had been exposed.
To correct for multiple comparisons between subtype B and each non-B subtype, significant results included those z values exceeding three in absolute value, according to a 0.05 Benjamini–Hochberg false discovery rate criterion [41]. This method is a sequential Bonferroni-type procedure that is appropriate for situations in which multiple statistically significant associations are expected. The coefficients in the binomial response model were ranked in ascending order, and each hypothesis of rank r was compared with a significance cutoff of 0.05 (false discovery rate) multiplied by r/n, where n was 99 for the protease mutations and 240 for the RT mutations (i.e., number of comparisons).
Results
HIV-1 Subtypes
Sequences were obtained from 3,686 persons, in 56 countries, infected with non-B HIV-1 subtypes (Figure 2; Table 1), including 1,997 persons from whom both protease and RT sequences were available, 908 persons from whom only protease sequences were available, and 933 persons from whom only RT sequences were available. Sequences from untreated individuals were isolated between 1983 and 2003. Sequences from treated individuals were isolated between 1993 and 2003. A total of 2,311 (82%) protease and 2,381 (83%) RT sequences were obtained from plasma samples. The remaining sequences were obtained from peripheral blood mononuclear cells.
Figure 2 Number of Treated and Untreated Persons Infected with B and Non-B HIV-1 Subtypes from Whom Protease and/or RT Sequences Were Available for Analysis
Table 1 Geographical Origin of Persons Infected with Non-B Subtypes
Numbers of persons from countries for which ten or more sequences were available are shown in parentheses. aSubtypes H, J, and K. DOI: 10.1371/journal.pmed.0020112.t001
Antiretroviral Treatments
Of the participants with non-subtype-B viruses, 1,533 (42%) were receiving antiretroviral drugs at the time of sequencing: 1,140 were receiving NRTIs, 527 PIs, and 766 NNRTIs. According to subtype, 89% to 100% had received one or more NRTIs, 22% to 76% had received one or more PIs, and 32% to 55% had received one or more NNRTIs. Among treated persons infected with subtype B viruses, 98% had received NRTIs, 66% had received PIs, and 34% had received NNRTIs (Figure 3).
Figure 3 Proportions of Persons Receiving Treatment with Specific NRTIs, NNRTIs, and PIs
The number of persons with non-B virus receiving the NRTI tenofovir (ten), the NNRTI delavirdine (three), and the PIs amprenavir (13) and lopinavir (28) are not shown. 3TC, lamivudine; ABC, abacavir; AZT, zidovudine; D4T, stavudine; DDC, zalcitabine; DDI, didanosine; EFV, efavirenz; IDV, indinavir; NFV, nelfinavir; NVP, nevirapine; RTV, ritonavir; SQV, saquinavir.
Mutation Prevalence
Twenty-two (22%) protease and 87 (36%) RT positions were conserved in all subtypes regardless of the presence or absence of therapy. Twenty-four (24%) protease and 38 (16%) RT positions were conserved in untreated persons but were mutant in more than 1% of treated persons. The remaining 53 (53%) protease and 115 (48%) RT positions were polymorphic, or mutant in more than 1% of untreated persons.
To assess the impact of viral subtype and treatment on the distribution of mutations in protease and RT, a binomial response model using subtype and treatment as explanatory variables was used to predict whether a position was wild-type (matching the consensus B site) or mutant. This model identified three types of positions: (i) positions in sequences from untreated people more likely to be mutated in non-B than in B subtypes (subtype-specific polymorphisms); (ii) positions in sequences of the same subtype more likely to be mutated in treated than in untreated people (subtype-specific treatment-related positions); and (iii) positions for which the effect of treatment differed significantly between non-B and B subtypes (subtype–treatment interactions).
Subtype-specific polymorphisms
Figure 4 shows the mutation prevalence according to subtype for 37 protease and 41 RT subtype-specific polymorphisms (significant θ1; see Methods). Twenty-eight of the protease and 26 of the RT subtype-specific polymorphisms were polymorphic in untreated subtype B viruses, whereas nine of the protease and 15 of the RT were conserved in subtype B. Subtype-specific polymorphisms at conserved positions in untreated subtype B viruses were generally present in a small number of subtypes at low levels (<5%). Notable exceptions included protease positions 45 and 74 and RT positions 40, 43, 104, 195 and 238.
Figure 4 Subtype-Specific Polymorphisms
Positions in protease (left) and RT (right) at which mutation frequency varied significantly between subtype B and at least one non-B subtype in untreated persons. Positions are shown along the x-axes, and the frequency of mutation for each subtype is shown along the y-axes. Positions related to drug resistance in subtype B are boxed. Bar colors denote statistical significance: black is statistically significant (Zθ1 ≥ 3); gray is borderline significant (1 ≤ Zθ1 < 3); white is not statistically significant (Zθ1 < 1).
Six subtype-specific polymorphisms in protease (positions 10, 20, 33, 36, 82, and 93) and five in RT (98, 116, 179, 227, and 238) occurred at sites known to be associated with drug resistance in subtype B viruses. These positions (with the exception of positions 116, 227, and 238 in RT) were also polymorphic in subtype B. M184I was present in six monophyletic CRF01_AE isolates from untreated persons [42,43] and was therefore not considered to be a subtype-specific polymorphism. These six sequences also displayed G→A hypermutation [44], possibly explaining the M184I change (ATG→ATA) and further complicating the significance of this finding.
Subtype-specific polymorphisms at four protease and three RT drug-resistance positions represented the consensus sequence for at least one non-B subtype: K20I in subtypes G and CRF02_AG, M36I in subtypes A, C, D, F, G, CRF01_AE and CRF02_AG, V82I in subtype G, and I93L in subtype C for protease; and A98S in subtype G, V179I in subtype A, and K238R in CRF01_AE for RT. Each of the non-B subtypes was significantly more polymorphic than subtype B at protease positions 20, 36, and 41 and RT positions 35, 39, and 207.
Subtype-specific treatment-related mutations.
Figure 5 shows the mutation prevalence according to subtype for 31 protease and 36 RT subtype-specific treatment-related positions significantly more likely to be mutant in treated than untreated persons in at least one non-B subtype (significant θ2; see Methods). The protease positions included 16 of the 22 known PI resistance positions and 15 additional treatment-related positions. The RT positions included 28 of the known 33 RTI resistance positions and eight additional treatment-related positions. Although each of the known PI- and RTI-resistance positions occurred in at least one non-B subtype, three of the 22 protease positions and five of the 33 RT positions included mutations that occurred too infrequently for a significant association with treatment to be detected in our analysis.
Figure 5 Subtype-Specific Treatment-Related Mutations
Positions in protease (left) and RT (right) at which mutations were significantly more prevalent in HIV-1 isolates from treated than from untreated persons infected with the same subtype. Positions are shown along the x-axes, and the proportion of mutant sequences in treated persons for each subtype is shown along the y-axes. For protease (left), treated persons are those receiving one or more PIs. For RT (right), treated persons are those receiving one or more NRTIs. Positions related to drug resistance in subtype B are boxed. Bar colors denote statistical significance: black is statistically significant (Zθ2 ≥ 3); gray is borderline significant (1 ≤ Zθ2 < 3); white is not statistically significant (Zθ2 < 1).
Of the 15 treatment-related protease positions not known to be associated with drug resistance, eight were also significantly associated with treatment in subtype B viruses (positions 13, 23, 43, 45, 62, 66, 74, and 85), and two have been previously reported to be associated with treatment in subtype B viruses (positions 22 and 83) [45]. The remaining five subtype-specific treatment-related protease positions included positions 6, 15, 19, 37 and 64, which—although highly polymorphic in many subtypes—are associated with treatment in subtype C (positions 6 and 64), CRF02_AG (position 15), subtype F (position 19), subtype A (position 37), and CRF01_AE (position 64).
Of the eight treatment-related RT positions at sites not known to be associated with drug resistance, seven were also significantly associated with treatment in subtype B viruses (positions 68, 203, 208, 218, 221, 223, and 228) and one (position 102) was associated with treatment in subtype C but not B.
Subtype–treatment interactions
The subtype of the sequence significantly influenced the effect of treatment (significant θ3; see Methods) on 20 protease positions (10, 12, 13, 14, 15, 20, 37, 53, 62, 63, 64, 65, 67, 71, 73, 74, 77, 82, 88, and 89) and 11 RT positions (35, 39, 48, 98, 104, 106, 121, 162, 166, 179, and 238). For example, RT position 98 was mutant in 7% of untreated and 16% of treated persons with subtype B viruses (approximately 2-fold difference) and in 1% of untreated and 14% of treated persons with CRF01_AG viruses (14-fold difference). Other positions less likely to be mutated in subtype B than in non-B viruses in response to treatment included protease residues 14 (subtype A); 13 and 64 (subtype C); 37 and 65 (subtype F); 71 (subtype G); 62 and 64 (CRF01_AE); and 15 and 71 (CRF02_AG); and RT residues 35 (subtype A); 98 and 106 (subtype C); 35 and 98 (subtype G); and 98 (CRF02_AG).
At other positions, treatment had a larger effect on subtype B viruses than on one or more non-B subtypes. For example, protease position 20 was mutant in 2% of untreated and 24% of treated persons with subtype B viruses (approximately 12-fold increase with treatment) and in 11% of untreated and 42% of treated persons with subtype C viruses (approximately 4-fold increase with treatment). These positions included protease residues 10, 20, and 63 (subtype A); 20, 53, 63, 74, and 82 (subtype C); 13 and 20 (subtype D); 10, 14, 20, and 77 (subtype F); 20, 67, 73, 82, and 88 (subtype G); 20, 63, 82, and 89 (CRF01_AE); and 20 (CRF02_AG); and RT residues 39 and 179 (subtype A); 35, 48, 121, and 166 (subtype C); 39 (subtype D); 39 (subtype F); 39 and 104 (subtype G); 162 and 238 (CRF01_AE); and 39 (CRF02_AG).
These 31 positions with subtype–treatment interactions included 12 known drug-resistance positions. Of these, seven protease (10, 20, 53, 63, 77, 82, and 88) and two RT (179 and 238) resistance positions were more likely to be mutated in subtype B than in one or more non-B subtypes in response to treatment. One protease position (71) and two RT positions (98 and 106) were more likely to be mutated in one or more non-B subtypes.
Known Drug-Resistance Mutations
Figure 6 shows the amino acid substitutions present at drug-resistance positions in protease and RT sequences from untreated and treated persons infected with B and non-B subtypes. Fourteen of the 22 known PI-resistance positions occurred in subtype A, 20 in subtype C, 16 in subtype D, 20 in subtype F, 18 in subtype G, 17 in CRF01_AE, and 17 in CRF02_AG. Thirteen of the 18 known NRTI-resistance positions occurred in subtype A, 18 in subtype C, 13 in subtype D, 15 in subtype F, 18 in subtype G, 18 in CRF01_AE, and 16 in CRF02_AG. Ten of the 15 known NNRTI-resistance positions occurred in subtype A, 15 in subtype C, 11 in subtype D, 13 in subtype F, 14 in subtype G, 13 in CRF01_AE, and 12 in CRF02_AG.
Figure 6 A Amino Acid Differences from Consensus B Sequence at Drug-Resistance Positions in Protease and RT according to Subtype
(A) shows data for protease, and (B) shows data for RT. In both, the first line lists the drug-resistance positions. The second line shows single-letter amino acid codes for the consensus B sequence. For each subtype (left column), the frequency of specific mutations in untreated persons is shown above the dashed line, whereas the frequency of specific mutations in treated persons is shown below the dashed line. Positions with significant differences in mutation frequency between B and non-B subtypes (p < 0.01, according to χ2 test with Yate's correction) are circled. A pound sign indicates an insertion.
Figure 6 B Amino Acid Differences from Consensus B Sequence at Drug-Resistance Positions in Protease and RT according to Subtype
(A) shows data for protease, and (B) shows data for RT. In both, the first line lists the drug-resistance positions. The second line shows single-letter amino acid codes for the consensus B sequence. For each subtype (left column), the frequency of specific mutations in untreated persons is shown above the dashed line, whereas the frequency of specific mutations in treated persons is shown below the dashed line. Positions with significant differences in mutation frequency between B and non-B subtypes (p < 0.01, according to χ2 test with Yate's correction) are circled. A pound sign indicates an insertion.
In all, 106 of 113 (94%) different amino acid substitutions at 55 known subtype B drug-resistance positions (22 protease and 33 RT) were also present in at least one non-B subtype. In an exploratory analysis, which was not controlled for multiple comparisons, the frequencies of 24 mutations at 14 protease positions and 32 mutations at 19 RT positions differed between subtype B and one or more non-B subtypes.
Discussion
This collaborative analysis was designed to determine whether and to what degree the genetic mechanisms of HIV drug resistance are shared between subtype B and non-B viruses. Mutations responsible for drug resistance in subtype B viruses have been characterized by three types of studies: (i) those that identify mutations selected in viruses of persons receiving antiretroviral therapy, (ii) those that quantify the effect of specific mutations on in vitro drug susceptibility, and (iii) those that examine the effectiveness of treatment regimens in persons with viruses containing known or suspected drug-resistance mutations. This study, which identifies mutations arising in non-B viruses during antiretroviral therapy, is a necessary step for designing laboratory and clinical studies of potential drug-resistance mutations.
Do the known subtype B drug-resistance mutations also occur in non-B subtypes? We found that each of the 55 known subtype B drug-resistance mutations occurred in at least one non-B isolate. Of these, 44 (80%) were significantly associated with drug therapy in non-B isolates. The remaining 11 mutations were uncommon in subtype B and all non-B subtypes, making it difficult to determine whether they were also significantly associated with therapy. Phenotypic susceptibility testing of non-B viruses with treatment-selected mutations is necessary to confirm and quantify the contribution of these mutations to drug resistance in the genetic context in which they arise.
Do non-subtype B viruses from persons with virologic failure develop novel mutations? Fifteen protease and eight RT positions not generally considered to be drug-resistance positions were significantly associated with treatment in at least one non-B subtype. However, mutations at 17 of these 23 positions were also associated with treatment in subtype B viruses. Therefore, of the 67 mutations associated with treatment in at least one non-B subtype, 61 were also associated with treatment in subtype B. For the six mutations associated with therapy in at least one non-B subtype but not in subtype B, the associations were at the borderline of significance and require confirmation.
Among untreated persons, non-B subtype-specific polymorphisms occurred at 37 protease and 41 RT positions. Most of these non-B polymorphic positions are also polymorphic in subtype B viruses, and several act as accessory drug-resistance mutations in subtype B viruses. Phenotypic susceptibility testing of non-B viruses with such polymorphic accessory mutations is needed to confirm that these naturally occurring viruses are fully susceptible to current antiretrovirals—a supposition that appears be true based on the excellent virologic responses of non-B viruses to antiretroviral treatment in observational studies.
We made two simplifications in this study to increase the statistical power of our analyses. These will become unnecessary in future analyses as sufficient numbers of sequences from persons with well characterized treatment histories become available. First, we did not distinguish between different substitutions at the same position; all differences from consensus B were considered mutations. Second, viruses were classified only by the classes of drugs to which they were exposed rather than by individual drugs or drug regimens. Therefore, our analyses could not detect differences between subtype B and other subtypes that depend on specific mutations or specific drugs. Indeed, two such differences have been reported: (i) V106M is the most common substitution at RT position 106 in subtype C viruses whereas V106A predominates in subtype B viruses [33,35,46], and (ii) although the protease mutations D30N and L90M both develop in non-B viruses during nelfinavir therapy, D30N occurs more commonly in subtype B, whereas L90M occurs more commonly in subtypes C, G, and CRF01_AE [17,34,47,48].
Although the clinical samples in this study were originally obtained for a variety of purposes, including clinical management, the sequences of these samples represent experiments of nature that reveal the mutations associated with continued HIV-1 replication in the presence of selective antiretroviral therapy. The accurate identification of treatment-related mutations in such a cross-sectional analysis is challenging, however, because misclassification can result from the transmission of drug-resistant viruses, differences in specific HIV-1 variants among different human populations (population stratification), and the many statistical comparisons required as a result of HIV-1 sequence variability.
The transmission of drug-resistant HIV-1 viruses weakens cross-sectional analyses because some untreated persons may have been infected with viruses already containing treatment-related mutations. To mitigate this effect, we excluded isolates from untreated persons containing two or more non-polymorphic known drug-resistance mutations, because this pattern is not consistent with natural sequence variation. However, as noted in the Methods, an analysis that included these isolates did not alter any of the significant findings in the study. Conversely, the possibility that resistance mutations transmitted between persons in our dataset inflated the amount of resistance among persons receiving treatment was mitigated by excluding any isolate differing from another isolate at less than 1% of its nucleotides.
HIV-1 evolution is driven by genetic drift, immunologic pressure, and selective drug pressure. Population stratification can be a confounding factor when viral lineages with different founder mutations (resulting from drift or immunologic pressure) are exposed to different degrees of antiretroviral selection pressure. To distinguish mutations developing in multiple individuals as a result of selective drug pressure from mutations originating in a fewer number of founder viruses, we reconstructed the ancestral sequences at each node of a phylogenetic tree for each subtype and counted the number of times each mutation was predicted to have developed within that subtype. Because of the limited ability of phylogenetic methods to estimate accurate trees for large numbers of related sequences (i.e., belonging to the same HIV-1 subtype), only those positions for which the majority of mutations (≥75%) appeared to result from new mutations were considered to be selected by antiretroviral therapy.
Because this analysis was, to our knowledge, the first to simultaneously examine all protease and most polymerase-coding RT positions in multiple subtypes, and because multiple associations between mutation and treatment were expected, we used a relatively lenient correction for multiple comparisons in order to minimize the number of missed associations. Nonetheless, of the 67 positive associations detected in this study, 61 were also present in persons with subtype B viruses and have previously been reported [45,49].
In conclusion, most of the protease and RT positions associated with drug resistance in subtype B viruses are selected by antiretroviral therapy in one or more non-B subtypes as well. Conversely, we found no evidence that non-B viruses develop resistance by mutations at positions that are not associated with resistance in subtype B viruses. Based on currently available data, global surveillance efforts and genotypic assessments of drug resistance should focus primarily on the known subtype B drug-resistance mutations.
Supporting Information
Dataset S1 List of GenBank Accession Numbers for Non-Subtype-B Sequences Used in This Study
(59 KB PDF).
Click here for additional data file.
Accession Numbers
The GenBank (http://www.ncbi.nlm.nih.gov/Genbank/) isolates and accession numbers for the reference subtype specimens discussed in this paper are U455/subtype A (M62320), HXB2/subtype B (K03455), C2220/subtype C (U46016), NDK/subtype D (M27323), 93BR020/subtype F (AF005494), SE6165/subtype G (AF061642), 90CR056/subtype H (AF005496), SE9173c/subtype J (AF082394), 97EQTB11C/subtype K (AJ249235), CM240/CRF01_AE (U54771), IbNG/CRF02_AG (L39106), YBF30/Group N (AJ006022), and ANT70C/Group O (L20587). The accession numbers for the non-subtype-B sequences used in this study are listed in Dataset S1.
Patient Summary
Background
There are many different subtypes of HIV-1. The most common one in more developed countries is subtype B and that is the one which has been studied most and used in drug development. However, worldwide, other subtypes are more frequent. All HIV-1 subtypes acquire mutations, and some of these cause resistance to the drugs used to treat HIV. It is not clear whether the same mutations that cause drug resistance to subtype B virus are also important in causing resistance to non-subtype-B viruses.
What Did the Researchers Do?
They compared the viral sequences of 3,686 people with non-subtype-B HIV, and 4,769 with subtype B virus, all with known treatment histories. They found that the mutations known to cause drug resistance in subtype B virus also occur in non-subtype B, and the majority of mutations in non-subtype B also occur in subtype B.
What Do These Findings Mean?
It seems that largely the same mutations occur in both subtype B and non-subtype-B viruses. However, some mutations were only present in low numbers, so more work will need to be done before their role is clear. Also, the authors did not look at mutations and their relation to each different drug a patient had—only the general type of drug. Nor did they look at what happens when different mutations occur at one place in a virus. However, for now the current strategy of focusing on assessing the mutations seen in subtype B virus seems a reasonable approach to take when assessing surveillance of drug resistance while more work is done to follow up these findings.
Where Can I Get More Information?
TheBody.com has a section on drug resistance: http://www.thebody.com/treat/resistance.html
The Aidsmap Web site has many patient information sheets, including on resistance: http://www.aidsmap.com
D. Katzenstein is the recipient of a Doris Duke Distinguished Clinical Scientist award, and the support of the Doris Duke Charitable Foundation for R. Kantor and the project is gratefully acknowledged. R. Shafer was supported by grants from the National Institutes of Allergy and Infectious Diseases (AI46148–01) and National Institute of General Medical Sciences (5P01GM066524–02). We thank Paul Weidle for data submission. We thank Matthew J. Gonzales, Soo-Yon Rhee, and Tommy Liu for computation assistance and Jonathan Taylor for help with the statistical analysis. The Doris Duke Charitable Foundation did not have any role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Author contributions. R. Kantor, D. A. Katzenstein, B. Efron, J. M. Schapiro, and R. W. Shafer designed the study. R. Kantor, D. A. Katzenstein, B. Efron, A. P. Carvalho, B. Wynhoven, P. Cane, J. Clarke, S. Sirivichayakul, M. A. Soares, J. Snoeck, C. Pillay, H. Rudich, R. Rodrigues, A. Holguin, K. Ariyoshi, M. B. Bouzas, and R. W. Shafer analyzed the data. R Kantor, D. A. Katzenstein, P. Cahn, W. Sugiura, V. Soriano, L. F. Brigido, Z. Grossman, L. Morris, A.-M. Vandamme, A. Tanuri, P. Phanuphak, J. N. Weber, D. Pillay, P. R. Harrigan, R. Camacho, and R. W. Shafer interpreted the data. R. Kantor, D. A. Katzenstein, and R. W. Shafer contributed to writing the paper.
Citation: Kantor R, Katzenstein DA, Efron B, Carvalho AP, Wynhoven B, et al. (2005) Impact of HIV-1 subtype and antiretroviral therapy on protease and reverse transcriptase genotype: Results of a global collaboration. PLoS Med 2(4): e112.
Abbreviations
NRTInucleoside reverse transcriptase inhibitor
NNRTInon-nucleoside reverse transcriptase inhibitor
PIprotease inhibitor
RTreverse transcriptase
RTIreverse transcriptase inhibitor
==== Refs
References
Gao F Bailes E Robertson DL Chen Y Rodenburg CM Origin of HIV-1 in the chimpanzee Pan troglodytes troglodytes
Nature 1999 397 436 441 9989410
Hahn BH Shaw GM De Cock KM Sharp PM AIDS as a zoonosis: Scientific and public health implications Science 2000 287 607 614 10649986
Korber B Muldoon M Theiler J Gao F Gupta R Timing the ancestor of the HIV-1 pandemic strains Science 2000 288 1789 1796 10846155
Robertson DL Anderson JP Bradac JA Carr JK Foley B HIV-1 nomenclature proposal Science 2000 288 55 56 10766634
Osmanov S Pattou C Walker N Schwardlander B Esparza J Estimated global distribution and regional spread of HIV-1 genetic subtypes in the year 2000 J Acquir Immune Defic Syndr 2002 29 184 190 11832690
Gonzales MJ Machekano RN Shafer RW HIV-1 reverse transcriptase and protease subtypes: Classification, amino acid mutation patterns, and prevalence in a Northern California clinic-based population J Infect Dis 2001 184 998 1006 11574914
Dumans AT Soares MA Machado ES Hue S Brindeiro RM Synonymous genetic polymorphisms within Brazilian human immunodeficiency virus type 1 subtypes may influence mutational routes to drug resistance J Infect Dis 2004 189 1232 1238 15031792
Gonzalez LM Brindeiro RM Aguiar RS Pereira HS Abreu CM Impact of nelfinavir resistance mutations on in vitro phenotype, fitness, and replication capacity of human immunodeficiency virus type 1 with subtype B and C proteases Antimicrob Agents Chemother 2004 48 3552 3555 15328124
Shafer RW Eisen JA Merigan TC Katzenstein DA Sequence and drug susceptibility of subtype C reverse transcriptase from human immunodeficiency virus type 1 seroconverters in Zimbabwe J Virol 1997 71 5441 5448 9188616
Shafer RW Chuang TK Hsu P White CB Katzenstein DA Sequence and drug susceptibility of subtype C protease from human immunodeficiency virus type 1 seroconverters in Zimbabwe AIDS Res Hum Retroviruses 1999 15 65 69 10024054
Toni T Masquelier B Bonard D Faure M Huet C Primary HIV-1 drug resistance in Abidjan (Cote d'Ivoire): A genotypic and phenotypic study AIDS 2002 16 488 491 11834964
Palmer S Margot N Gilbert H Shaw N Buckheit R Tenofovir, adefovir, and zidovudine susceptibilities of primary human immunodeficiency virus type 1 isolates with non-B subtypes or nucleoside resistance AIDS Res Hum Retroviruses 2001 17 1167 1173 11522186
Apetrei C Descamps D Collin G Loussert-Ajaka I Damond F Human immunodeficiency virus type 1 subtype F reverse transcriptase sequence and drug susceptibility J Virol 1998 72 3534 3538 9557632
Adje C Cheingsong R Roels TH Maurice C Djomand G High prevalence of genotypic and phenotypic HIV-1 drug-resistant strains among patients receiving antiretroviral therapy in Abidjan, Cote d'Ivoire J Acquir Immune Defic Syndr 2001 26 501 506 11391173
Weidle PJ Malamba S Mwebaze R Sozi C Rukundo G Assessment of a pilot antiretroviral drug therapy programme in Uganda: Patients' response, survival, and drug resistance Lancet 2002 360 34 40 12114039
Vergne L Malonga-Mouellet G Mistoul I Mavoungou R Mansaray H Resistance to antiretroviral treatment in Gabon: Need for implementation of guidelines on antiretroviral therapy use and HIV-1 drug resistance monitoring in developing countries J Acquir Immune Defic Syndr 2002 29 165 168 11832686
Cane PA de Ruiter A Rice PA Wiselka M Fox R Resistance-associated mutations in the human immunodeficiency virus type 1 subtype C protease gene from treated and untreated patients in the United Kingdom J Clin Microbiol 2001 39 2652 2654 11427587
Pillay D Walker AS Gibb DM De Rossi A Kaye S Impact of human immunodeficiency virus type 1 subtypes on virologic response and emergence of drug resistance among children in the Paediatric European Network for Treatment of AIDS (PENTA) 5 trial J Infect Dis 2002 186 617 625 12195348
Frater AJ Dunn DT Beardall A Ariyoshi K Clarke JR Comparative response of African HIV-1-infected individuals to highly active antiretroviral therapy AIDS 2002 16 1139 1146 12004272
Grossman Z Vardinon N Chemtob D Alkan ML Bentwich Z Genotypic variation of HIV-1 reverse transcriptase and protease: Comparative analysis of clade C and clade B AIDS 2001 15 1453 1460 11504976
Kebba A Atwine D Mwebaze R Kityo C Nakityo R Therapeutic responses to AZT + 3TC + EFV in advanced antiretroviral naive HIV type 1-infected Ugandan patients AIDS Res Hum Retroviruses 2002 18 1181 1187 12487824
Landman R Schiemann R Thiam S Vray M Canestri A Once-a-day highly active antiretroviral therapy in treatment-naive HIV-1-infected adults in Senegal AIDS 2003 17 1017 1022 12700451
Alexander CS Montessori V Wynhoven B Dong W Chan K Prevalence and response to antiretroviral therapy of non-B subtypes of HIV in antiretroviral-naive individuals in British Columbia Antivir Ther 2002 7 31 35 12008785
Weidle PJ Downing R Sozi C Mwebaze R Rukundo G Development of phenotypic and genotypic resistance to antiretroviral therapy in the UNAIDS HIV Drug Access Initiative—Uganda AIDS 2003 17 Suppl 3 S39 S48 14565608
Djomand G Roels T Ellerbrock T Hanson D Diomande F Virologic and immunologic outcomes and programmatic challenges of an antiretroviral treatment pilot project in Abidjan, Cote d'Ivoire AIDS 2003 17 Suppl 3 S5 S15 14565604
Laurent C Kouanfack C Koulla-Shiro S Nkoue N Bourgeois A Effectiveness and safety of a generic fixed-dose combination of nevirapine, stavudine, and lamivudine in HIV-1-infected adults in Cameroon: Open-label multicentre trial Lancet 2004 364 29 34 15234853
Palmer S Alaeus A Albert J Cox S Drug susceptibility of subtypes A, B, C, D, and E human immunodeficiency virus type 1 primary isolates AIDS Res Hum Retroviruses 1998 14 157 162 9462926
Loemba H Brenner B Parniak MA Ma'ayan S Spira B Genetic divergence of human immunodeficiency virus type 1 Ethiopian clade C reverse transcriptase (RT) and rapid development of resistance against nonnucleoside inhibitors of RT Antimicrob Agents Chemother 2002 46 2087 2094 12069959
Caride E Brindeiro R Hertogs K Larder B Dehertogh P Drug-resistant reverse transcriptase genotyping and phenotyping of B and non-B subtypes (F and A) of human immunodeficiency virus type I found in Brazilian patients failing HAART Virology 2000 275 107 115 11017792
Caride E Hertogs K Larder B Dehertogh P Brindeiro R Genotypic and phenotypic evidence of different drug-resistance mutation patterns between B and non-B subtype isolates of human immunodeficiency virus type 1 found in Brazilian patients failing HAART Virus Genes 2001 23 193 202 11724274
Velazquez-Campoy A Vega S Freire E Amplification of the effects of drug resistance mutations by background polymorphisms in HIV-1 protease from African subtypes Biochemistry 2002 41 8613 8619 12093278
Ariyoshi K Matsuda M Miura H Tateishi S Yamada K Patterns of point mutations associated with antiretroviral drug treatment failure in CRF01_AE (subtype E) infection differ from subtype B infection J Acquir Immune Defic Syndr 2003 33 336 342 12843744
Brenner B Turner D Oliveira M Moisi D Detorio M A V106M mutation in HIV-1 clade C viruses exposed to efavirenz confers cross-resistance to non-nucleoside reverse transcriptase inhibitors AIDS 2003 17 F1 F5 12478089
Grossman Z Paxinos EE Averbuch D Maayan S Parkin NT Mutation D30N is not preferentially selected by human immunodeficiency virus type 1 subtype C in the development of resistance to nelfinavir Antimicrob Agents Chemother 2004 48 2159 2165 15155216
Grossman Z Istomin V Averbuch D Lorber M Risenberg K Genetic variation at NNRTI resistance-associated positions in patients infected with HIV-1 subtype C AIDS 2004 18 909 915 15060438
Lole KS Bollinger RC Paranjape RS Gadkari D Kulkarni SS Full-length human immunodeficiency virus type 1 genomes from subtype C- infected seroconverters in India, with evidence of intersubtype recombination J Virol 1999 73 152 160 9847317
Shafer RW Jung DR Betts BJ Human immunodeficiency virus type 1 reverse transcriptase and protease mutation search engine for queries Nat Med 2000 6 1290 1292 11062545
D'Aquila RT Schapiro JM Brun-Vezinet F Clotet B Conway B Drug resistance mutations in HIV-1 Top HIV Med 2003 11 92 96 12837968
Rhee SY Gonzales MJ Kantor R Betts BJ Ravela J Human immunodeficiency virus reverse transcriptase and protease sequence database Nucleic Acids Res 2003 31 298 303 12520007
Hue S Clewley JP Cane PA Pillay D HIV-1 pol gene variation is sufficient for reconstruction of transmissions in the era of antiretroviral therapy AIDS 2004 18 719 728 15075506
Benjamini Y Hochberg Y Controlling the false discovery rate: a practical and powerful approach to multiple testing J R Stat Soc Ser B 1995 57 289 300
Lan NT Recordon-Pinson P Hung PV Uyen NT Lien TT HIV type 1 isolates from 200 untreated individuals in Ho Chi Minh City (Vietnam): ANRS 1257 study. Large predominance of CRF01_AE and presence of major resistance mutations to antiretroviral drugs AIDS Res Hum Retroviruses 2003 19 925 928 14601592
Tovanabutra S Beyrer C Sakkhachornphop S Razak MH Ramos GL The changing molecular epidemiology of HIV type 1 among northern Thai drug users, 1999 to 2002 AIDS Res Hum Retroviruses 2004 20 465 475 15186520
Rose PP Korber BT Detecting hypermutations in viral sequences with an emphasis on G –> A hypermutation Bioinformatics 2000 16 400 401 10869039
Wu TD Schiffer CA Gonzales MJ Taylor J Kantor R Mutation patterns and structural correlates in human immunodeficiency virus type 1 protease following different protease inhibitor treatments J Virol 2003 77 4836 4847 12663790
Morris L Pillay C Chezzi C Lupondwana P Ntsala M Low frequency of the V106M mutation among HIV-1 subtype C-infected pregnant women exposed to nevirapine AIDS 2003 17 1698 1700 12853755
Gomes P Diogo I Goncalves MF Carvalho P Cabanas J Different pathways to nelfinavir genotypic resistance in HIV-1 subtypes B and G [abstract]. Ninth Conference on Retroviruses and Opportunistic Infections; 2002 February 24–28; Seattle, Washington. Abstract number 46. Available: http://www.retroconference.org//2002/Abstract/12937.htm
2002 Accessed 14 March 2005
Sugiura W Matsuda Z Yokomaku Y Hertogs K Larder B Interference between D30N and L90M in selection and development of protease inhibitor-resistant human immunodeficiency virus type 1 Antimicrob Agents Chemother 2002 46 708 715 11850252
Gonzales MJ Wu TD Taylor J Belitskaya I Kantor R Extended spectrum of HIV-1 reverse transcriptase mutations in patients receiving multiple nucleoside analog inhibitors AIDS 2003 17 791 799 12660525
| 15839752 | PMC1087220 | CC BY | 2021-01-05 10:39:46 | no | PLoS Med. 2005 Apr 26; 2(4):e112 | utf-8 | PLoS Med | 2,005 | 10.1371/journal.pmed.0020112 | oa_comm |
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 10.1371/journal.pmed.0020116SynopsisInfectious DiseasesInfectious DiseasesMedicine in Developing CountriesThe Complicated Art of Tracking Dengue Synopsis4 2005 26 4 2005 2 4 e116Copyright: © 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.
Nonstationary Influence of El Niño on the Synchronous Dengue Epidemics in Thailand
==== Body
Compared with malaria, dengue fever has a rather lower profile in the public mind, although to those who have had it, it leaves a great impression. The name dengue fever is derived from the Swahiliwords Ki denga pepo (“it is a sudden overtaking by an evil spirit”), which gives an idea of the rapid onset of the disease. The dengue virus is carried by the mosquito Aedes aegypti, and the disease often occurs as epidemics. Although the classic illness is a fairly benign acute febrile syndrome, it may be very painful—hence the English nickname, breakbone fever. The virus can also cause a much more serious illness known as dengue hemorrhagic fever, which can progress to dengue shock syndrome. There are four main serotypes of the dengue RNA virus; dengue hemorrhagic fever is more likely to occur during dengue infection in people with preexisting active or passive (e.g., maternally acquired) immunity who are exposed to a different dengue virus serotype. In contrast to classic dengue, the hemorrhagic fever and shock syndromes are mostly diseases of children and, if untreated, have a mortality of around 50%.
Aedes aegypti, the main vector of dengue (Photo: CDC/Robert S Craig)
Around two-fifths of the world's population are now at risk of the disease (one estimate is that 80 million people are infected each year). The number at risk will increase as population growth, urbanization, international travel, and climate change influence transmission of the disease. Understanding how all these factors interact is important in planning for disease outbreaks. However, the incidence of dengue is not easily predictable, varying with season, and also between years. For example, although dengue is most prevalent in the wet season, dengue epidemics have also been associated with drought in some countries. El Niño is the best known climatic event affecting climate between years, and some research already suggests that there is a relationship between the timing of dengue epidemics and El Niño in the Pacific Islands and in other countries.
Previous research has uncovered traveling waves of dengue in Thailand, but the cause of these has been obscure. In a paper in this month's PLoS Medicine Bernard Cazelles and colleagues looked at the details of the relationship between dengue incidence and El Niño in Thailand. Their results, based on complex mathematical analysis, do not provide easy answers for those who might want to plan for dengue outbreaks, though they do go some way to helping to understand the complex interplay between the various factors. In essence, the researchers found that there was a significant association between El Niño oscillations, climate variables, and dengue hemorrhagic fever incidence with a 2- to 3-year repeat, for both Bangkok and the rest of Thailand. However this association was significant only for the years 1986–1992, and outside these years factors other than climate were probably responsible for triggering the disease outbreaks.
| 0 | PMC1087221 | CC0 | 2021-01-05 10:39:45 | no | PLoS Med. 2005 Apr 26; 2(4):e116 | utf-8 | PLoS Med | 2,005 | 10.1371/journal.pmed.0020116 | oa_comm |
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 1583974710.1371/journal.pmed.0020120Correspondence and Other CommunicationsEmergency MedicineNeurology/NeurosurgeryOpthalmologyPrimary CareEmergency MedicineNeurologyOphthalmologyStrokeEnophthalmos Is Not Present in Horner Syndrome CorrespondenceDaroff Robert
1
1CASE School of MedicineCleveland, OhioUnited States of AmericaE-mail: [email protected]
Competing Interests: The author has declared that no competing interests exist.
4 2005 26 4 2005 2 4 e120Copyright: © 2005 Robert Daroff.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.
Painful Horner Syndrome As a Harbinger of Silent Carotid Dissection
==== Body
The case report by Nautiyal et al. [1] is an instructive reminder that the first episode of an acute painful Horner Syndrome should prompt imaging of the ipsilateral internal carotid artery, since carotid dissection (as well as other conditions, such as high-grade stenosis) needs to be ruled out. Unfortunately, the authors perpetuate the extremely common misconception that enophthalmos accompanies ptosis and miosis in human Horner Syndrome. It is only an illusion of enophthalmos caused by the ptosis. This is evident in the left eye of their patient in Figure 1 of the case report.
Actual measurement with exophthalmometry clearly demonstrates the lack of enophthalmos. As stated by Loewenfeld ([2], p. 1139), “Animals such as cats, rats, or dogs have enophthalmos on the side of the sympathetic lesion. But in man, the enophthalmos is only apparent. The small palpebral fissure makes the eye look sunken in on the affected side, but the position of the globe in the orbit remains virtually unchanged. This has been found by all workers who have measured the supposed enophthalmos objectively.” Loewenfeld cites four supportive references.
Thompson and Miller ([3], p. 964) provide four additional references that the enophthalmos “is apparent rather than real.”
Citation: Daroff R (2005) Enophthalmos is not present in Horner Syndrome. PLoS Med 2(4): e120.
==== Refs
References
Nautiyal A Singh S DiSalle M O'Sullivan J Painful Horner syndrome as a harbinger of silent carotid dissection PLoS Med 2005 2 e19 10.1371/journal.pmed.0020019 15696206
Loewenfeld IE The Pupil: Anatomy, physiology, and clinical applications, Volume 1 1999 Boston Butterworth-Heinemann 2 v
Thompson HS Miller NR Miller NR Newman NJ Disorders of pupillary function, accommodation, and lacrimation Walsh and Hoyt's Clinical Neuro-ophthalmology 1998 1 Baltimore Williams and Wilkins 961 1040 5th ed
| 15839747 | PMC1087222 | CC BY | 2021-01-05 10:39:49 | no | PLoS Med. 2005 Apr 26; 2(4):e120 | utf-8 | PLoS Med | 2,005 | 10.1371/journal.pmed.0020120 | oa_comm |
==== Front
Aust New Zealand Health PolicyAustralia and New Zealand Health Policy1743-8462BioMed Central London 1743-8462-2-71581119210.1186/1743-8462-2-7CommentaryAustralian public health policy in 2003 – 2004 Lin Vivian [email protected] Priscilla [email protected] School of Public Health, Faculty of Health Sciences, La Trobe University, Bundoora Victoria 3086 Australia2005 6 4 2005 2 7 7 3 9 2004 6 4 2005 Copyright © 2005 Lin and Robinson; licensee BioMed Central Ltd.2005Lin and Robinson; 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.
In Australia, compared with other developed countries the many and varied programs which comprise public health have continued to be funded poorly and unsystematically, particularly given the amount of publicly voiced political support.
In 2003, the major public health policy developments in communicable disease control were in the fields of SARS, and vaccine funding, whilst the TGA was focused on the Pan Pharmaceutical crisis. Programs directed to health maintenance and healthy ageing were approved. The tertiary education sector was involved in the development of programs for training the public health workforce and new professional qualifications and competencies. The Abelson Report received support from overseas experts, providing a potential platform for calls to improve national funding for future Australian preventive programs; however, inconsistencies continued across all jurisdictions in their approaches to tackling national health priorities. Despite 2004 being an election year, public health policy was not visible, with the bulk of the public health funding available in the 2004/05 federal budget allocated to managing such emerging risks as avian flu. We conclude by suggesting several implications for the future.
==== Body
Introduction
Public health is a small component of the health system, both in terms of budgetary allocation at either state or national level and in terms of the number of practitioners. It incorporates a myriad of activities; legislation and regulation for health protection, preventive services directed at specific diseases and populations, and health promotion programs geared towards particular risk factors and vulnerable groups in the community. As such, it looks like a disparate collection of programs and investments.
In Australia, there is also confusion about the very terminology of 'public health'. Despite its extensive history and global understanding, in Australia the term is used variously; to refer to publicly funded health services, and interventions (regardless of the funding source) which are aimed at primary prevention and the promotion and protection of the public health ('rats and drains'). This has led to an increasing number of jurisdictions adopting the label 'population health'.
Renovation of the public health system has been on the international agenda for some years. In the US, the Institute of Medicine released reports during 2003 about the public health workforce required for 21st century challenges [1], as well as re-visited and updated its landmark report, The Future of Public Health in the 21st Century[2]. In the UK, following the path-breaking review of the NHS by Derek Wanless[3] the Treasury commissioned him, in 2003, to undertake a review of whole-of-government effort in public health. Arising in part from the challenges that confronted Canada during the outbreak of sudden acute respiratory syndrome (SARS) in 2003, a new public health agency, at arms length from government, is being created.
Public health in Australia, meanwhile, remained fragmented – by programs, across jurisdictions (particularly the states and territories) – and without a systematic approach to funding, organisation, or conceptualisation. In 2003/04, the gap between rhetoric and funding continued to be noticeable, along with the tension between framing priorities for popular appeal versus the technical language of the evidence base.
This article will examine some of the indicative developments of public health in Australia in 2003/04. The key developments are identified, and a number of them are selected for in-depth analysis. In this article, we use the traditional meaning of the term 'public health' and focus on activities which are usually designed to promote and protect the health of the population. The drivers for these developments, their short term implications and some signposts for the future are suggested.
2003/04 in Retrospect: A brief chronicle
While early global anxiety over SARS occupied headlines between February and May, the more persistent popular headline in 2003 focused on obesity. Summits were held in NSW and Victoria, while the National Obesity Taskforce was convened under the auspice of the Australian Health Ministers Council (AHMC).
When Kay Patterson was the Federal Health Minister, she declared that prevention was the fourth pillar of Medicare and she wanted to be 'Minister for Prevention'. Indeed, the 2003/04 federal budget, although limited, contained a bundle of initiatives entitled "Prevention on the Health Agenda". In particular, a number of immunisation and health promotion programs were included.
Significant amongst the funding initiatives for public health announced in 2003/04 was government support for the meningococcal vaccine. Although this was the culmination of many months of careful planning, a perception existed that this only occurred after considerable public interest in and anxiety about deaths from outbreaks of this disease.
Further changes to the recommended schedule in 2003 were made by the Australian Technical Advisory Group on Immunisation (ATAGI), in particular the inclusion of pneumococcal and varicella vaccines; however, these did not result in similar prescribed vaccine programs or in similar funding. These three developments are reviewed in greater detail in the next section.
The National Public Health Partnership (NPHP) and the AHMC adopted the influenza pandemic plan in October 2003, and with the advent of the newly-identified disease SARS, as well as outbreaks of meningococcal disease, management and prevention of communicable diseases was prominent. Following on from the significant funding boost for bioterrorism preparedness in 2002/03, public health preparedness became a more generic theme.
The arrival of SARS occupied the national popular and political imagination as well as tested the infrastructure capacity of public health. Australia fared well during the outbreak. Apart from escaping with only six Australian cases, it provided an opportunity to establish a coordinated approach between the Commonwealth and the states/territories and also contributed to the global epidemiological investigation and prevention effort. SARS also prompted amendments to the Quarantine Act [4].
While the recall following the Pan Pharmaceutical crisis put the Therapeutic Goods Administration (TGA) under the spotlight, it also managed to conclude negotiations that had been in train for several years on a Trans-Tasman regulatory regime and authority. Also on the regulatory front, the Australian New Zealand Food Regulation Ministerial Council endorsed a nutrition, health and related claims policy guidelines and established a review of genetically modified (GM) labeling of foods [5]. All these developments pointed to the global nature of public health, and the intersection between public health activities and the economy.
Policy development in public health has never been confined to a set of health programs, and in 2003/04, the lead was often taken from outside the health sector. Most significant was the adoption of the National Agenda for Early Childhood [6], pushed by public health advocates for child health since the mid 1990s. The National Public Health Partnership responded by coordinating a scoping of child health strategies across Australia. Elsewhere in Government, "Promoting and Maintaining Good Health" was adopted as one of the National Research Priorities [7]. Healthy ageing also emerged as a policy theme in Ageing Research.
Public health workforce development was pursued outside the mainstream education and training arrangements for public health in universities. The Community Services and Health Training Board commissioned a consultative process to develop population health competencies for the Vocational Education and Training (VET) sector [8]. New population health qualifications and competencies were proposed for incorporation into the Health Training Package – including certificates in population health and in environmental health, and diplomas in population health and in indigenous environmental health.
The release in 2003 of the report "Returns on Investments in Public Health: an epidemiological and economic analysis" [9] (often referred to as the Abelson report), may have a significant impact in subsequent years. Commissioned several years earlier by the Population Health Division of the Department of Health and Ageing (DoHA), the report experienced a relatively low profile until Derek Wanless visited from the UK. Having chaired a review that contributed to a significant budgetary increase for the NHS, Wanless had been commissioned by the British Treasury to examine prevention across government. In September 2003, at a meeting in Canberra with senior officials across key agencies, Wanless marveled at the value of the Abelson report, described in more detail below.
Although 2004 was an election year, public health policy was neither visible during the campaign or in policy development more generally. The Federal Government's initiative to wind up the National Occupational Health and Safety Commission received little publicity and comment, even though it indicated the Commonwealth's increasing tendency to pursue its own pathway, separate from states and territories, and to bring the functions of statutory bodies into departments.
Jurisdictional and annual reports show that across the states and territories, there were multiple plans, draft guidelines, meetings, episodic training and programs across a broad range of areas. Some health issues are being taken up across jurisdictions – particularly tobacco control, sexually transmitted infections, Aboriginal health, and vaccination. Innovative activities were reported in some jurisdictions, such as a new Health Impact Assessment Branch and a new public health training program in Western Australia. There was, however, no apparent consistency in health priorities across the nation, and an apparent divergence in the interests of the states/territories and the federal government.
Obesity: Old or new frontier for health promotion?
While the "prevention and management of overweight and obesity" agenda may have appeared to many observers as a new issue in 2003, its arrival was preceded by several years of intensive work. The NHMRC had released Acting on Australia's Weight: Strategic plan for the prevention of overweight and obesity in 1997 [10], the same year the ABS published the findings from the 1995 National Nutrition Survey, revealing that 45% of men and 29% of women in Australia were overweight, with an additional 18% of men and women classified as obese [11]. Furthermore, overweight and obesity were more common in lower socio-economic groups, in rural populations, in some immigrant groups, and in Aboriginal and Torres Strait Islander (ATSI) peoples.
Despite longstanding national cooperation on nutrition (since the days of the National Better Health Program in the late 1980s), and even more recent national cooperation on physical activity, public and political imagination was not captured until the same issues were recast as 'obesity', with a focus in particular on childhood obesity. Following from the NSW Childhood Obesity Summit in late 2002, the Australian Health Ministers agreed that a national approach was required and established a National Obesity Taskforce [12].
In 2003, NSW Health released it's response to the Summit recommendations and supported the vast majority of the 145 resolutions [13]. The Victorian Department of Human Services also held a summit [14], while Healthy Weight 2008 – Australia's Future was released by the Commonwealth [15]. The NHMRC joined in with release in late 2003 of clinical practice guidelines for general practitioners and other health professionals [16].
While the specifics vary, the major themes and strategies are captured in Healthy Weight 2008. These are summarised in the Table 1.
Table 1 Major themes and strategies in public health weight programs
A. SETTINGS KEY STRATEGIES
Child care Good practice standards that incorporate physical activity guidelines and dietary guidelines for children
Schools As above plus active transport to school and programs to reduce excessive TV watching and computer games
Primary care services Support GPs to screen body mass index and implement lifestyle scripts
Community-based support programs for management of overweight
Family and community care services As per childcare settings
Maternal and infant health Extend healthy eating and active living programs; breastfeeding support programs; disseminate information resources for parents; 'baby friendly' accreditation for hospitals and health services;
Neighborhoods and community organisations Healthy eating and active living initiatives within existing programs; support improved physical and infrastructure planning; develop good practice 'tools'
Workplaces Support active transport programs; healthy eating and active living support for parents with young children
Food supply Accreditation for food service outlets; cold chain management initiatives; encourage reduction of energy content and size of servings in food industry
Media and marketing Monitor effectiveness of Children's Television Standards
B. NATIONAL ACTION KEY STRATEGIES
Support for families and community-wide education Social marketing; promotion of fruits and vegetables; national awards for settings-based programs
'Whole of Community' demonstration areas Demonstration projects, with professional support unit and clearinghouse; dissemination and professional development strategy
Evidence and performance monitoring Surveillance system and tracking indicators; policy research
Coordination and capacity-building Leadership program for obesity prevention; support professional networks for dissemination of good practice
The Commonwealth strategy is, however, relatively weak on intersectoral policy and regulatory measures. As an illustrative example of the contrast at the state level, implementation in NSW now ranges from school physical activity and nutrition survey, to a school canteen strategy, to negotiating with Commercial Television Australia about their code of practice on advertising in peak children's viewing hours. The Commonwealth apparently chose not to consider how it might exercise its relevant taxation or legislative powers, despite the history of health promotion pointing to the importance of public policy measures beyond the health system.
An examination of the manner in which the obesity issue was framed, and the details contained in the national strategy, raises a number of issues and questions:
- Why was framing the issues as 'obesity' more successful than the focus on 'nutrition' and 'physical activity'? Why did 'obesity' gain traction while the other terms did not?
- Why did the Commonwealth opt for the softer programmatic approach, rather than tackle obesity with stronger public policy measures (such as taxation and regulation), and demonstrate its national leadership capacity?
- Was the absence of stronger public policy measures because 'obesity' is regarded as largely a health issue, rather than a whole-of-government issue? Or was the Government waiting to see if the US opposed the WHO Global Strategy on account of the strength of the industry lobby?
- After a number of years of public concern about eating disorders and whether they arise in part because of promotion of certain types of body image, was the 'obesity' label a backward step for mental health and a return to traditional images of beauty?
- Is there a risk that people, including children, who are labeled as 'overweight and obese' will be stigmatised? To what extent have the voices of affected communities been incorporated into the development of national strategies, if at all?
- Given the correlation between obesity and socioeconomic disadvantage, how would the proposed strategy not exacerbate those inequalities?
- Were children targeted because they are a "captive audience" and therefore easy targets or did the evidence suggest the best return on investment (in terms of health gain and managing demand on the health care system) would come from a focus on children?
- Was the move to appeal to a populist agenda, while simultaneously progressing the longer-term agenda of tackling health inequalities through multi-sectoral partnerships, a triumph for public health advocates?
These complex threads are interwoven. For the moment, the publicly enunciated agenda represents a confluence of a number of rationales.
Vaccines: From evidence base to funding
During 2003–4 three new vaccines were added to the schedule of recommended vaccines for Australians (an additional change to the schedule, recommending that polio immunisation be changed from oral to injected (IPD) vaccine, will not be discussed here). These vaccines protect against serogroup C meningococcal disease, some strains of Pneumococcal disease, and chicken pox (varicella) [17]. For the first time, not all of these recommended vaccines will be funded by Government.
Prior to the introduction of these vaccines, the quality of information about the epidemiology and burden of disease caused by these three infections was extremely variable. Meningococcal disease has been notifiable for many years, and in Australia almost all is caused by serogroups B and C. Whilst serogroup B predominantly occurs in young children, a new strain of serogroup C [18] was causing increasing anxiety amongst public health professionals, microbiologists, staff of accident and emergency departments, intensive care units and of course the public and media.
The cause of anxiety amongst health professionals was based on the fact that this new strain carried a high fatality rate with severe after-effects in a high proportion of survivors. The attack rate, although still small, was increasing exponentially each year and reaching an important trigger point, and the majority of cases were now healthy teenagers and young adults. Although an initial accelerated catch-up programme was introduced for teenagers (the major risk group), the new conjugated vaccine was also introduced to the childhood schedule at age one, as from that age, only one dose (at a cost of $30–$60) was considered necessary for full protection from serogroup C disease.
Pneumococcal disease became notifiable in 2001, however, with such a short surveillance history, not much is certain locally, epidemiologically speaking, about risk groups and effects (although there is no reason to suppose that it has a different epidemiological pattern from other developed countries). Pneumococcal disease is thought to occur at least four times as often as meningococcal disease, is known to carry major sequelae and has a high case fatality rate. For some time it has been known to be even more common amongst the indigenous Australian population with attack rates of up to 1 in 500 each year, knowledge which underpinned the 1999 decision to target Aboriginal people for free vaccination as soon as the new vaccines became available. Unfortunately at about $120 per dose, conjugate pneumococcal vaccine is very expensive and, for the protection of the very young children who bear the brunt of this disease, it is licensed only to be given as a three dose course, making provision of this vaccine to all Australian children prohibitively expensive.
Varicella, predominantly a childhood disease, is caused by a Herpes virus known as herpes virus 3 or varicella-zoster virus or VZV. It is not notifiable in Australia; therefore no epidemiological population data are available. A reliable varicella vaccine has been available since the mid 1990s in the USA and is part of American routine immunisation schedule. This vaccine became available in Australia in 2000, at a cost of about $75–$90 per dose, with two doses being required for full protection.
In 2003 the Commonwealth provided its periodic update on the Australian Standard Vaccination Schedule, the list of vaccines it provides as appropriate at no cost to all Australians [19]. For the first time it differed from the National Immunisation Program recommendations in that besides meningococcal serogroup C conjugate vaccines, pneumococcal vaccine, varicella vaccine and also inactivated polio (injected) vaccine were also recommended: however, funding was only secured for meningococcal conjugate vaccines, with a continuation of the provision of pneumococcal vaccines for indigenous children. As a result, although recommended, pneumococcal and varicella vaccines were not funded and parents would have to decide whether or not to pay for them.
These funding decisions had important implications. Vaccines protect most of their recipients from unpleasant and sometimes life-threatening disease. One view, subscribed to in the UK, is that ethically, children should not be denied access because of their parents' inability to pay. These vaccines have been the subject of several cost-benefit studies, with generally favourable to extremely favourable pro-vaccination results. Table 2 summarises the various models for framing policy.
Table 2 New VPD awareness matrix
Argument – do we have good epidemiological evidence What are the political risks in not funding this vaccine? What are the risks perceived by health professionals What are the perceived risks and outrage by general public Ethics – what is in it for the stakeholders?
Meningococcal sg C disease Yes: notifiable disease for many years – good detailed and longitudinal evidence High political risk; Low Public frightened- recent high level of awareness amongst public, news coverage biased ++ to worst cases [26] Much public support for gvt Vaccine provider contracts
Pneumococcal disease Some: notifiable since 2000 so some local evidence, more from published materials from overseas Med-low: public not highly aware of significance in children High Public not anxious; news highlights occasionally but less general awareness (See * below) Some public support for gvt Vaccine provider contracts
Varicella Little – not notifiable Low – viewed by many people as an insignificant and mild disease of childhood Med Very little; whilst parents know this to be an unpleasant disease there is a general lack of awareness of complications, and vaccine not considered a high priority [27] Vaccine provider contracts
(*Whilst there are several papers about the reasons older people, their families and health care providers use pneumococcal vaccines, there do not seem to be any published peer-reviewed studies of parental understanding of pneumococcal disease). Source [26] [27].
The policy of funding meningococcal serogroup C vaccine was built on a sustained program of epidemiological evidence, ethical decision-making and public support (and was arguably honed by public pressure). Pneumococcal disease and varicella vaccination programs however, were neither supported by good local epidemiological evidence nor respectable levels of public awareness about these diseases. There had not been a similar program of sustained policy building to support or drive a decision to fund these vaccines. As a funding policy, this was noteworthy in that it marked a departure from previous policies where all recommended vaccines were fully funded by governments. National vaccination policy is designed to advise vaccination policy makers and practitioners of the most up-to-date thinking about optimal vaccination schedules for Australian children, and is not therefore proscriptive, unlike the United Kingdom (UK). Changing or adding vaccines to the recommended schedule is therefore an advisory matter, and the question of funding the vaccination program is decided separately.
Cost benefit studies indicate pneumococcal polysaccharide and conjugate vaccines can be cost-effective although vaccine costs clearly affect ratios of cost to benefit greatly [20,21]. Varicella vaccine is more contentious, because this disease is more severe in older cases, and it is possible that one result of a vaccination program could be an increase in older cases (and therefore severe disease). Whilst the vaccine undoubtedly works, there is no consensus about precisely who should be vaccinated for maximum population health as well as cost benefit, and again potential financial savings are highly dependant upon vaccine costs [22,23].
The costs of preventive vaccine programs and curative medicine are funded from different sources. Vaccines are currently funded by the Commonwealth and subsidised through the states according to local vaccination policies, whilst the costs of curing cases of these diseases is broadly funded through the Medicare and private health insurance systems. Savings to Medicare and health insurance funds, as a result of successful vaccination programs, are not automatically transferred to the Commonwealth to fund the vaccine programs. Savings – or costs – in one area are of little interest or importance to other program areas.
In 2004 the Government revised this funding policy, providing funding for conjugate pneumococcal vaccines population immunisation program for all children under seven years of age (as well as specific people in other risk groups) to commence in January 2005. The Australian Technical Advisory Group on Immunisation (ATAGI) completed Ministerial reports on both varicella and polio (injected as well as oral) vaccination late in 2004, and it is possible that programs for these vaccines will also be funded in the future.
Federal budget: Prevention on the Health Agenda?
The 2002/2003 Federal Budget papers stated that "the Government is committed to making disease prevention and health promotion a fundamental pillar of the health system": however, this was not evident in the subsequent 2003/2004 budget. The Government's Focus on Prevention Package in 2002/03 aimed to incorporate disease prevention into the core business of the primary health care system and was reflective of how the public health agenda was evolving at the national level [24]. The package was comprised largely of a range of measures directed at specific diseases, plus a bundle of initiatives for general practitioners, also referred to as the "primary health care system".
Amongst health conditions affecting Australians, breast cancer received the most attention, with the National Breast Cancer Centre being funded to develop a partnership approach to the review and dissemination of new information, along with information, support and management initiatives for rural women diagnosed with breast cancer. Hepatitis C also received some attention, with funding for national education and prevention projects. Financial support was offered for the SARS efforts that had been undertaken by states and territories, in particular for providing medical personnel at international airports. A clear process for assessing priorities under the broad banded National Public Health Program was also flagged.
For purposes of the budget, primary health care was defined as general practitioners, and the measures funded included:
• "Lifestyle prescriptions" to help GPs "raise community awareness and understanding of benefits of preventive health";
• Collaborative approach to learning, training education and support systems;
• Coordinated care plans for people with chronic or terminal conditions; and
• Involvement in multidisciplinary case conferencing.
The budget did not adopt a comprehensive approach to the primary health care system, perhaps because many community health services, which represent the other important arm for delivery of public health services, are the responsibility of states. The timetable for renewing Public Health Outcome Funding Agreements (PHOFAs) between the Commonwealth and states and territories in 2004 raised in the minds of some stakeholders, the possibility that the Commonwealth might adopt a more comprehensive and strategic approach, linking public health and primary health care funding streams.
Judging by the actual quantum of funds made available in the 2003/2004 budget, it would seem that most elements from the package did not actually receive additional funding, as shown in Table 3. Indeed, many of the GP initiatives, previously cast as improving primary health care, were subsequently packaged as 'prevention'.
Table 3 Additional federal funds (in millions) for new public health activities between 2003–2007
INITIATIVES 2003/04 2004/05 2005/06 2006/07
Community awareness 2.1 1.5 0.7 -
Primary care providers working together 2.7 4.6 4.6 4.5
Priority setting - - - -
SARS 1.7 - - -
National Breast Cancer Centre - - - -
Support for women with breast Cancer - - - -
Hep C prevention and education - - - -
Sharing healthcare - - - -
Annual health assessment for older Australians - - - -
Meningococcal C campaign 1.3 0.4 0.4 0.4
Preventing falls in older people - - - -
Coordinated care planning - - - -
Multidisciplinary case conferencing - - - -
Enhanced divisional quality use of medicines 10.9 17.0 19.2 21.6
GP education, support and community linkages - 1.4 - 1.7 - 1.8 - 1.8
Source: [28]
The combination of these measures reflected a tight fiscal climate, with little growth in the overall health budget, as well as that of other portfolios. It was also a package that demonstrated relatively limited imagination, with support for established issues (such as breast cancer) and re-packaging general practice measures that were already in train. With Medicare spending "uncapped" (and targeted public health programs "capped"), attaining more prevention dollars through the GP sector may appear to be one of the few ways to 'grow' dollars for prevention. Although this could be considered to be consistent with the Ottawa Charter of "reorienting health services", many GPs are not trained in a population-based approach to practice, and simply providing new for payments to all represents an undifferentiated, uncoordinated and untargeted approach to prevention. If there is limited support to GPs, and little monitoring, then these measures are unlikely to translate into improved health outcomes.
Meanwhile, the strategic framework for chronic disease prevention, adopted by the National Public Health Partnership in 2001, still lacked a policy and budgetary response in 2003/04 and in 2004/05, while states/territories adopting varying measures singly.
Funding for the Tough on Drugs strategy was announced outside the Focus on Prevention package; perhaps due to the fact that the Tough on Drugs was the responsibility of the Parliamentary Secretary therefore requiring a separate communications strategy, or because the Prime Minister has a strong personal interest in the illicit drug strategy. The range of measures funded (which included introduction of retractable needle and syringe technology, addressing problems related to increased availability and use of psycho stimulants, establishing a research fund, supporting alcohol and drug workforce development needs, promoting access to drug treatment in rural areas, and tackling problems faced by drug users with concurrent mental health problems) certainly suggested more serious government interest and commitment to illicit drugs.
The 2004/05 Budget indicated the Federal Government's agenda in public health had narrowed considerably. $33 million new funding was announced for measures to address emerging risks (such as emergency medicines stockpile, disease surveillance and public health laboratories, health security legislation and incidence response), but only $5.2 million new funding was made available for promotion of healthy lifestyles related to national health priorities (for addressing such as tobacco, alcohol, drugs, injury, and cancer). [25]
Drivers for Policy and Implications for Public Health
During the course of the Howard Government, there has been a gradual process of re-casting the "landscape" of interest groups and policy constituencies. Strong support for breast cancer and zero-tolerance on illicit drugs contrasts sharply with the delays experienced in renewal of the National HIV/Hepatitis C Strategy. The new prominence given to meningococcal vaccine, child health and obesity creates space for other interest groups: even if the re-framing was shaped by nutrition and physical activity lobbies, other clinical interests have been brought into the picture. These developments illustrate how 'political' considerations are important in determining 'public health policy'.
It was interesting however, to observe the interest in prevention from outside the health portfolio, particularly from Treasury. This was motivated in part by the Intergenerational Report and concerns about both the sustainability of Medicare as well as the social and economic cost burden arising from an ageing society. This helped to ensure interest in the Abelson Report[9].
Few countries have conducted research on return of investment from prevention efforts. Australia was praised by Derek Wanless at a high-level consultation for completing such an analysis, during his visit to Canberra while conducting a review for the UK Treasury, "Securing Good Health for the Whole Population"[26]. His final report pointed to Australia and Netherlands as two countries that were increasingly using economic evaluation in public health programs. It will be interesting to see if public health policy analysts and Treasury officials draw on this report in future years. In the future it will be interesting to see if the focus on high-visibility programs can demonstrate short-term economic returns.
Given 2004 was an election year, the "political economy" of prevention programs could arguably have become a focus of future public health policy, with the 2003/4 agenda providing the Government with the opportunity to gauge public reaction to this new positioning and design their election campaign appropriately. This was, however, not the case. The American emphasis on 'preparedness' appears not to resonate with the Australian public in the same way.
From the perspective of public health policy advocates, some lessons that can be drawn from 2003/04 are:
• Government's response to public health proposals are shaped by its understanding of the popular interest and desire to communicate directly with the general public;
• Longer term public health issues which have struggled to gain support can be progressed if they are cleverly shaped to fit the Government's "formula";
• Develop and nurture new advocates, particularly in seeking to engage with the broader health system; and
• Work with the media as partners rather than adversaries
These lessons need to be learned well and quickly, to assist with moving the forum for public health policy debate more into the public domain; beyond an essentially "in house" discourse between politicians, researchers and public health advocates. If a more engaged and informed community takes up a public health issue, government will be more likely to respond.
List of abbreviations
ANTA – Australian National Training Authority
AHMC – Australian Health Ministers Council
ATAGI – Australian Technical Advisory Group on Immunisation
ATSI – Aboriginal and Torres Strait Islander
DoHA – Department of Health and Ageing
GM – genetically modified (foods)
GP – general practitioner
NHMRC – National Health and Medical Research Council
NPHP – National Public Health Partnership
PHOFA – Public Health Outcome Funding Agreement
SARS – Sudden Acute Respiratory Syndrome
TGA – Therapeutic Goods Administration
VET – Vocational Education and Training
WHO – World Health Organization
UK – United Kingdom
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
VL conceived the format of this article and contributed the sections on policy, governance and finance. PR contributed the sections on communicable disease control and vaccination strategies.
Acknowledgements
Staff of the Faculty of Health Sciences, La Trobe University for useful discussion and comments of previous drafts of this paper, and Sheryll Kay for research assistance.
==== Refs
Institute of Medicine Who will keep the public healthy?
Institute of Medicine The Future of Public Health in the 21st Century
Wanless D Securing Our Future Health: Taking A Long-term View
Australian Government Department of Health and Ageing
Australian Government Department of Health and Ageing Newsletter of Food Standards Australia New Zealand Food Standards News
Australian Government Commonwealth Department of Family and Ageing Towards the Development of a National Agenda for Early Childhood
Australian Government Department of Science Education and Training Promoting and Maintaining Good Health
Community Services and Health Training Board Population health
Abelson P Returns on investment in public health
National Health and Medical Research Council Acting on Australia's Weight: Strategic plan for the prevention of overweight and obesity
A.B.S. National Nutrition Survey: Nutrient Intakes and Physical Measurements Australia
N.S.W. Health Childhood Obesity Summit Communique
N.S.W. Health N.S.W. Childhood Obesity Summit Government Response
Department of Human Services V A Healthy balance - Victorians respond to obesity
Australian Government Department of Health and Ageing Healthy Weight 2008 - Australia's Future
National Health and Medical Research Council Guidelines for the development and implementation of clinical practice guidelines
National Health and Medical Research Council Australian Immunisation Handbook, 8th edition, 2004
Tribe D Zaia A Griffith J Robinson P Li H Taylor K Hogg G Increase in meningococcal disease associated with the emergence of a novel ST-11 variant of Serogroup C Neisseria meningitidis in Victoria, Australia, 1999-2000. Epidemiology & Infection, 2002 128 7 14 11895093
National Health and Medical Research Council The Australian Standard Vaccination Schedule
Darkes M.J.M. PGL Pneumococcal vaccine (Prevenat[TM]; PNCRM7): a review of its use in the prevention of Streptococcus pneumoniae infection. Pediatric drugs 2002 4(9) 609 630 12175274
Ray GT Butler JC Black SB Shinefield HR Fireman BH Lieu TA Observed costs and health care use of children in a randomized controlled trial of pneumococcal conjugate vaccine. Paediatric Infectious Diseases Journal 2000 21(5) 361 365
Thiry MI Beutels P van Damme PI van Doorslar E Economic evaluations of varicella vaccination programmes: a review of the literature. Pharmacoeconomics 2003 21(1) 13 38 12484801
Brisson M Edmunds WJ Varicella vaccination in England and Wales: cost-utility analysis. Archives of Disease in Childhood 2003 88(10) 862 869 14500303 10.1136/adc.88.10.862
Australian Government Department of Health and Ageing Annual Report 2002-3
Wanless D Securing Good Health for the Whole Population
Irwin DJ Miller JM Milner PC Patterson T Richards RG Williams DA Insley CA Stuart JM Community immunisation programme in response to an outbreak of invasive Neisseria meningitidis serogroup C infection in the Trent region of England 1995-1996. Public Health Med 1997 19(2) 162 170
Taylor JA Newman RD Parental attitudes towards varicella vaccination. Archives of Pediatric and Adolescent Medicine 2000 154 (3)) 302 306
Australian Government Department of Health and Ageing: Focus on Prevention
| 15811192 | PMC1087471 | CC BY | 2021-01-04 16:38:28 | no | Aust New Zealand Health Policy. 2005 Apr 6; 2:7 | utf-8 | Aust New Zealand Health Policy | 2,005 | 10.1186/1743-8462-2-7 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-701579039610.1186/1471-2105-6-70SoftwarePaircomp, FamilyRelationsII and Cartwheel: tools for interspecific sequence comparison Brown C Titus [email protected] Yuan [email protected] Eric H [email protected] R Andrew [email protected] Division of Biological Sciences, California Institute of Technology, Pasadena, CA 91125, USA2 Center for Computational Regulatory Genomics, California Institute of Technology, Pasadena, CA 91125, USA2005 24 3 2005 6 70 70 18 11 2004 24 3 2005 Copyright © 2005 Brown 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
Comparative sequence analysis is an effective and increasingly common way to identify cis-regulatory regions in animal genomes.
Results
We describe three tools for comparative analysis of pairs of BAC-sized genomic regions. Paircomp is a tool that does windowed (ungapped) comparisons of two sequences and reports all matches above a set threshold. FamilyRelationsII is a graphical viewer for comparisons that enables interactive exploration of several different kinds of comparisons. Cartwheel is a Web site and compute-cluster management system used to execute and store comparisons for display by FamilyRelationsII. These tools are specialized for the discovery of cis-regulatory regions in animal genomes. All tools and their source code are freely available at .
Conclusion
These tools have been shown to effectively identify regulatory regions in echinoderms, mammals, and nematodes.
==== Body
Background
Comparative sequence analysis is fast becoming a standard method for discovering cis-regulatory modules [1]. The technique relies on the signatures of conservation left by functional genomic regions as the background sequence evolves. It is often the only way to computationally discover cis-regulatory modules in animal genomes when definite knowledge of upstream regulators is lacking, and it can serve as an excellent complement to experimental techniques.
Paircomp, FamilyRelationsII (FRII), and Cartwheel are an integrated system for comparing two BAC-sized (~100 kb) genomic sequences, viewing the comparison, manipulating thresholds and views, and extracting the results. These tools and their predecessors, seqcomp and FamilyRelations, have been used extensively in the years since we first made them available [2]. However, the addition of Cartwheel, a Web server system for performing, storing, and revisiting analyses, makes this combined toolkit considerably more useful to the experimental biologist.
The first analysis done with FamilyRelations was a comparison of the otx region between two sea urchins; 11 of the 17 conserved blocks were shown to drive expression of a reporter [3]. Kirouac and Sternberg [4] showed that features conserved between C. elegans and C. briggsae encode functional regulatory regions. Romano and Wray [5] used FamilyRelations to show that primary sequence identity was conserved in only part of the previously identified endo16 cis-regulatory region, when the L. variegatus sequence was used as a partner to the S. purpuratus sequence. Leung et al. [6] used FRII to analyze regions in which NFKB bound to verify that the regions were conserved between mouse and human. And, most recently, Revilla-i-Domingo et al. [7] identified a small conserved region in the delta genomic locus as a cis-regulatory element responsible for localized expression of delta in S. purpuratus. Similar analyses of the regulation of gatae, krox, wnt8, brachyury, tbrain, foxa and deadringer in S. purpuratus are forthcoming from this lab. While most published use of FRII and Cartwheel has been in sea urchins and nematodes, users have reported that the tools accurately identify regulatory regions in vertebrates and plants.
FRII and Cartwheel are specialized for identifying conservation within relatively small genomic regions, and can be used for comparing BAC sequences between organisms for which no whole genome assembly exists (e.g. S. purpuratus/L. variegatus). The exhaustive "dot-plot"-style search algorithm used (described below) assumes nothing about the relative positioning or orientation of regulatory regions and can be used to detect rearrangements that might be missed by a global alignment algorithm (see e.g. [4]). Because of these features, FRII and Cartwheel are particularly useful in targeted searches for regulatory regions.
In this paper, we present these effective tools for comparative sequence analysis to the wider biological community.
Implementation
Paircomp is a program for doing windowed comparisons of two sequences. It is an expanded reimplementation of the seqcomp program [2]. Paircomp contains several algorithms for doing exhaustive fixed-width-window sequence comparisons, optimized for different parameters. The default algorithm uses a sliding window to do a "rolling comparison" and runs in time O(NxM) for two sequences of lengths N and M. Paircomp is written in C++ and has a Python interface.
FamilyRelationsII (FRII) is a graphical viewer for sequence analyses. It is a C++ reimplementation of the original Java/Jython FamilyRelations [2]. FRII uses the cross-platform FLTK windowing toolkit to present a common interface on Windows, Mac OS X, and Linux/X11.
Cartwheel is a server-side system that presents a uniform interface for job coordination and execution. It has several components, including a Web interface through which users can establish analyses; a remote interface for programs to retrieve analysis data; and a batch job queueing system based on a method of parallel processing known as a Linda tuple space. All of the components are built on top of a PostgreSQL database. Cartwheel is written in Python and provides libraries in Python, Java, and C++ for remote access.
A technical history of the design decisions made in the implementation of these tools has been published online ([8], article "Python in Bioinformatics").
Availability
FRII is freely available for download in a binary distribution for Mac OS X and Windows [9]; FRII will also run under most UNIX distributions but must be compiled individually. The Center for Computational Regulatory Genomics at Caltech maintains a public Cartwheel server [10]. A tutorial for FRII is available online [11], and an example homework assignment for an undergraduate class is also available. The source code for paircomp, FRII and Cartwheel and all their components is freely available under the L/GPL through the above Web sites. Paircomp, FamilyRelationsII and Cartwheel are Copyright © 2001–2004 the California Institute of Technology.
Results and discussion
Paircomp
Several different classes of algorithms are available for comparing two genomic sequences. Windowed comparisons do an exhaustive comparison of two sequences with a fixed-width window, and record strict (ungapped) sequence identity within that window [2,12]. Local alignment algorithms such as BLAST search for common "words" of DNA in a pair of sequences and build a gapped alignment around these words [13]. These gapped alignments are often scored by overall length, so that e.g. a 500 bp match at 90% is ranked higher than a 200 bp match at 90%. Global alignment algorithms such as AVID [14] and LAGAN [15] seek to build a start-to-end gapped alignment of syntenic genomic regions. Windowed comparisons and local alignment algorithms usually search for matches in both forward and reverse complement directions, while global alignment algorithms typically try to build an alignment without inversions. Implementations of all three strategies for genomic comparisons have been publicly available for some time: Dotter and seqcomp implement windowed comparisons [2,12]; PipMaker uses a local alignment algorithm, blastz [16,17]; and Vista relies on a global alignment generated by AVID [18]. All three comparison strategies have been successful at finding regulatory regions [1,19].
Of the three general classes of algorithms, we chose to use windowed comparisons in our search for cis-regulatory modules. Our decision was based on several criteria. First, these comparisons report matches based solely on strict sequence identity with no gapping, unlike alignment algorithms. This is a good ab initio requirement when comparing sequences in search of cis-regulatory modules, whose evolution is still poorly understood; in particular, binding sites could be sensitive to indels, which are somewhat elided in gapped alignments. Moreover, we had no a priori expectation for the locations, sizes, or degrees of similarity of conserved regions, necessitating an exhaustive search strategy that did not bias scores based on the length or position of matches. And, finally, from a user-interface perspective the parameters for paircomp – windowsize and threshold – are simple and intuitively linked to the results. Our success with this basic approach means that we have not needed to move to alternative algorithms.
Paircomp is a standalone program that executes windowed comparisons (see Methods). It searches for matches in both the forward and reverse complement directions. Paircomp runs within Cartwheel; the results are stored in a database and communicated to FRII.
Cartwheel
Cartwheel is a Web site through which analyses are executed and from which analyses are loaded into FamilyRelationsII. It provides an easy-to-use interface through which to establish a set of analyses on a pair of sequences. Cartwheel also allows the annotation of sequences with a variety of features; features can be uploaded to Cartwheel in the standard GFF format. A tutorial for setting up pairwise comparisons is available online [11].
FamilyRelationsII
FamilyRelationsII, or FRII, displays comparisons of BAC-sized genomic sequences of lengths ~100 kb. It is a graphical program that runs directly from a desktop and loads data from the Cartwheel server. From within FRII, users can zoom in to look more closely at features, alter scoring thresholds for comparisons, change the color of features, and turn on or off the display of specific analyses. FRII can also display closeup views of comparisons and alignments against DNA and protein sequence.
Figure 1 shows the main FRII view of a comparison between the otx locus in S. purpuratus and L. variegatus, two sea urchins that diverged approx. 50 mya. The genomic sequences were obtained from BAC libraries as described in [3]. In the case of S. purpuratus, the BAC contains the entire otx coding region; the L. variegatus sequence contains only the 5' region of the gene, and not the final exon.
The comparison shown is a paircomp comparison performed with a 20 bp window at 90% and then displayed at a 95% threshold. The general colinearity of the matches suggests that the majority of the similar regions are conserved with respect to size, orientation, and relative distance from the exons. This colinearity is typical of conserved features in our comparisons. The diagonal lines crossing the comparison often identify low complexity regions such as simple sequence repeats present throughout both genomic regions. This pairwise mapping view is one of the two large-scale views in FRII; the other large-scale view is a dot-plot view, shown in Figure 2.
Figure 2 shows a dot-plot view of an expanded region of the comparison, centered on the first exon of the α-otx transcript. In addition to the exon itself, there is patchy conservation throughout the region; again, this is typical of many comparisons. This view also shows that all of the elements are collinear on scales of ~10 kb.
In both the dot-plot and pairwise mapping view, multiple comparisons done with different parameters can be displayed in different colors. The threshold for the matches shown can be adjusted until the desired view is obtained, and sequence can be exported from any of the views via a pop-up menu.
Once a threshold is chosen, the user can expand the view of a particular region. Figure 3 shows a closeup view of the region outlined in blue in Figure 2. The sequence shown in Figure 3 is a small patch of conservation upstream of the first exon, displayed at a 19/20 threshold. Here the user scans along the sequence and visually compares both the boundaries of the matches and the complexity of the sequence. Sequences are directly exported to other applications via the "paste" buffer.
FRII also performs searches for motifs using the IUPAC notation in which e.g. W represents A or T. This feature allows users to search for matches to known "consensus" binding sites for transcription factors. Searches are either stored on the Cartwheel server and displayed as individual features on FRII views, or executed directly in FRII. One particularly convenient feature is the ability to ask for motifs that have mismatches in up to 5 positions; this lets users search for weaker matches to known consensi.
Other analyses
FRII displays a variety of analyses. In addition to paircomp windowed comparisons, FRII displays and manipulates Vista-style comparisons, BLAST and blastz comparisons, BLAST database searches, cDNA and protein comparisons, and the results of several different gene finders (genscan, geneid, and hmmgene [20-22]). All of these analyses may be executed directly on the Cartwheel server, excepting only Vista comparisons using the (default) AVID alignment program. The data for Vista comparisons must be uploaded from the results returned by the Vista Web site; however, Vista-style comparisons with the LAGAN global alignment tool are executed directly on Cartwheel.
Discovering and analyzing regulatory regions
We and others have successfully used paircomp, FRII, and Cartwheel to discover a number of regulatory regions (see Introduction). Once we have a pair of genomic regions to compare, the steps we follow are essentially invariant from region to region:
1. We set up two to three paircomp analyses at the following windowsizes and thresholds: 10 bp/90%; 20 bp/80%; 50 bp/60%.
2. We match the cDNA or protein of interest against both regions, to determine where the coding regions lie.
3. We also compare the RefSeq database from NCBI against both regions, to find other genes in the region.
4. We load these analyses into FRII and zoom in to a view that includes as much intergenic sequence around the gene as is possible without also including other genes. We then adjust the thresholds on the 20 bp and 50 bp analyses until we obtain a roughly collinear pattern of conserved blocks. Typical values for these thresholds are 80–100% for a 20 bp windowed comparison, and 60–80% for a 50 bp windowed comparison.
5. We use the closeup view to extract the conserved blocks, and design PCR primers to isolate all of the contiguous blocks of conserved sequence. We then individually subclone or fuse them into a GFP reporter construct together with a basal promoter. These constructs are then introduced into the sea urchin by microinjection and analyzed for appropriate spatiotemporal expression.
In our experience, we have always been able to identify the relevant enhancer elements using this procedure. A similar procedure in which putatively negative elements are fused with a ubiquitous driver of expression often identifies necessary repressive elements. Also note that one caveat of these procedures is that for some genes, e.g. transcription factors, there are often many regions that appear to do nothing. These may be regulatory regions that affect expression at times or in places that are not under consideration, or could be other genomic features not relevant to gene regulation.
Conclusion
Paircomp, FamilyRelationsII, and Cartwheel are an effective, easy-to-use set of tools for analyzing conservation in BAC-sized genomic regions. Over 100 people are currently using them, and they have been effective in finding regulatory regions in a variety of organisms. In this paper we have described the tools and provided an introduction for biologists who wish to use them.
Availability and requirements
See Implementation, above, for information on server-side software.
Project name: FamilyRelationsII
Project home page:
Operating systems: Mac OS X, Windows NT/XP, UNIX/Linux (X Windows)
Programming language: C++
License: GPL/LGPL
No restrictions placed on use.
Authors' contributions
CTB designed and implemented the majority of the functionality described. YX implemented a significant portion of the XML-RPC functionality used for client-server interaction. EHD laid out the design requirements, aided in writing the paper, and supervised the development of FRII. RAC is responsible for running the servers and did the majority of bug testing, and also contributed to the paper.
Acknowledgements
Tristan De Buysscher and Madeleine Price, under the supervision of Dr. Barbara Wold, developed the original seqcomp and contributed to FamilyRelations. Ramon Cendejas and Kevin Berney aided in the development of features and helped exercise the Cartwheel server; a complete list of contributors to FamilyRelationsII and Cartwheel can be found on the Cartwheel Web site, under Developers. We especially thank Carolina Livi, Pei-Yun Lee, Dr. Ellen Rothenberg and Dr. Erich Schwarz for extensive user-interface testing over the years. Dr. Ellen Rothenberg and Dr. Erich Schwarz both contributed significantly to discussions of new features; in addition, Sagar Damle, Tracy Teal and Dr. Erich Schwarz gave many helpful comments on this paper. We also thank two anonymous reviewers for their comments. CTB is supported by National Institutes of Health Grant GM61005, and the Beckman Institute Center for Computational Regulatory Genomics is supported by National Institutes of Health Grant RR15044.
Figures and Tables
Figure 1 A paircomp comparison of the otx gene locus from S. purpuratus (top) with L. variegatus (bottom). We used paircomp to compare all 20 bp subsequences from a 160 kb S. purpuratus BAC with a 62 kb L. variegatus BAC; those 20 bp subsequences with a exact match of 19/20 or 20/20 bases are connected with a red line. Only the 80 kb surrounding the otx gene is shown on the top. Matches to the known S. purpuratus cDNA sequence are shown in red on the top sequence, and TBLASTX matches in L. variegatus to the same cDNA sequence are shown in blue on the bottom sequence. The L. variegatus genomic sequence does not extend to cover the 3' region of the coding sequence. On the top of the view are tabs to switch between the "pair view" (shown) and the "dot plot" view (see Figure 2). On the right side of the view are control buttons that allow the user to change both the color and the threshold at which matches are displayed. The user can also view a closeup of a region by selecting the region on the sequence (e.g. as on the bottom sequence, where a region from 40 kb to 61.6 kb is selected) and then pressing the "View closeup..." button. An example closeup view is shown in Figure 3.
Figure 2 A "dot-plot" style view of a subregion of the otx comparison (see Figure 1). The top sequence is a zoomed-in view of the otx genomic region from S. purpuratus, as in Figure 1; the region runs from 119.6 kb to 133.0 kb. The side sequence is a zoomed-in view of the orthologous region from L. variegatus, running from 38.5 kb to 51.5 kb. The region surrounding the first exon (in red) of the sp α-otx transcript is selected on the top (S. purpuratus) sequence, and the corresponding TBLASTX matches are highlighted on the left (L. variegatus) sequence in blue. The selection box in the center of the view contains the paircomp matches in this region, showing only 20 bp matches that match at 19/20 or 20/20 (corresponding to a 95% threshold). A closeup view of this region, showing the DNA sequence of the two regions with the corresponding matches, is shown in Figure 3.
Figure 3 A closeup view of the paircomp comparison of the genomic sequence surrounding the first exon of otx in S. purpuratus (top sequence) and L. variegatus (bottom sequence). The top half of the closeup view shows orthologous 2 kb genomic regions (126.2 kb – 128.3 kb in the S. purpuratus BAC, 44.4 kb – 46.5 kb in the L. variegatus BAC). Matches of 19/20 or 20/20 bases are drawn in red between the sequences, and the exon matches from Figure 2 are shown in black on the sequence lines. The bottom half of the closeup view shows the part of the sequence selected in blue on the top half of the view. Lines are drawn in black between individual matching bases, and the matching bases are colored in red. Note that both blocks shown match at 19/20 because of the single mismatch in the middle of the blocks.
==== Refs
Cooper GM Sidow A Genomic regulatory regions: insights from comparative sequence analysis Curr Opin Genet Dev 2003 13 604 610 14638322 10.1016/j.gde.2003.10.001
Brown CT Rust AG Clarke PJ Pan Z Schilstra MJ De Buysscher T Griffin G Wold BJ Cameron RA Davidson EH Bolouri H New computational approaches for analysis of cis-regulatory networks Dev Biol 2002 246 86 102 12027436 10.1006/dbio.2002.0619
Yuh CH Brown CT Livi CB Rowen L Clarke PJ Davidson EH Patchy interspecific sequence similarities efficiently identify positive cis-regulatory elements in the sea urchin Dev Biol 2002 246 148 161 12027440 10.1006/dbio.2002.0618
Kirouac M Sternberg PW cis-Regulatory control of three cell fate-specific genes in vulval organogenesis of Caenorhabditis elegans and C. briggsae Dev Biol 2003 257 85 103 12710959 10.1016/S0012-1606(03)00032-0
Romano LA Wray GA Conservation of Endo16 expression in sea urchins despite evolutionary divergence in both cis and trans-acting components of transcriptional regulation Development 2003 130 4187 4199 12874137 10.1242/dev.00611
Leung TH Hoffmann A Baltimore D One nucleotide in a kappaB site can determine cofactor specificity for NF-kappaB dimers Cell 2004 118 453 464 15315758 10.1016/j.cell.2004.08.007
Revilla-i-Domingo R Minokawa T Davidson EH R11: a cis-regulatory node of the sea urchin embryo gene network that controls early expression of SpDelta in micromeres Dev Biol 2004 274 438 451 15385170 10.1016/j.ydbio.2004.07.008
PyZine online magazine
FamilyRelations Web site
Caltech Cartwheel server, "Woodward"
FamilyRelations tutorial
Sonnhammer EL Durbin R A dot-matrix program with dynamic threshold control suited for genomic DNA and protein sequence analysis Gene 1995 167 GC1 10 8566757 10.1016/0378-1119(95)00714-8
Altschul SF Gish W Miller W Myers EW Lipman DJ Basic local alignment search tool J Mol Biol 1990 215 403 410 2231712 10.1006/jmbi.1990.9999
Bray N Dubchak I Pachter L AVID: A global alignment program Genome Res 2003 13 97 102 12529311 10.1101/gr.789803
Brudno M Do CB Cooper GM Kim MF Davydov E Green ED Sidow A Batzoglou S LAGAN and Multi-LAGAN: efficient tools for large-scale multiple alignment of genomic DNA Genome Res 2003 13 721 731 12654723 10.1101/gr.926603
Elnitski L Riemer C Petrykowska H Florea L Schwartz S Miller W Hardison R PipTools: a computational toolkit to annotate and analyze pairwise comparisons of genomic sequences Genomics 2002 80 681 690 12504859 10.1006/geno.2002.7018
Schwartz S Zhang Z Frazer KA Smit A Riemer C Bouck J Gibbs R Hardison R Miller W PipMaker – a web server for aligning two genomic DNA sequences Genome Res 2000 10 577 586 10779500 10.1101/gr.10.4.577
Frazer KA Pachter L Poliakov A Rubin EM Dubchak I VISTA: computational tools for comparative genomics Nucleic Acids Res 2004 32 W273 279 15215394
Yi TM Walsh K Schimmel P Rabbit muscle creatine kinase: genomic cloning, sequencing, and analysis of upstream sequences important for expression in myocytes Nucleic Acids Res 1991 19 3027 3033 2057360
Burge C Karlin S Prediction of complete gene structures in human genomic DNA J Mol Biol 1997 268 78 94 9149143 10.1006/jmbi.1997.0951
Parra G Blanco E Guigo R GeneID in Drosophila Genome Res 2000 10 511 515 10779490 10.1101/gr.10.4.511
Krogh A Using database matches with for HMMGene for automated gene detection in Drosophila Genome Res 2000 10 523 528 10779492 10.1101/gr.10.4.523
| 15790396 | PMC1087472 | CC BY | 2021-01-04 16:02:50 | no | BMC Bioinformatics. 2005 Mar 24; 6:70 | utf-8 | BMC Bioinformatics | 2,005 | 10.1186/1471-2105-6-70 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-741579039010.1186/1471-2105-6-74SoftwareProviding visualisation support for the analysis of anatomy ontology data Dadzie Aba-Sah [email protected] Albert [email protected] School of Mathematical and Computer Sciences, Heriot-Watt University, Edinburgh EH14 4AS, Scotland2 Medical Research Council, Human Genetics Unit, Western General Hospital, Edinburgh EH4 2XU, Scotland2005 24 3 2005 6 74 74 18 8 2004 24 3 2005 Copyright © 2005 Dadzie and Burger; 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
Improvements in technology have been accompanied by the generation of large amounts of complex data. This same technology must be harnessed effectively if the knowledge stored within the data is to be retrieved. Storing data in ontologies aids its management; ontologies serve as controlled vocabularies that promote data exchange and re-use, improving analysis.
The Edinburgh Mouse Atlas Project stores the developmental stages of the mouse embryo in anatomy ontologies. This project is looking at the use of visual data overviews for intuitive analysis of the ontology data.
Results
A prototype has been developed that visualises the ontologies using directed acyclic graphs in two dimensions, with the ability to study detail in regions of interest in isolation or within the context of the overview. This is followed by the development of a technique that layers individual anatomy ontologies in three-dimensional space, so that relationships across multiple data sets may be mapped using physical links drawn along the third axis.
Conclusion
Usability evaluations of the applications confirmed advantages in visual analysis of complex data. This project will look next at data input from multiple sources, and continue to develop the techniques presented to provide intuitive identification of relationships that span multiple ontologies.
==== Body
Background
Improvements in technology have resulted in the ability to perform increasingly complex experimental research, generating large amounts of multi-dimensional data. Research in biology today involves the exploration and analysis of the large, shared databases available to the wider scientific community. This reduces duplication of effort and allows more extensive research to be done, by providing independent data sources for verifying hypotheses formulated.
Conversely, technology has been unable to manage the data it has spawned; data management, analysis and visualisation tools struggle to meet requirements [1]. This is due largely to the constantly evolving information needs of the biologists working with the increasingly large amounts of heterogeneous data generated [2]. Bioinformatics developed to exploit information technology in biological data analysis [3,4].
This paper looks at the development of visual solutions for intuitive analysis of anatomy ontologies. The primary data source is the Edinburgh Mouse Atlas Project (EMAP), developed at Edinburgh's Medical Research Council's (MRC) Human Genetics Unit (HGU), documenting the developmental stages of the mouse embryo.
Prototypes are being developed for visualisation in 2D (two-dimensions) and 3D, to highlight relationships between and within different anatomy components. This will help uncover knowledge about gene expression, structure and function as cells and tissues evolve into defined organs, to track normal development and evolution.
Visual solutions for effective analysis of biological data
At least a basic understanding of the structure of the anatomy ontology data is required to provide effective design and development with mappings to spatial representations that capture users' mental models of the data. Visualisations generated should provide biologists with data overviews, followed by the ability to study regions of interest (ROIs) in detail, within the context of the overall data set, to highlight patterns within the data [5,6]. Support for interactive data exploration should be provided; functionality for browsing and searching and for manipulation of data structures allows analysis from multiple perspectives.
Visual encoding of textual data exploits humans' highly developed perceptual abilities to decrease the cognitive load associated with complex data analysis [6,7]. A number of visualisation methods and techniques already exist for complex data analysis, both within and outwith the field of bioinformatics, including 2D and 3D scatter plots, self-organising maps (SOMs), parallel coordinates, 2D and 3D hierarchical graphs, information maps, murals and cubes, perspective walls, virtual landscapes, cityscapes, and physical space metaphors such as rooms, windows and desktops. Hyperbolic or fish-eye views and lenses, magic and semantic lenses, and dynamic query systems aid detailed study of regions of interest (ROIs) especially in large data sets. [3,6,8-10]
In order to ascertain what would provide, individually or in concert, optimal visual data analysis solutions for the study of anatomy ontologies, it is necessary to assess existing tools and techniques to determine their applicability to the data sets of interest and the tasks biologists perform. It is important to provide analysis solutions that the different target users with varying research backgrounds are able to use to interpret data required for their work effectively. [11]
The Edinburgh Mouse Atlas Project
The working EMAP browser (see Figure 1) employs a collapsible, indented text index with mappings to corresponding anatomical components in 2D and 3D digital models for the developmental stages of the mouse embryo.
Extensions required to EMAP browsers
More intuitive methods are required for data analysis, especially where comparisons between multiple data sets, such as lineage across stages, are to be made. A major advantage in visual representation of the data over EMAP's text indices is the availability of an overview in addition to the ability to study ROIs in detail.
The ontology data is structured hierarchically from a root, with unique sub-components together forming complete (super) components [12], using part-of relationships. A graphical layout that exploits this structure should aid discovery of relationships within the data [6,8]. Visualisations can make use of physical attributes such as shape, colour and size to encode properties of and relationships between data elements. The following sections discuss advantages in visual data analysis:
1. Representation of complex relationships
As all the developmental stages for a specified anatomy are combined into the complete, abstract organism persistence of components across stages may result in non-unique entries, with different paths to the root.
An illustration from the mouse anatomy ontology would be the component second polar body, which exists on the first level below the root for the first three stages of development. However in the 4th stage the second polar body forms a part of the extraembryonic component. Figure 2 shows the two occurrences of the second polar body in the ontology for the abstract organism, on the first and the second levels below the root.
The second polar body could be regarded as having multiple parentage in the abstract organism, as demonstrated for the component G in the index in Figure 3.
2. Grouping
Grouping of anatomical components based on user-specified criteria, to provide different perspectives of the data structure, is a function required for which no graphical support exists in the EMAP browsers. Grouping enables users to focus on relationships within data based on different classification methods for structuring data than those used for constructing the ontologies. A typical example would be the skin that covers the entire organism. Skin is not represented as a super-component in any one stage; however there may be components in each stage that together make up the skin of the organism at that point in development. A group component skin could be created that links to all components that comprise the skin in a stage, using the default part-of relationship again to form a complete whole, while preserving the uniqueness of data components [12] (see Figure 15). Note that this will result in multiple parentage for unique nodes, similar to that occurring in Figures 2 and 3.
3. Lineage
Lineage across stages is currently displayed using plain text descriptions of a component's ancestors and descendants in consecutive text boxes arranged along a horizontal plane. The main disadvantage associated with this is the need to scroll through up to 28 text boxes to identify all the Theiler stages (TS) of the mouse embryo, for example, through which a specified component persists (see Figure 4).
4. Visual querying
It is necessary to extend the sub-string searching provided in the EMAP browsers to retrieve related information from external data sources. Lack of integration between databases and the different search and query tools provided for these data sources [1,4] however present problems for information retrieval (IR). Differences in data quality resulting from varied methods for collection and storage, and different data annotation techniques may result in large semantic and syntactic differences in information stored, increasing difficulty in the retrieval and use of data from different sources [13].
Transparent mappings to external data sources that hide their underlying complexity should aid searching and IR. Visual, dynamic querying, illustrated in Figure 5, employing semantic lenses and dynamic query sliders, encourages data exploration by removing the need for users to learn high-level query languages and syntax. Complex queries can be formed intuitively using an interface that provides immediate, visual feedback and allows simplified modification and/or reversal of actions [14], without increasing cognitive load.
Searching currently makes use of free text fields for input, highlighting search hits within the overview. Modification to use a semantic lens could extract information of interest and use this as input for a refined search. The sub-tree containing the component parts of the heart, say, could be extracted using a filter that takes heart as input and node property print name to filter out non-relevant data. Related information from other data sources could then take the results of this search as input to extract gene expression, say, on the components of interest.
5. Simultaneous analysis of multiple anatomy ontologies
Determination of lineage and the comparison of ontologies for different organisms require simultaneous visualisation of multiple data sets. Intuitive comparison between different data sets requires visualisations that highlight mappings between related data elements (see also Figure 14).
Related work in hierarchical data visualisation
Several applications already exist for the visualisation of large, hierarchically structured data sets. It is important to examine specific applications to determine if users' requirements cannot be satisfied with existing visualisation solutions. Those applications found to most closely approach the requirements of this project are summarised below, detailing features they provide and their limitations for the data analysis required.
1. Protégé
A Java-based knowledge modelling tool, Protégé incorporates multiple hierarchical visualisation applications to aid the construction, editing and visualisation of ontologies. These include OntoViz, which makes of the GraphViz visualisation libraries for graphical representations of hierarchical data. OWLViz, specifically developed to visualise OWL ontologies, also makes use of GraphViz. A major problem with extending GraphViz is that it requires use of the non-standard languages DOT and lefty.
TGViz uses Touchgraph, developed in Java, for dynamic layout of nodes in a connected network graph. TouchGraph encodes node properties using colour, and provides clustering of like data, as well as geometric and hyperbolic zoom. Functionality for searching and for saving graphs as image files is also provided.
SHriMP, providing modular components built using Java Beans, is combined with Protégé to form Jambalaya, a tool that provides fish-eye views that make use of a continuous zoom for overviews of large data sets. Data abstraction, employing nesting and hiding of data, is followed by extraction of sub-sets to separate windows to allow focus on detail in ROIs. Encoding of data nodes using colour and depth cueing in 3D helps to distinguish more important data. [15]
2. Piccolo
Developed using Java2D and Swing, Piccolo provides support for zoom and the use of multiple cameras or viewpoints. Piccolo has been customised for visualisation of network and hierarchical data. GINY, the Graph Interface Library, extends Piccolo to visualise protein and gene interaction and expression in Cytoscape. GINY uses colour coding of gene expression to aid comparison of data sub-sets. Data reduction is achieved by clustering related data into encapsulating, composite nodes.
SpaceTree extends Piccolo to produce rooted, node-link hierarchical graphs that combine (physical and semantic) zoom, panning and folding of sub-graphs to provide maximum screen space to ROIs. Main disadvantages of SpaceTree include the loss of the overview when analysing ROIs. Also, though the provision of (textual) detail in nodes is useful this reduces screen real estate available for the overview.
Zoomgraph extends Piccolo's zoomable interface to provide semantic zoom into ROIs. An advantage in Zoomgraph is the ability to define the properties of data nodes and links. [16]
3. Walrus
Using Java3D, Walrus employs a hyperbolic projection for the visualisation of directed graphs in 3D space. Although its hyperbolic layout provides an ideal method for displaying the kind of hierarchical data under study Walrus is fairly specialised; it uses its own non-standard file format. Further, the structure of graphs once loaded cannot be altered, and only one graph can be loaded at a time. [17]
4. Hypergraph
This Java application provides a hyperbolic layout in 2D that allows interactive repositioning of nodes to provide more magnification to ROIs, with hyperlinks to further detail in external files. [18]
5. VRMLgraph
Developed using Java, VRMLgraph draws arbitrary node-edge graphs in 3D. Very little functionality is implemented beyond the drawing of nodes and links connecting them; the main benefit of this application is that it can take advantage of built-in navigation cues and capabilities in VRML for 3D perspective and cameras/viewpoints. [19]
Motivation for project
The applications and toolkits described above incorporate visualisation techniques for exploration of overviews, and with the exception of Walrus, detailed analysis in ROIs. Given the proven capability of hyperbolic layouts for navigation and exploration of large data sets it would be useful to harness the simple but effective implementation of the hyperbolic layout in Hypergraph. Protégé as it stands provides functionality that would still need adaptation to allow creation of groups as required for the anatomy ontology data. Functionality for folding trees, searching, and highlighting of user paths, and encoding of data properties require further development to satisfy user requirements.
Harnessing existing technology that performs effective analysis of complex data, applied in the tools studied above, will provide some of the functionality required to aid visual data analysis. However it is still necessary to develop novel techniques for analysis of the anatomy ontology data, building on existing methods that have proven useful for visualisation of complex data. Intuitive comparison of multiple data sets and the tracing of lineage through the anatomy ontologies cannot be obtained using the functionality available in 2D tools; occlusion would be too high to allow useful analysis at any level of detail. 3D tools would reduce the problem of occlusion significantly. However 3D visualisations typically distribute data throughout the space available, clustering related data nodes around focal points. Distinguishing individual nodes and the data sets to which they belong is difficult. Drawing physical links between related nodes across data sets may result in crossing of links, reducing the ability to recognise these relationships.
The next section describes the 2D visualisation browser developed that builds on existing techniques to provide intuitive visual data analysis. This is followed by an evaluation carried out to determine if the visualisations provided improve the textual analysis currently performed. Finally a 3D browser develops novel techniques that provide intuitive recognition of relationships across data sets, described in the sub-section Novel techniques for visual analysis of the ontology data under Implementation.
Implementation
A prototype has been developed in Java to provide visualisations of the ontology data. A rooted directed acyclic graph (DAG) in 2D provides an overview of each data set, with the ability to study ROIs in detail. Anatomy components are represented using a 2D circle (node), while relationships between nodes are represented by (1D) lines linking nodes, encoding relationships using colour. Functionality implemented is described below, highlighting changes and additions made after the performance of the heuristic evaluation summarised in the next section.
Description of visualisation prototype developed
1. Layout
Only the top three layers of the DAG are drawn when the visualisation is first generated, in order to provide more screen space for data nodes and minimise occlusion; typical of data with a hierarchical structure, the number of nodes in a level increases as one descends the tree. The top layers provide an abstraction of the data set from which users may continue to extract more detail.
The default layout of the DAG, with the lowest level of occlusion due to node labels, is left-right (L-R), shown in Figure 6. A top-down (T-D) layout is also available. A radial layout, illustrated in Figure 7, was suggested during the heuristic evaluation as an option for more optimal use of screen space. Layout initially provided equal space to all nodes in each level. However as can be seen in Figure 8 this makes poor use of screen space; some areas have high occlusion while others are sparsely populated. The improved layout in Figure 9 weights the layout for the first layer below the root, where the largest bias occurs in node distribution, with weight dependent on number of (immediate) sub-nodes.
2. Textual detail
Labels have a default setting of component name for nodes, and (primary) relationship between nodes for links. Labels may be set to any of a node's properties, and all node and link properties may be displayed on request. The option to hide labels may be used to reduce occlusion, in which case holding the mouse over a node or link of interest brings the focus to it and pops up its label.
3. Highlighting & ghosting
Nodes and links of interest may be highlighted for emphasis, while less important data can be suppressed by ghosting it out.
4. Selection of ROIs
A rectangle can be drawn to select nodes in close proximity to each other, making it possible to perform actions on the selection simultaneously.
5. Expansion & collapsing of sub-trees
One solution to occlusion is to collapse sub-trees to hide less relevant data, providing more screen space to visible nodes.
6. Zoom
The prototype has four implementations of zoom: ROIs can be (re)drawn in a separate window, providing magnification by drawing the same number of nodes in a larger area, shown in Figure 10. A sub-tree may also be drawn in a separate window, providing both a semantic and a physical zoom.
The ability to zoom into an ROI within the context of surrounding information was suggested by users as a useful option. Employing an implementation that makes use of a hybrid between a semantic lens and a continuous hyperbolic zoom, this expands any one sub-tree while collapsing all others beyond that level. The ROI is drawn using maximum screen space, with minimal rearrangement of the graph, as Figure 11 shows. This also removes the extra cognitive load required to map between separate but related visualisations held in multiple windows.
Physical zoom of the entire tree was not implemented initially because this requires scrolling through the DAG to move to ROIs, with an associated loss of context. Users however suggested the reduction in occlusion due to nodes being pushed further apart would make it easier to explore detail in ROIs, compensating for the loss of the overview. This option has now been implemented.
7. Searching/querying
A custom dialog allows users to perform sub-string searches on any of a node's properties. Textual results showing component ID and print name for nodes that satisfy a query are supplemented with highlighting of hits in the graph.
8. Grouping of nodes
Functionality is provided for the creation of group nodes, linked to nodes already existing in the DAG, to group related nodes (see Grouping under the section Extensions required to EMAP browsers and Figure 15).
9. Tracing lineage
Graphical layout of the data aids determination of lineage by allowing users to trace the ancestors or descendants of a node using a data overview.
10. Resetting to default state
The option to remove all formatting applied to nodes returns to the default rendering of data elements.
Heuristic evaluation of the 2D anatomy browser
A heuristic evaluation was carried out with a small group of target users comprising biologists and computer scientists working at the MRC on EMAP. This involved a demonstration of the functionality available for interaction with and manipulation of the visualisations generated. Users commented on usefulness (or not) of aspects of the system, and made suggestions for improvements, additions and changes to the system. This led to the redesign and re-implementation of the prototype, in preparation for a structured user evaluation.
A major issue encountered in the visualisation of complex data is occlusion (see Figure 8), an acute problem in the visualisation solution developed, especially due to data labels. The evaluation highlighted further issues with occlusion in the visualisations, and provided suggestions for reduction of this problem, illustrated in Figures 6, 7, 9, 10 and 11.
A second major problem identified is the exponential increase in system response time with data load, with a significant negative impact on interaction, illustrated in the graphs in Figure 12. This problem is even more acute for remote execution in X-windows, probably due to enhancements for Swing in Windows which have the reverse effect in X-windows [20].
Novel techniques for visual analysis of the ontology data
The main advantages in 3D visualisation of large data sets are the larger amount of space available for data display, due to the added dimension of depth, and the higher degree of freedom for exploration and navigation. Natural perspective in 3D also provides increased magnification as one approaches the user's viewpoint. Using 3D makes possible the simultaneous display of multiple data sets (see Figure 13), with far less occlusion than would occur in 2D.
This project requires the analysis of multiple data sets, to identify relationships that span multiple stages in one organism or map relationships between different organisms. Beyond a threshold of about 200 nodes, however, significant occlusion occurs in 2D.
The 2D browser makes use of existing techniques for visual analysis of individual data sets, with modifications and extensions as required to provide optimal analysis. To improve data analysis in areas of high occlusion different kinds of zoom have been implemented, including the development of a hybrid lens (see Zoom under the section Description of visualisation prototype developed). A uniform zoom expands a sub-tree of interest to make maximum use of space in the main window; without relocation surrounding nodes would be obscured, as occurs for magic lenses. A hyperbolic lens could be used to move surrounding data to the periphery of the window; however constantly changing hyperbolic layouts have the disadvantage of destroying the mental models users form of data structure. The hybrid lens developed folds nodes on the same level as the node whose sub-tree gains the focus. Surrounding context is still maintained to a large degree, while users concentrate on analysis of the ROI. However this and other methods for reducing occlusion in the 2D browser provide only a partial solution, and only for individual data sets.
Increasing to three the number of dimensions used for visualisation provides more space in which to hold data notes, further reducing occlusion of data. 3D however comes with its disadvantages, detailed in the following section, with the most significant being disorientation when navigating through 3D space. Evidence however exists for the increased usability of between 2.n (n > 0) and 3 dimensional [8] visualisation of data; users are able to move out of the area holding the data and fly over or below ROIs. This reduces the feeling of immersion into the data and the disorientation this causes. Overviews of data sets are obtained that improve users' mental models of data structure, and users are able to move back into the data as required to analyse ROIs in detail.
In order to take advantage of the space provided in the extra dimension without losing the benefit of simpler analysis in 2D a novel system has been developed that continues to draw individual data sets in 2D, layered uniformly in parallel along the horizontal axis (see Figure 13). The physical space between the layered DAGs serves as the information space used to hold relationships between node pairs belonging to different ontologies. Links are drawn across the space between related data sets, using colour codes to represent the different relationships that exist. Users are able to transfer learning in the use of the 2D browser to visualisation in 3D, continuing to analyse individual data sets in relative isolation, each lying in its own plane. Because relationships spanning data sets lie in separate planes and are drawn parallel to those holding the DAGs they stand out and are easily recognised, especially when viewed from a point above or below the data area, shown in Figure 14.
Figure 15 illustrates the solution implemented for grouping of nodes in the 3D anatomy browser. This shows the benefits gained by relocating the group created to a plane parallel to that in which the DAG lies, removing the crossing of links that occurs in 2D. Placing group nodes in a separate plane has the added advantage of removing the increase in occlusion that occurs in 2D due to the addition of further nodes and links in the plane holding the DAG. The Structured evaluation of the visualisation prototypes detailed further in this paper confirmed that grouping of nodes in 3D was found to be more intuitive than for the 2D browser.
Issues encountered in the 3D visualisations
The main difficulty encountered interacting with the 3D browser is maintaining control over the data structure during navigation. Built-in functionality for navigation in Java3D employs the keyboard and/or the mouse, allowing translation, rotation and zoom. The lack of a history function means that it is difficult to return to specific points in the data structure; the only option available to users for recovery when the data structure is moved out of the viewing area and the bounding sphere within which behaviours are active, or when users become lost in the data, is to return to the centre of the universe (and default camera viewpoint).
Occlusion of more distant data in 3D, disorientation of users during navigation, and the complexity associated with the creation and support of 3D visualisations in terms of software and hardware required mean that 3D may not necessarily provide a better option than 2D. Employing 2D for the visualisation of individual stages or for abstracted views of individual, complete anatomy ontologies, with an extension to 3D as multiple stages and/or ontologies are compared may provide an optimal solution. Functionality for switching between the 2D and 3D views should help to resolve the disorientation that occurs in 3D by allowing users to focus on only data of interest in 2D without the distraction of extraneous information. Continuing to place individual data sets in 2D planes reduces disorientation by giving users more control over the level to which they become immersed in the data; users are able to move above the area holding the data and observe its structure from above or below. This provides an overview that highlights all relationships occurring across data sets in addition to displaying each data set lying in its own 2D plane.
To ease integration of the prototypes with the EMAP Anatomy Browser it is important that their graphical user interfaces (GUIs) do not deviate significantly from that of the EMAP browser. This should reduce users' learning curves, increasing willingness to use the prototypes. For this reason it may be useful to continue especially GUI development in Java. This also eliminates the problem of cross-platform compatibility and enables development for use on the Internet. Java3D, using OpenGL for low level rendering acceleration, may seem the obvious choice for the extension of the 2D browser. However limited support for development using Java3D may result in restrictions in functionality and that may have little future support. Further, Java's cross-platform compatibility comes with a cost in performance due to the overhead incurred in the interpretation of Java byte code into native machine code, manifested in the large increase in system response time observed as data load increases.
Another option would be to develop the 3D browser using OpenGL (with C/C++), to take advantage of OpenGL's advanced 3D modelling capability, hardware acceleration and larger user base and support. GL4Java bindings from OpenGL to Java could then be used to provide an interface with the benefits of Java.
Structured evaluation of the visualisation prototypes
In order to establish if user requirements have been fully and correctly captured and implemented it is necessary to evaluate the visualisation prototypes with target users. This provides measures of user satisfaction and determines the effectiveness and efficiency of the systems developed. If the new system developed is to be used it will have to provide advantages over tools in current use, improving productivity without increasing users' work load [7,20,21]. A structured evaluation should identify usability issues and highlight features that improve data analysis. Requirements for changes to, additions and extensions to functionality also need to be identified. Sources of error and poor system response should be minimised; effective system feedback and good error management contribute to user satisfaction [5].
To guide the evaluation two main hypotheses were tested:
H0: Visual analysis of especially large, complex data sets provides advantages over textual analysis.
H1: Visualisation in 3D provides advantages for analysis over 2D that justify the larger amount of support required.
Preparation for structured evaluation
1 Task scenarios
To perform optimal data analysis users must possess the information required to achieve their goals: domain knowledge and how functionality in systems maps to this. To be effective a user evaluation should simulate the working processes of typical users in their normal working environments. A set of task scenarios were developed, detailing successful completion criteria and maximum goal completion times. These benchmarks could then be used to compare expected/ideal user behaviour to actual.
A heuristic evaluation, involving a walk through the scenarios developed, was carried out to ascertain that the scenarios capture typical user tasks in a normal work environment and allow users to explore functionality offered. The walk-through also highlighted further development required to increase intuitiveness of the prototypes, summarised in Table 1 below, and carried out prior to the structured evaluation.
2 Custom questionnaires
Custom questionnaires based on the QUIS, the Questionnaire for User Interface Satisfaction [5], were prepared to elicit subjective measures of user satisfaction and usability of the prototypes: a pre-evaluation questionnaire to collect demographic information and a post-evaluation questionnaire to collect information on usability of the system. The (custom) questionnaires were reviewed by an HCI (Human Computer Interaction) expert and revised where necessary.
3 Pilot test & review of evaluation procedure
A pilot test was then carried out with an (independent) HCI expert with a specialisation in visualisation. This examined the design of the evaluation procedure and checked that usability requirements had been integrated into it.
Results and discussion
Implementation of structured evaluation procedure
1 User backgrounds
Two main target user groups of the prototypes have been identified: the primary target – biologists, and the secondary target – computer scientists working in bioinformatics.
Ten users in total performed the evaluation, six from the MRC and the remaining four, research students at the School of Maths and Computer Sciences (MACS) at Heriot-Watt University. With the exception of one user, an HCI expert from MACS, all users had some experience in bioinformatics, and most had at least tried out the EMAP browsers and/or had some association with XSPAN (The Cross Species Anatomy Network).
2 Methodology
A flow diagram was used to record users' paths to complete each task, noting additionally, users' reactions and comments, errors made, requests for help and responses given. Software logging captured interaction with the menus and toolbar (transparently), and a built-in timer was used to record task completion times.
Verbal help was provided by the developer to supplement the help files which had not been completed prior to the structured evaluation.
Analysis of evaluation data
Qualitative feedback was obtained from the custom and the SUS (System Usability Scale) questionnaires administered, and from observation of users recorded during each evaluation. Because the test user population is only ten statistical analyses may not provide very reliable indications of satisfaction or otherwise with the prototypes. Therefore the only statistical tests performed are those looking at means (within a 95% confidence interval (CI)) and extreme values of rankings. Mean user satisfaction rankings for the post evaluation questionnaire, SUS scores, and task completion times are shown below:
1 SUS scores
The SUS scores obtained, as shown in Figure 16, indicate eight out of nine users with a score above the mid mark (50). The highest score obtained was 90 (out of 100) and the lowest was 37.5. The mean SUS score is 62.5.
2 Satisfaction ratings
Overall means for satisfaction ratings for individual users, shown in Figure 16, have eight out of ten users with rankings above the central mark 5, with five of those values lying above the mean for the entire sample. One user ranked the 3D browser above the central mark and the 2D below, and the last ranked both below the central mark. Mean rankings for the 3D browser were higher than for the 2D, both over the whole user population and by user group (see Figure 16). Rankings by target user group saw biologists, the main target, rating usability higher than computer scientists.
Further analysis of (mean) rankings for each item show quietness of system, consistency of terms used and good relation to users' normal work, eased ability to determine lineage, low time to achieve proficiency using the systems, eased data analysis, and the visualisations providing advantages over the text indices fell in the top ten for both the 2D and 3D browsers. Hiding of sub-trees to reduce occlusion and reliability of the system were other items ranked in the top ten for the 2D browser. Consistency of messages on screen, and the options for zoom also fell in the top ten rankings for the 3D browser.
Large variations in system speed had the worst ranking for both browsers, followed by average time to perform tasks. Occlusion of data, level of support for error recovery, ease of navigation through data, and flexibility of the system all fell within the lowest ranked items for both browsers. Additionally, for the 2D browser, the system being satisfying, ease of reading text on screen, ghosting of nodes and occlusion of data specifically for the T-D layout fell in the bottom ten rankings. Ability to identify errors and their sources, difficulty of the system and the needs of experienced and inexperienced users being taken into account all fell within the bottom ten rankings for the 3D browser.
3 Task completion times
Task completion times are summarised in Figure 17.
It is observed that times for T2-2D (task 2, 2D browser), T3-2D, T7-2D, T1-3D and T2-3D are very high compared to expected maximum completion times. Analysis of the qualitative feedback from the flow diagrams for each user provides clues that explain these results, highlighting areas where users had difficulty identifying or understanding functionality required to complete tasks. Restrictions for data input, poor wording or ambiguity in function labels and poor presentation of additional textual information to users contributed to most of the problems encountered. Learning was exhibited by users performing repeated tasks in 3D in shorter times than expected. A significant example is for grouping of nodes; users also found grouping in 3D to be more intuitive than for the 2D browser (see section on Novel techniques for visual analysis of the ontology data).
Changes made to prototypes based on structured evaluation
A number of changes and additions have been made to the prototypes, based on the evaluation results. More detailed information is provided on node and link properties, and functionality is available for insertion of user comments. The DAG now automatically displays more levels if required to uncover hidden information requested by users. Further options for search include searching on hidden nodes and within a sub-tree only. Specific search hits in the DAG may also be isolated and highlighted from within the search dialog. Ghosting now fades out links to and from ghosted nodes and hides their labels.
The functions for creating groups have been moved to the Edit from the View menu, to map to users' semantic interpretation of the function. Options for selection of nodes now include retrieval from the graph by clicking, choosing from a drop-down list or by entering component IDs in a free text field. Functionality for editing and removing groups has also been implemented, and users may extract a group of interest for analysis in isolation in a 2D (sub) window.
Input dialogs when first opened now prompt users with the current value of the variable of interest. Where practicable free text fields have been replaced with sliders to prevent errors by allowing input only within legal ranges. An editable legend detailing encoding properties for nodes and links has been provided.
Full implementation of functionality in sub-windows allows tighter coupling between the main and sub-windows. Functionality for switching between the 2D and 3D browsers without exiting either application has been implemented. The ability to save user sessions (to text or image files) aids incremental analysis. Finally, the help files now also provide implementation detail and clarification of ambiguous terminology.
Conclusion
The EMAP browsers in current use employ textual indices for the analysis of anatomy ontology data for the mouse embryo, with mappings to 2D slices in 3D regions of models of the embryos at each stage of development. The lack of an overview for these large data sets increases users' cognitive load during data analysis; more intuitive methods for analysis are required to aid research.
Visualisation of large, complex data sets exploits human perceptual abilities to increase cognition and ease data analysis, by providing spatial representations that map to users' mental models of data structure. This paper explores the use of hierarchical graphs for the visualisation and analysis of the complex, hierarchically structured, multi-dimensional ontology data under study, assessing different visualisation techniques available for analysis of this data and their merits and limitations. Prototypes are being developed for 2D and 3D visual analysis of the ontologies, to provide alternative and novel solutions where methods and techniques available are unable to meet users' needs. Visualisations generated provide overviews of each data set with functionality for further interactive analysis of ROIs.
Evaluation of the prototypes being developed aimed to test two main hypotheses: whether visualisation provides advantages over textual representation for data analysis, and whether 3D visualisation provides enough advantages over 2D to justify the larger amount of development and support required for it.
Analysis of the data collected during the heuristic and structured evaluations showed users found the visual representations of the data to provide significant advantages over text for determining data structure, and more intuitive methods for search and query. However increasing levels of occlusion with data set size render visualisations of the larger data sets difficult to use. Functionality for data reduction and the ability to analyse data of interest in isolation offer improved solutions for analysis. Further novel solutions are required to resolve more fully the problem of occlusion.
Evaluation results showed the 3D browser slightly more intuitive for use than the 2D. The larger amount of space available in 3D makes it possible to display more data with a lower amount of occlusion than occurs in 2D. Functions for searching, grouping of data and locating information using the 3D browser were also found to be more intuitive. The large variations in system response time encountered in 2D, a major source of user dissatisfaction, did not pose as significant a problem in 3D.
Further work
Further functionality for the 3D browser will therefore be implemented, looking especially at providing solutions for the most significant problems encountered in 2D.
Further development on the prototypes will build visual structures that provide effective, interactive encoding and display of multiple relationships between and within different data sets.
General applicability of software developed
The visualisation applications have been developed to work with data from EMAP, stored in XML or HTML files. Storage in self-describing XML eases analysis of the data structure; the information obtained is used to generate the visual representations that provide intuitive analysis of the anatomy ontologies under study.
The next stage of the project is looking at reading data directly from the EMAGE (Edinburgh Mouse Atlas of Gene Expression) database. The browsers are built so that a layer exists between the data read in and the generation of the visualisations. Therefore reading in data from different sources only requires writing additional parsers for the new data sources, and feeding the parsed data into the visualisation system.
The visualisation browsers however expect specific data properties; the systems also require updating to read in and process additional or different data properties from those used to describe the structure of the mouse embryos, to allow customisation for differently structured ontologies. Visualisations for other data sources containing ontology data may then be generated after writing custom parsers for the data they store.
Availability and requirements
Project names: EMAP and XSPAN
Project home page:
Operating system(s): Platform independent
Programming language: Java
Other requirements: Java3D with OpenGL
License: None
Any restrictions to use by non-academics: None
Authors' contributions
ASD did the background research, built the prototypes and carried out the evaluations, and wrote the paper. AB was one of the reviewers in the heuristic evaluation, provided suggestions in building the prototypes, reviewed the draft of the article and approved submission.
Acknowledgements
Appreciation goes to the several researchers at the MRC and the research students in MACS who took part in the evaluations, and for the useful comments and suggestions made. Special thanks go to Gus Ferguson who spent a lot of time helping prepare the structured evaluation.
Figures and Tables
Figure 1 The EMAP anatomy browser. The EMAP Anatomy Browser, showing a 3D reconstruction of the mouse embryo at Theiler Stage 14. Components selected on the 2D section (in blue) are highlighted in the corresponding collapsible text index (in red) for that stage. Even with some levels collapsed the inability to obtain an overview of the entire data set can be seen.
Figure 2 Multiple parentage for the abstract mouse. Persistence of components across multiple stages may result in non-unique component names for the abstract organism, but still with unique paths to the root. The second polar body is found in different levels in multiple stages. In the abstract mouse it appears twice; the two nodes with identical names are highlighted and the path each traces to the root is drawn.
Figure 3 Employing visualisation for the representation of complex relationships within data. The collapsible text index contains a component G with multiple parentage; it is necessary to provide two separate text entries to show descent from each parent. The corresponding directed acyclic graph (DAG) could either represent the data with two separate nodes (clones) or display the node G only once, with a link to it from each parent. Advantages in the latter representation include a reduction in the amount of data displayed to users.
Figure 4 Using visualisation to trace lineage intuitively. A more intuitive visual solution to the current textual method (top) for displaying lineage across stages is shown to the right. The lineage paths trace evolution of a component from the root to another component across a number of stages.
Figure 5 Visual, dynamic querying. A semantic lens is used to provide both a physical and semantic zoom. A specified node is then as input for a search and dynamic sliders are used to set search criteria. The provision of search ranges and data sources available aids users in determining information available to perform directed searches. Finally, transparent formatting of search syntax removes the need for users to learn complex query syntax.
Figure 6 The L-R layout for the first five levels of TS11
Figure 7 The equivalent radial layout for TS11 drawn In Figure 6. There is more uniform distribution of nodes for this layout than for the L-R or T-D. This reduces the problem of high density of nodes in areas of the L-R and T-D layouts while other areas remain largely unoccupied.
Figure 8 Unweighted distribution of nodes in L-R layout for TS16. The bottom half of the graph has a very high level of occlusion; TS16 contains 572 nodes. Without weighting the top half of the graph lies mostly empty, containing only 8 nodes.
Figure 9 Weighted distribution of nodes in L-R layout for TS16. Compared to the unweighted distribution for the same DAG in Figure 8 this layout makes better use of space, aiding analysis in areas in areas of high density by reducing occlusion.
Figure 10 Zooming into a selection area using a sub-window. The selection area in the main window is redrawn in a sub-window, providing magnification by drawing the same number of nodes in a larger area.
Figure 11 Semantic and geometric zoom in the main window. A sub-tree of interest is expanded to take advantage of maximum screen space while suppressing surrounding data of less interest, providing both a semantic and a geometric zoom.
Figure 12 System response time with load for the 2D browser. Plots showing the exponential increase in system response time with data load for the 2D browser.
Figure 13 Extension to 3D to allow simultaneous visualisation of multiple datasets. Simultaneous display of four stages in a 3D window; DAGs are laid out in parallel planes equidistant from each other, with all root nodes lying in the same plane and the DAG growing downwards. The snapshot zooms into an ROI and shows textual detail brought up for a selected node (highlighted in red).
Figure 14 Relationships spanning data sets in the 3D browser. Colour-coded links are drawn between three DAGs to show relationships across stages, functionality that can also be used for tracing lineage.
Figure 15 Comparison of visualisation of grouping in 3D to 2D. Removing the group node to a plane parallel to that in which the DAG lies (in 3D) not only makes it easier to identify the group created, but also prevents the crossing of links that occurs for grouping in the 2D browser.
Figure 16 User satisfaction ratings for the structured evaluation of the visualisation prototypes. The two charts on the left show the mean overall satisfaction ratings for each of the 2D and 3D browsers. The first chart shows means for each user and the second, means by target user group. The colour coded reference lines for each chart show the mean overall satisfaction rating for each browser, while the black reference line shows the midpoint for the Likert scale (ranked 1–9) for the questions posed. The third chart plots SUS scores for each user (note there is no score for user 1).
Figure 17 Task completion times for the structured evaluation. The two charts on the left show mean task completion times within a 95% CI, while the third plots task completion times for each task for each user, compared to the expected maximum time (MTC) to complete each task.
Table 1 Proposals for changes to browsers prior to structured evaluation; changes implemented before evaluation are flagged.
Original implementation Suggestion(s) for improvement Implemented
Occlusion due to node labels: occurs even where occlusion of nodes is very low to none, especially for TD layout of the DAG 1. Hiding of labels (already implemented)
2. Interactive repositioning of nodes/labels
3. Drawing labels at a (user-defined) angle to horizontal plane √
X
X
Default labelling of nodes: set to component name Change to print name (full path to root) to aid differentiation between nodes with identical component names, e.g., TS12 has four nodes with component name mesenchyme. (Note that this solution increases the problem of occlusion due to node labels) √
(implemented in some cases)
Component detail: only component IDs provided in some cases Provision of print names in addition to component IDs, as otherwise required to look up names to identify nodes √ (partially implemented)
Search: Ability to highlight a specific search hit from within search dialog √
Creation of groups: group nodes could only be selected from list held in dialog Ability to click to select group nodes using the DAG
Ability to enter component IDs directly √
X
History/Undo function: not available Undo function encourages exploration, especially for 3D where navigation sometimes produces unexpected, undesirable results X
Storage of user sessions: not available Storage of user sessions – provides history function √
==== Refs
Benton D Bioinformatics – Principles and Potential of a New Multidisciplinary In Trends Biotechnology 1996 14 261 272
Gilbert DR Schroeder M van Helden J Interactive Visualization and Exploration of Relationships between Biological Objects Trends in Biotechnology 2000 18 487 494 11102659
Robinson AJ Flores TP Terry Gaasterland, Peter Karp, Kevin Karplus, Christos Ouzounis, Chris Sander, Alfonso Valencia Novel Techniques for Visualising Biological Information Proceedings of the Fifth International Conference on Intelligent Systems for Molecular Biology: Halkidiki 1996 The AAAI Press 241 249 21–25 June, 1997
Stevens RD Goble CA Baker P Brass A A Classification of Tasks in Bioinformatics Bioinformatics 2001 17 180 188 11238075
Shneiderman B Designing the User Interface: Strategies for Effective Human-Computer Interaction 1998 3 Addison-Wesley
Shneiderman B The Eyes Have It: A Task By Data Type Taxonomy for Information Visualizations Proceedings of the 1996 IEEE Symposium on Visual Languages: Boulder 1996 IEEE Computer Society 336 343 3–6 September, 1996
Wesson J Van Greunen D Research Papers: Human-Computer Interaction, Visualization and Collaboration: Visualisation of Usability Data: Measuring Task Efficiency Proceedings of the 2002 Annual Research Conference of the South African Institute of Computer Scientists and Information Technologists On Enablement Through Technology: Port Elizabeth 2002 11 18
Chalmers M Ingram R Pfranger C Adding Imageability Features to Information Displays Proceedings of the 9th Annual ACM Symposium on User Interface Software and Technology: Seattle 1996 ACM Press 33 39 6–8 November, 1996
Herman I Delest M Melançon G Tree Visualisation and Navigation Clues for Information Visualisation Computer Graphics Forum 1998 17 153 165
Herman I Melançon G Marshall MS Graph Visualization and Navigation in Information Visualization: A Survey IEEE Transactions on Visualization and Computer Graphics 2000 6 24 43
Purchase HC Calder P, Thomas B The Effects of Graph Layout Proceedings of 1998 Australasian Computer Human Interaction Conference: Adelaide 1998 IEEE Computer Society 80 86 30 November – 4 December 1998
Burger A Davidson D Baldock R Formalization of Mouse Embryo Anatomy Bioinformatics 2004 20 259 267 14734318
Bult CJ Richardson JE Blake JA Kadin JA Ringwald M Eppig JT Baldarelli RM Baya M Beal JS Begley DA Boddy WJ Bradt DW Butler NE Chu T Corbani LE Corradi J Davisson MT Garippa D Glass LH Grant PL Hill DP King BL Krupke DM Lennon-Pierce M Lutz CM Maltais LJ Mani P McCright I McKenzie L Naf D Ormsby JE Ramachandran S Reed DJ Shaw DR Szauter P Trombley LA Mouse Genome Informatics in a New Age of Biological Inquiry Proceedings of the IEEE International Symposium on Bio-Informatics and Biomedical Engineering 2000: Arlington 2000 29 32 8–10 November, 2000
Shneiderman B Williamson C Ahlberg C Dynamic Queries: Database Searching by Direct Manipulation Proceedings of the SIGCHI Conference on Human Factors in Computing Systems: Monterey 1992 ACM Press 669 670
The Protégé; Ontology Editor and Knowledge Acquisition System
Piccolo Home Page
Walrus Graph Visualization Tool
Hypergraph
VRMLgraph
Wilson S Kesselman J Java Platform Performance: Strategies and Tactics 2000 1 Addison-Wesley Pub Co
Preece J Rogers Y Sharp H Benyon D Holland S Carey T Human-Computer Interaction 1994 Addison-Wesley
Baldock R Bard J Burger A Burton N Christiansen J Feng G Hill B Houghton D Kaufman M Rao J Sharpe J Ross A Stevenson P Venkataraman S Waterhouse A Yang Y Davidson D EMAP and EMAGE: A Framework for Understanding Spatially Organised Data Neuroinformatics 2003 1 309 325 15043218
Batzoglou S Pachter L Mesirov J Berger B Lander ES Human and Mouse Gene Structure: Comparative Analysis and Application to Exon Prediction Proceedings of the Fourth Annual International Conference on Computational Molecular Biology 2000: Tokyo 2000 46 53
Burger A Systematic Nomenclature for Embryo Anatomy The Manchester Bioinformatics Week: Ontology Workshop: Manchester 2002 23–28 March 2002
Burger A XSPAN: A Cross-Species Anatomy Network The Standards and Ontologies for Functional Genomics (SOFG) Conference: Hinxton 2002 9 17–20 November 2002
Chambers JM Cleveland W Kleiner B Tukey P Graphical Methods for Data Analysis 1983 Wadsworth and Brooks/Cole Publishing
Faulkner X 2000 Usability Engineering, Macmillan Press Ltd
Graham M Kennedy J Benton D Towards a Methodology for Developing Visualisations International Journal of Human-Computer Studies 2000 53 789 807
Jordan PW Thomas B Weerdmeester BA McClelland IL Usability Evaluation in Industry 1996 Taylor & Francis
Lindholm T Yellin F The Java(TM) Virtual Machine Specification 1999 2 Addison-Wesley Pub Co
Nielsen J Molich R Jane Carrasco Chew, John Whiteside Heuristic Evaluation of User Interfaces Proceedings of the SIGCHI Conference on Human Factors in Computing Systems: Empowering People: Seattle 1990 249 256 April 01 – 05, 1990
Nielsen J 1993 Usability Engineering, Academic Press, Inc
Preece J A Guide to Usability: Human Factors in Computing 1993 Addison-Wesley
Robertson GG Card SK Mackinlay JD Information Visualization Using 3D Interactive Animation. Special Issue on Graphical User Interfaces Communications of the ACM 1993 36 57 71
Rubin J Handbook of Usability Testing: How to Plan, Design, and Conduct Effective Tests 1994 John Wiley & Sons, Inc
Slaughter L Norman KL Shneiderman B Assessing Users' Subjective Satisfaction with the Information System for Youth Services (ISYS) VA Tech Proc of Third Annual Mid-Atlantic Human Factors Conference: Blacksburg 1995 164 170 26–28 March, 1995
| 15790390 | PMC1087473 | CC BY | 2021-01-04 16:02:48 | no | BMC Bioinformatics. 2005 Mar 24; 6:74 | utf-8 | BMC Bioinformatics | 2,005 | 10.1186/1471-2105-6-74 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-811580197910.1186/1471-2105-6-81DatabaseColumba: an integrated database of proteins, structures, and annotations Trißl Silke [email protected] Kristian [email protected]üller Heiko [email protected] Thomas [email protected] Ina [email protected] Robert [email protected]ömmel Cornelius [email protected] Ulf [email protected] Institute of Informatics, Humboldt-Universität zu Berlin, Unter den Linden 6, 10099 Berlin, Germany2 Institute of Biochemistry, Charité Universitätsmedizin Berlin, Monbijoustraß e 2a, 10117 Berlin, Germany3 Zuse Institute Berlin, Takustrasse 7, 14195 Berlin, Germany4 Technische Fachhochschule Berlin, Seestr. 64, 13347 Berlin, Germany2005 31 3 2005 6 81 81 15 11 2004 31 3 2005 Copyright © 2005 TrißI 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
Structural and functional research often requires the computation of sets of protein structures based on certain properties of the proteins, such as sequence features, fold classification, or functional annotation. Compiling such sets using current web resources is tedious because the necessary data are spread over many different databases. To facilitate this task, we have created COLUMBA, an integrated database of annotations of protein structures.
Description
COLUMBA currently integrates twelve different databases, including PDB, KEGG, Swiss-Prot, CATH, SCOP, the Gene Ontology, and ENZYME. The database can be searched using either keyword search or data source-specific web forms. Users can thus quickly select and download PDB entries that, for instance, participate in a particular pathway, are classified as containing a certain CATH architecture, are annotated as having a certain molecular function in the Gene Ontology, and whose structures have a resolution under a defined threshold. The results of queries are provided in both machine-readable extensible markup language and human-readable format. The structures themselves can be viewed interactively on the web.
Conclusion
The COLUMBA database facilitates the creation of protein structure data sets for many structure-based studies. It allows to combine queries on a number of structure-related databases not covered by other projects at present. Thus, information on both many and few protein structures can be used efficiently. The web interface for COLUMBA is available at .
==== Body
Background
Biological databases have become a major resource for researchers in life science. With the constantly increasing number of data deposited and the computational tools evolving, the focus of research has shifted from the study of a single gene towards an intra- and inter-species comparison of genes and gene products. This trend can also be seen in the field of structural biology, where the number of protein structures deposited in the Protein Data Bank, PDB [1] is increasing rapidly. However, looking at the structure alone is not sufficient for a comprehensive study of the various types of relationships between proteins. Other types of information, such as functional and structural annotations of proteins, also have to be taken into account.
Oberg and colleagues [2] compared the results from infrared and circular dichroism spectroscopy with the actual 3D structure of a protein to gain insight into the relationship between assigned protein secondary structures and spectral band shape. To carry out this study, they had to compile a set of proteins based on the folding classification as defined by CATH, the content of secondary structure elements computed by the DSSP program, and the commercial availability of the proteins. Martin and colleagues [3] systematically explored the relationship between the folding classification from CATH and the classification of proteins into ENZYME classes. For that purpose, they needed groups of structurally resolved proteins belonging to one of the six main ENZYME classes. In both examples, the first step in the experiments was the compilation of a set of protein structures based on the structure itself and on folding classification, sequence properties, enzymatic activity, and other types of information.
Researchers have several possibilities to collect information on protein structures. First, entries in the PDB itself contain a set of full text information and often are annotated with links to external data sources. However, PDB entries are not curated, only archived by the PDB team. This has two consequences. First, the data are not constantly updated and therefore quickly become out-of-date. Second, the annotation provided by different submitters is highly heterogeneous and does not follow a standardized nomenclature. As a consequence, searching the PDB for annotations is an error-prone task. Annotations may be incomplete or inconsistent with standard nomenclature, spelling errors and uncontrolled usage of abbreviations prevent an efficient textual search, and literature references or links to functional and structural databases may be outdated or missing. Examples of such problems are described in [4]. This lack has led to a number of second-party databases that digest PDB entries and attach a wealth of links to relevant databases. The two best-known sources of that kind are probably PDBsum [5] and the IMB Jena Image Library [6]. Both store hyperlinks to external databases and not the actual information. Therefore they are well suited for human browsing of single entries, but inadequate for working with sets of structures and their properties. Imagine a researcher wants to compile the set of DNA binding proteins from mammals resolved by X-ray crystallography with a resolution lower than 3.2 Å. Solving this task can be achieved by using either PDBsum or the IMB databases, but it requires extensive manual work or the coding of specialized scripts [7].
To overcome this problem, we created COLUMBA, a database of information on protein structures that physically integrates information from twelve protein structure related data sources into a single data warehouse. Besides the protein structures themselves, COLUMBA covers structural and sequence-based classification schemes, functional annotation, secondary structure elements, and participation in metabolic pathways. Links between these data and the protein structures, both on the chain, compound, and entry level, are either taken from the second-party databases or are computed inside COLUMBA, leading to more accurate and more current information than available in the PDB itself and as current as possible, we compute links between chains and Swiss-Prot entries based on sequence similarity, thus cross-referencing 68% of the PDB entries to a Swiss-Prot sequence.
Construction and content
Data sources
COLUMBA is centered around PDB entries [1]. For each entry we store general information like the experimental method, resolution, deposition date, and author. Each PDB entry is organized in compounds, which represent biological units, and each compound has one or more chains. A compound, for which an enzyme classification (E.C.) number exists, is annotated with information from ENZYME [8] for the enzyme name and biochemical reaction, and with data from the Kyoto Encyclopedia of Genes and Genomes, KEGG [9] for the participation of that enzyme in metabolic pathways. COLUMBA also integrates data from the Roche Biochemical Pathway Map [10].
To gain information about protein domains, entries from the protein-fold classification databases SCOP [11] and CATH [12] are linked to protein chains. Furthermore, each chain is assigned to a PISCES cluster [13]. PISCES groups protein chains according to their sequence identity and experimental properties into culled sets. For each chain, the secondary structure is computed using the DSSP program [14]. Links to Swiss-Prot entries [15] were retrieved from the PDBSprotEC database [16]. Exploiting the links from Swiss-Prot to other databases, PDB chains are connected to the NCBI Taxonomy database [17] and functional annotation from Gene Ontology [18].
Architecture and database schema
All data sources integrated into COLUMBA describe specific aspects of either PDB entries itself, their compounds, or their chains. We never mix data from different data sources with each other. This partitioning is directly reflected in the database schema (see Figure 1), where we model each data source as a different dimension in which protein structures are annotated. Each data source occupies its own, specialized subschema within the overall schema of COLUMBA. As an example, the subschema of KEGG consists of three tables, one for the metabolic pathway names, one for the enzyme names, and the third table stores information about enzymes participating in pathways. Each subschema is linked to the central subschema representing PDB entries. This "separation of concerns" is also reflected in the Web interface.
Integration of data sources into COLUMBA
COLUMBA is implemented on top of the open source database system PostGreSQL [19]. It currently integrates data from the twelve data sources as shown in Table 1. The data from the original sources are available in different formats, such as flat files, database dump files, or pure HTML pages. We use parsers, written in Python and Perl, respectively, to populate COLUMBA with the data obtained in non-relational representation. For PDB we use our own parser derived from the BioPython project [20]. To upload Swiss-Prot, Gene Ontology, and NCBI Taxonomy we use the parsers and schema provided in the BioSQL project [21]. After parsing each data source into a separate database schema, the data in those schemas are mapped into the COLUMBA target schema. Program source of our parsers is available on request. The connections between data sources and the PDB data are generally established by using existing links. Links from PDB to ENZYME, KEGG, and the Boehringer map are obtained through the E.C. number given in PDB entries. DSSP secondary structures are computed directly on the chains. The connection between PDB chains and Swiss-Prot entries is established by using the information from the PDBSprotEC database [16]. Swiss-Prot is also used as intermediate information for connecting PDB entries to the NCBI Taxonomy and Gene Ontology Annotation [22].
Annotation workflow
The annotation workflow populates the COLUMBA data warehouse and establishes connections between PDB entries and the other data sources. Each data source is represented by a software module implementing a fixed interface. Once a new PDB entry is written into COLUMBA, a workflow manager triggers each module, which adds annotations to the entry. The implementation of modules varies according to the nature of the data source. For instance, the DSSP module calls the DSSP program to compute the secondary structure for each chain, whereas the SCOP module searches PDB and chain identifiers in external files. Our annotation pipeline is able to handle logical dependencies between different modules. This architecture allows to include a new data source by just extending the database schema for the new tables, and implementing an appropriate module.
Content of COLUMBA
COLUMBA is populated with data using the annotation workflow described in the previous section. New entries from the Protein Data Bank are added regularly to COLUMBA, and links to the other data sources are established upon this import. Data sources with a release policy, such as Swiss-Prot, SCOP or CATH are updated according to new releases. All other data sources are updated as new data becomes available. Table 2 lists the number of PDB entries, broken down to compounds and chains that have an annotation in the respective sources and combinations of sources.
Utility
COLUMBA is a relational, integrated database of information on protein structures and is specially designed to support the creation of sets of protein structures sharing annotations in any of the data sources. Sets as those described in the introduction can be compiled with a few mouse-clicks using COLUMBA.
Web interface
COLUMBA can be searched through a web interface available at . The interface allows two types of queries: Full text search as well as data source and attribute specific searches. In both cases, the query results in a list of PDB entries with their corresponding chains.
For convenience and as a quick-start, COLUMBA can be searched by using a standard keyword search over all textual fields in COLUMBA (Figure 2A), including the annotation given by the PDB, enzymatic, metabolic, taxonomic, and the protein-fold classification information. Keywords can be combined using logical AND, OR, and NOT operators. The keyword search performs a simultaneous request over the content of all integrated data sources, and is thus a quick and easy-to-use option for finding interesting protein structures. However, it does not allow for source- or attribute specific queries, e.g., for finding all protein structures, which are specifically annotated in CATH as containing a Rossmann fold. The main focus of COLUMBA is the compilation of sets of structures sharing properties from different second-party databases. To support such queries, we have created a specialized web interface based on the paradigm of query refinement. This process is best understood as having an initial data set, which is subsequently reduced by applying different filters. In our case, the initial data set contains the entire set of PDB entries. For each of the data sources integrated into COLUMBA, the user may specify source-specific filter conditions using a proper web form (see Figure 2B). The source specific forms can be found by using the labeled buttons on the left side of the web page. After entering conditions in a form, those PDB entries that do not fulfill the stated conditions are removed from the current set of results. Several forms can be used consecutively, thus restricting the original set of all PDB entries by conditions on multiple data sources. Conditions on different sources are always logically connected by an AND. The available search operators depend on the specific field and data source, ranging from numerical comparisons to substring matching and traversal of ontological structures. To guide the user, COLUMBA constantly shows the current number of qualifying PDB entries after each query step in the header of the page. This demonstrates the consequences of adding, deleting, or changing conditions and helps to prevent the over-specification of search conditions leading to empty sets. Note that the full-text search can be used as an additional restriction condition on the result set, which has turned out to be a quite powerful feature of the search interface.
Once the user has specified all desired conditions, COLUMBA computes the qualifying set of protein structures. This list of results (see Figure 3A) gives basic information, such as PDB ID, experimental method, compound name, and chains for each entry. The PDB entry ID links to the COLUMBA Explorer view for that particular entry. The Explorer (see Figure 3B) shows all information stored in COLUMBA for that PDB entry. This includes the experimental method and resolution for each entry and compound name, metabolic information, and the source organism for each compound. Detailed information is given for each chain, including protein-fold classification from SCOP and CATH, data from the according Swiss-Prot entry, Gene Ontology annotation, and NCBI taxon name. These data can also be viewed or downloaded in XML format. We also provide on-line molecular visualisation via JMol [23], and links to the original data items in the respective databases.
To further enhance the search capabilities of the web interface, it is possible to upload a file containing a set of PDB identifiers. Thus, a user can view all data in COLUMBA for the entries in his list and create subsets of protein structures from the list by entering conditions on second-party annotations. Thereby, the COLUMBA web interface greatly reduces the required time to collect additional information for entries in any list of PDB entries.
Example of use
Consider a query for all compounds from ENZYME class '1.-.-.-' containing a chain with a TIM barrel fold (see Figure 3A). To compute this set, a user first specifies 'TIM barrel' in the full text search form, which returns all PDB chains with the keyword 'TIM barrel' in any of the data sources, including the PDB, SCOP, and CATH annotation. Next, the set of all proteins fulfilling this condition can be intersected with the result of the search for the ENZYME class in the metabolism form. The intersection contains 95 PDB structures. However, using the full text search is only one option for finding the appropriate answer. In general, different answers are possible for a given question depending on the preferences and trust of the user in different databases. Consider again the example given above. If a user has high confidence in either CATH or SCOP, he may specify a condition using the CATH or SCOP form, respectively, instead of performing the full text search for 'TIM barrel'. This results in 79 entries when relying only on CATH and 90 entries for SCOP. The user might even want to restrict the search to only those chains that are annotated as containing a 'TIM barrel' fold in both CATH and SCOP. The returned set has 79 PDB protein structures. These differences result from the fact that COLUMBA usually only takes the cross-references given in the original data and does not curate or amend the content of the integrated databases.
Applications of COLUMBA
The web interface is designed to compile sets of protein structures sharing properties from protein structure related sources, but it is possible to tackle more sophisticated issues by exploiting the relational data warehouse of COLUMBA. We show a number of applications where we used SQL (Structured Query Language) to retrieve information.
A research question concerning the participation of enzymes in metabolic pathways arose from an article from Martin et al. [3] that investigated the relationship between the protein classification of ENZYME and the folding classification of CATH. One finding at that time was that the known enzymes in the glycolytic pathway contained a very limited set of different CATH architectures and topologies. This naturally raises the questions whether this is the case for other metabolic pathways as well. We used COLUMBA to address this problem, combining PDB data, information on metabolic pathways from KEGG, and the CATH classification.
Each KEGG pathway consists of a number of enzymes related to a PDB compound. Those compounds are linked to the respective chains, which in turn are cross-referenced to CATH classes, architectures, and topologies. We computed the number of occurrences of CATH classes for the set of enzymes in a pathway containing more than 10 enzymes and having a coverage of at least 50% with PDB structures that are also annotated in CATH. For all qualifying pathways the figures are given in Table 3.
The first striking fact is that only 26% of the enzymes participating in KEGG pathways do have annotated chains in CATH. This is because just 34% of the enzymes in KEGG are structurally resolved, of which several are not annotated by CATH. The enzymes within the annotated set contain four times as many domains with an Alpha/Beta class in CATH than Mainly Alpha and Mainly Beta, respectively. In comparison, for all proteins annotated by CATH the Alpha/Beta class only occurs twice as often as each of the other two folds.
In Figure 4 the subdivision of all enzymes (Figure 4A) as well as of selected pathways into classes, architectures, and topologies from the protein-fold classification CATH is shown in 'CATH wheels'. The predominant CATH architecture in all three 'CATH-wheels' is the 3-Layer(aba) Sandwich, with the 'Rossmann fold' comprising the biggest topology. In Figure 4B the Pyrimidine metabolism is shown. As we can see, the shares of the different classes are almost equal to the distribution of classes in all enzymes. Figure 4C shows the Glycolysis/Gluconeogenesis pathway, where in 1998 only 11 enzymes were known. These enzymes exhibited mostly an Alpha/Beta fold. By now, 24 enzymes are resolved and structurally classified by CATH, which lead to more domains that differ from the predominant Alpha/Beta fold. As more and more enzymes become structurally resolved in the future, this picture will shift yet again.
Discussion
Related work
The most frequent approach to the interconnection of data on protein structures that are spread over multiple original data sources is the usage of hyperlinks. Examples are PDBsum [5] and the IMB Jena Image Library [6]. This method is well suited for human browsing of single entries, but as soon as it comes to handling sets of objects, following many hyperlinks becomes a tedious and time consuming task. Efficient handling of sets can only be achieved if data are physically integrated into a single system. In the protein structure world, there are three main such databases apart from COLUMBA. 3DinSight [24] focuses on visualization of sequence features such as PROSITE patterns or altered positions in the 3D structure. iProClass [25] concentrates on protein sequence and integrates 50 different databases using so-called 'rich links'. Finally, BioMolQuest [26] integrates in total four data sources, thus storing only a subset of the information available in COLUMBA. Currently, the Protein Data Bank itself is preparing a new web interface to provide not only the links to related sources, but the actual information from SCOP, CATH, and the Gene Ontology. These are only a subset of the sources integrated in COLUMBA. COLUMBA's functionality could also have been achieved by implementing specific modules for SRS. However, we early on decided to use relational database technology instead of the highly proprietary SRS languages and methods.
Two groups currently address the problem of inconsistent use of terminology in the PDB: the PDB uniformity project [4] and the Macromolecular Structure Database MSD [27]. Both projects aim at correcting PDB entries, unifying terminology, and adding or updating links to scientific references. The MSD also addresses the linkage of PDB chains to Swiss-Prot entries. We hope that these efforts will make our work easier in the near future, for instance if the PDB entries themselves come with consistent and structured taxonomic information.
COLUMBA currently integrates twelve data sources concerned with different aspects of protein sequences and structures. Notably, COLUMBA does not store the coordinates of structures themselves but is designed to enable users to find 'the right' set of structures based on annotations. This is by intention, since there are already many programs that can efficiently parse, visualize, or compare protein structures from PDB files.
An important design principle of COLUMBA is that it never mixes data from different sources into a single table. Each data source is considered as a dimension in which PDB entries, compounds, and chains are annotated. We call this approach multidimensional data integration [28], which is inspired by data warehouse design, where facts, e.g., sales, are described by dimensions, such as store, product, or customer [29]. The resulting database schema is called star schema in correspondence with the visual appearance. We also use a star schema like structure with the tables holding information from protein structures in the center of a set of tables containing the data from other data sources.
Our approach is in contrast to projects that aim at a tighter semantic integration, merging logically similar types of information into a single table. Such a semantic integration approach was for instance followed in the TAMBIS project [30]. However, we strongly believe that merging data from different databases is counterproductive for the biologist because it blurs important differences. On the other hand, keeping data separated inevitably leads to a certain degree of semantic redundancy, i.e., different schema elements provide the same type of information. For instance, functional annotation of proteins is encoded both in Swiss-Prot keywords and Gene Ontology terms; 'TIM barrels' are annotated in CATH, SCOP, and the PDB annotation itself. But this redundancy does not originate from data duplication, but rather from evidence obtained independently by different people or by different experiments. These evidences are important in their own right.
We believe that the advantages of our approach prevail for mainly two reasons:
• Users recognize the origin of the data they query and obtain as result. In our experience, biologists often have their favorite set of databases, where they know about the pitfalls and peculiarities. By keeping data separated, personal preferences or differences in trust in particular databases can be expressed and the results can be judged based on prior experience.
• Subtle differences in the semantics of fields of different databases are conserved. For instance, both Swiss-Prot keywords and GO annotations express functional annotation. However, the process of creating this annotation is quite different, and it is often meaningful to discriminate between the two.
Furthermore, separating data and software for the different data sources greatly simplifies system maintenance. Changes to data sources, including the deletion or addition of data sources, only affect a well defined part of the schema and of the web interface.
Our perception of considering annotation sources as dimensions describing some primary objects is also followed in the EnsMart project [31]. EnsMart uses a 'reversed star schema' to connect genes with different types of information, such as genomic position, transcription factors, or expression data. The data are queried through a generic web interface, which also allows source-specific queries and their combinations. Conceptually, EnsMart and COLUMBA are very similar, but they work on totally different types of data. Moreover, COLUMBA is directly designed for handling annotations of protein structures, which has advantages in terms of result visualization and search options.
Conclusion
COLUMBA has proven to be very useful for a number of tasks in our own structural research. Generating sets of structures, which previously required days of manual browsing or writing of parsers, now only takes a few mouse clicks, or an SQL query. Once the set of PDB entries and chains is obtained, there are many other programs for visualizing or comparing structures. COLUMBA's future development will further concentrate on annotation of structure in contrast to the structure and its coordinates itself. The next data sources to be integrated are those covering protein domains and motifs, i.e., InterPro [32] and its relatives. In the long run, we will push COLUMBA towards a medical orientation. Obvious candidates for being integrated are literature abstracts from Medline and the OMIM database [33]. The LIGAND database [34] will provide information about small molecules interacting with proteins to use COLUBMA for the prediction of drug target sides. Moving towards medical data is a natural next step since much of structural research, including our own [35] is concerned with drug development.
Availability
The database is available at .
Authors' contributions
ST designed and implemented the web-interface. KR was responsible for the implementation of the annotation workflow. HM helped importing PDB data. UL designed the overall architecture and co-supervises the project. TS maintains the server infrastructure. IK provided data on the classification of protein folds, and RP on protein ligands. CF co-supervised the project and was the main source of motivation for building an integrated protein annotation database.
Acknowledgements
This work is supported by BMBF grant no. 0312705B (Berlin Center for Genome-Based Bioinformatics). We thank Raphael Bauer, Rene Heek, and Stefan Günther for implementing many parts of Columba. We acknowledge the excellent work of the database maintainers of the different source databases and thank for their consent to integrate their data into COLUMBA.
Figures and Tables
Figure 1 Schematic entity-relationship model of COLUMBA. The dark gray part in the middle is the subschema that originates from the Protein Data Bank (PDB). The other subschemas are represented by a single box indicating the name of the data source and are grouped according to a broad classification of their content.
Figure 2 Screen shots of COLUMBA web-forms. (A) Interface for the full text search. (B) Query form for the metabolism information, where the result set can be restricted by information from ENZYME and KEGG.
Figure 3 Screen shots of COLUMBA query results. (A) Result set for a query requesting structures from the ENZYME class '1.-.-.-' combined with a full text condition on 'TIM barrel'. (B) COLUMBA Explorer detailed view of the PDB structure 1d3h.
Figure 4 The CATH wheel for KEGG pathways. The color of the CATH wheel represents the CATH classes, where yellow stands for alpha/beta, red for mainly alpha, blue for mainly beta, and green for Few Secondary Structures. The inner circle represents the CATH architectures (C.A.), where the width of each segment represents the number of enzymes found to exhibit that architecture. The outer circle stands for the Topology (C.A.T.). (A) shows the distribution of all enzymes participating in KEGG pathways with the '3-layer(aba) sandwich' representing the largest architecture. (B) shows the CATH wheel for the pathway 'Pyrimidine metabolism' while (C) for 'Glycolysis/Gluconeogenesis'.
Table 1 Data sources integrated in COLUMBA.
Source download page format Parsed by
PDB flat file BioPython
SCOP flat file BioPython
CATH flat file own
DSSP computed - own
ENYZME flat file BioPython
Boehringer HTML own
KEGG HTML own
Swiss-Prot flat file bioSQL
GO flat file bioSQL
GOA DB dump COPY
Taxonomy flat file bioSQL
PISCES DB dump own
The forth column gives the parsers used.
Table 2 Number of entries from the PDB.
from PDB to Number of entries
SCOP 42 908
CATH 32 825
DSSP 54 028
chains (total 60 241) Swiss- Prot 36 651
NCBI Taxonomy 36 651
Gene Ontology via GOA 36 008
PISCES 8 367
SCOP & CATH 32 439
SCOP & CATH & Swiss-Prot 27 972
Enzyme 12 510
Boehringer 5 029
7162 compounds (total 33 779) KEGG 7 162 7 162 9 172
Enzyme & KEGG 7 162
Enzyme & SCOP & CATH 9 172
Enzyme & SCOP & CATH & KEGG 5 054
Enzyme & Swiss-Prot 9 440
entries (total 26 104) all minus PISCES 2 868
all 621
They are divided into compounds and chains, which link to second-party databases and selected combinations of them.
Table 3 The number of enzymes for selected metabolic pathways from KEGG.
Metabolic pathway Enzyme total CATH class
Total with str. coverage a / b a b Few
all pathways 1 952 508 26,0 443 114 107 15
Fatty acid biosynthesis (path 1) 14 7 50,0 6 0 2 0
Oxidative phosphorylation 10 5 50,0 3 3 3 1
Streptomycin biosynthesis 14 7 50,0 6 0 1 0
Pyrimidine metabolism 59 30 50,8 29 6 5 0
Selenoamino acid metabolism 21 11 52,3 11 2 2 1
Pentose phosphate pathway 33 18 54,5 17 2 2 2
Methionine metabolism 23 13 56,5 13 3 1 1
One carbon pool by folate 23 13 56,5 13 2 1 0
Phe, Tyr and Trp biosynthesis 31 19 61,2 18 6 2 0
Glycolysis / Gluconeogenesis 38 24 63,1 24 2 5 2
Reductive carboxylate cycle (CO2 fixation) 13 9 69,2 8 3 1 1
Aminoacyl-tRNA biosynthesis 21 16 76,1 16 8 6 0
Carbon fixation 23 18 78,2 18 2 3 0
The sum of observations in CATH classes can be higher than the number of enzymes with structures from the pathway, because in one chain, several domains with distinct folds can occur.
==== Refs
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
Oberg K Ruysschaert J Goormaghtigh E Rationally selected basis proteins: A new approach to selecting proteins for spectroscopic secondary structure analysis Protein Sci 2003 12 2015 2031 12931000 10.1110/ps.0354703
Martin AC Orengo CA Hutchinson EG Jones S Karmirantzou M Laskowski RA Mitchell JB Taroni C Thornton JM Protein folds and functions Structure 1998 6 875 884 9687369 10.1016/S0969-2126(98)00089-6
Bhat T Bourne P Feng Z Gilliland G Jain S Ravichandran V Schneider B Schneider K Thanki N Weissig H Westbrook J Berman HM The PDB data uniformity project Nucleic Acids Res 2001 29 214 218 11125095 10.1093/nar/29.1.214
Laskowski RA PDBsum: summaries and analyses of PDB structures Nucleic Acids Res 2001 29 221 222 11125097 10.1093/nar/29.1.221
Reichert J Sühnel J The IMB Jena Image Library of Biological Macromelecules: 2002 update Nucleic Acids Res 2002 30 253 254 11752308 10.1093/nar/30.1.253
Stein L Creating a bioinformatics nation Nature 2002 417 119 120 12000935 10.1038/417119a
Bairoch A The ENZYME database in 2000 Nucleic Acids Res 2000 28 304 305 10592255 10.1093/nar/28.1.304
Kanehisa M Goto S Kavashima S Okuno Y Hattori M The KEGG resource for deciphering thegenome Nucleic Acids Res 2004 32 D277 D280 14681412 10.1093/nar/gkh063
Michal G Biochemical Pathways Boehringer Mannheim GmbH 1993
Murzin A Brenner S 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
Orengo C Michie A Jones S Jones D Swindells M Thornton J CATH – a hierarchic classification of protein domain structures Structure 1997 5 1093 1108 9309224 10.1016/S0969-2126(97)00260-8
Wang G Dunbrack RL Jr PISCES: a protein sequence culling server Bioinformatics 2003 19 1589 1591 12912846 10.1093/bioinformatics/btg224
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
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
Martin AC PDBSprotEC: a Web-accessible database linking PDB chains to EC numbers via SwissProt Bioinformatics 2004 20 986 8 14764547 10.1093/bioinformatics/bth048
Wheeler D Chappey C Lash A Leipe D Madden T Schuler G Tatusova T Rapp B Database resources of the National Center for Biotechnology Information Nucleic Acids Res 2000 28 10 14 10592169 10.1093/nar/28.1.10
Gene Ontology Consortium The Gene Ontology (GO) database and informatics resource Nucleic Acids Res 2004 32 D258 261 14681407 10.1093/nar/gkh036
PostGreSQL
BioPython
BioSQL
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 D266 14681408 10.1093/nar/gkh021
Jmol
An J Nakarna T Kubota Y Sarai A 3DinSight: an integrated relational database and search tool for the structure, function and properties of biomolecules Bioinformatics 1998 14 188 195 9545451 10.1093/bioinformatics/14.2.188
Wu CH Huang H Nikolskaya A Hu Z Barker WC The iProClass integrated database for protein functional analysis Comput Biol Chem 2004 28 87 96 15022647 10.1016/j.compbiolchem.2003.10.003
Bukhman YV Skolnick J BioMolQuest: integrated database-based retrieval of protein structural and functional information Bioinformatics 2001 17 468 478 11331242 10.1093/bioinformatics/17.5.468
Boutselakis H Dimitropoulos D Fillon J Golovin A Henrick K Hussain A Ionides J John M Keller P Krissinel E McNeil P Naim A Newman R Oldfield T Pineda J Rachedi A Copeland J Sitnov A Sobhany S Suarez-Uruena A Swaminathan J Tagari M Tate J Tromm S Velankar S Vranken W E-MSD: the European Bioinformatics Institute Macromolecular Structure Database Nucleic Acids Res 2003 31 458 462 12520052 10.1093/nar/gkg065
Rother K Müller H Trissl S Koch I Steinke T Preissner R Frömmel C Leser U Rahm E Columba: Multidimensional Data Integration of Protein Annotations DILS, Volume 2994 of Lecture Notes in Computer Science 2004 Springer 156 171
Chaudhuri S Dayal U An Overview of Data Warehousing and OLAP Technology SIGMOD record 1997 26 65 74
Baker PG Brass A Bechhofer S Goble C Paton N Stevens R Glasgow J, Littlejohn T, Major F, Lathrop R, Sankoff D, Sensen C, Montreal TAMBIS: Transparent Access to Multiple Bioinformatics Information Sources 6th Int Conf on Intelligent Systems for Molecular Biology 1998 Canada: AAAI Press, Menlo Park 25 34
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 14 160 169 14707178 10.1101/gr.1645104
Apweiler R Attwood T Bairoch A Bateman A Birney E Biswas M Bucher P Cerutti L Corpet F Croning M Durbin R Falquet L Fleischmann W Gouzy J Hermjakob H Hulo N Jonassen I Kahn D Kanapin A Karavidopoulou Y Lopez R Marx B Mulder N Oinn T Pagni M Servant F Sigrist C Zdobnov E The InterPro database, an integrated documentation resource for protein families, domains and functional sites Nucleic Acids Res 2001 29 37 40 11125043 10.1093/nar/29.1.37
McKusick V Mendelian Inheritance in Man A Catalog of Human Genes and Genetic Disorders 1998 Johns Hopkins University Press, Baltimore
Goto S Okuno Y Hattori M Nishioka T Kanehisa M LIGAND: database of chemical compounds and reactions in biological pathways Nucleic Acids Res 2002 30 402 404 11752349 10.1093/nar/30.1.402
Preissner R Goede A Frömmel C Dictionary of interfaces in proteins (DIP). Data bank of complementary molecular surface patches J Mol Biol 1998 280 535 550 9665855 10.1006/jmbi.1998.1878
| 15801979 | PMC1087474 | CC BY | 2021-01-04 16:02:50 | no | BMC Bioinformatics. 2005 Mar 31; 6:81 | utf-8 | BMC Bioinformatics | 2,005 | 10.1186/1471-2105-6-81 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-961582630910.1186/1471-2105-6-96SoftwarePREP-Mt: predictive RNA editor for plant mitochondrial genes Mower Jeffrey P [email protected] Department of Biology, Indiana University, Bloomington, IN, 47405, USA2005 12 4 2005 6 96 96 10 2 2005 12 4 2005 Copyright © 2005 Mower; 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 plants, RNA editing is a process that converts specific cytidines to uridines and uridines to cytidines in transcripts from virtually all mitochondrial protein-coding genes. There are thousands of plant mitochondrial genes in the sequence databases, but sites of RNA editing have not been determined for most. Accurate methods of RNA editing site prediction will be important in filling in this information gap and could reduce or even eliminate the need for experimental determination of editing sites for many sequences. Because RNA editing tends to increase protein conservation across species by "correcting" codons that specify unconserved amino acids, this principle can be used to predict editing sites by identifying positions where an RNA editing event would increase the conservation of a protein to homologues from other plants. PREP-Mt takes this approach to predict editing sites for any protein-coding gene in plant mitochondria.
Results
To test the general applicability of the PREP-Mt methodology, RNA editing sites were predicted for 370 full-length or nearly full-length DNA sequences and then compared to the known sites of RNA editing for these sequences. Of 60,263 cytidines in this test set, PREP-Mt correctly classified 58,994 as either an edited or unedited site (accuracy = 97.9%). PREP-Mt properly identified 3,038 of the 3,698 known sites of RNA editing (sensitivity = 82.2%) and 55,956 of the 56,565 known unedited sites (specificity = 98.9%). Accuracy and sensitivity increased to 98.7% and 94.7%, respectively, after excluding the 489 silent editing sites (which have no effect on protein sequence or function) from the test set.
Conclusion
These results indicate that PREP-Mt is effective at identifying C to U RNA editing sites in plant mitochondrial protein-coding genes. Thus, PREP-Mt should be useful in predicting protein sequences for use in molecular, biochemical, and phylogenetic analyses. In addition, PREP-Mt could be used to determine functionality of a mitochondrial gene or to identify particular sequences with unusual editing properties. The PREP-Mt methodology should be applicable to any system where RNA editing increases protein conservation across species.
==== Body
Background
RNA editing is a type of RNA processing (such as polyadenylation, intron splicing, and 5' end-capping) that inserts, deletes, or modifies nucleotides in an RNA transcript, thereby changing the information encoded by the genome. First discovered in trypanosome mitochondria [1], RNA editing has since been observed in a range of eukaryotes, including slime molds, amoeboid protozoans, plants, animals, and fungi, and also in viruses [2,3]. In plants, RNA editing converts cytidines to uridines and uridines to cytidines in mitochondrial and plastid, but not nuclear, transcripts. The frequency and type of conversion in each organelle is highly lineage-specific [4-7]. In angiosperms, for example, approximately 400 editing sites (all C to U) have been found in the 30 to 40 mitochondrial protein-coding genes [8-10], but only about 30 C to U sites were seen across over 100 plastid genes [11-13]. In contrast, both types of conversion are found with high abundance in mitochondrial and plastid transcripts of ferns and hornworts [4-6,14,15].
In all plant lineages, RNA editing most often alters the amino acid sequence encoded by protein-coding genes, but it can occasionally generate new start codons, create or remove stop codons, or cause silent changes that do not affect the protein sequence. RNA editing has also been observed in tRNA genes, untranslated regions, and introns [2], although the frequency of editing in these regions appears to be much lower. One feature observed immediately upon the discovery of RNA editing in plants is that edited transcripts code for proteins that are more conserved across species than proteins predicted from genomic DNA [16-18]. In fact, this tendency for codon "correction" was one of the clues that led to the discovery of RNA editing, since proteins predicted from early plant mitochondrial DNA sequences contained biochemically distinct amino acids at positions that are otherwise conserved throughout eukaryotes [16-18]. These initial observations have been repeatedly confirmed in almost all subsequent studies of editing in plants, with the most notable exceptions occurring in pseudogenes [19-21], which presumably have no selective constraints on their editing sites.
Because of the changes induced by RNA editing, the protein sequences encoded by mature mitochondrial transcripts are often quite different from what is encoded by the genomic DNA. In order to correctly analyze plant mitochondrial sequences in phylogenetic, molecular, or biochemical studies, RNA editing information must be known. Experimental determination, via direct comparison of the RNA transcript sequence and genomic DNA sequence, is the de facto standard for identification of sites of RNA editing. Given that these experimental analyses take time and cost money, however, two general approaches have been used to predict sites of RNA editing. The first relies on the possibility that the sequence context of an edited site contains information that signals the editing machinery or associated specificity factors. Indeed, experimental analyses of the surrounding sequence context indicate that nucleotides upstream and downstream are important in specifying sites of editing [22,23]. Furthermore, over 90% of editing sites have a pyrimidine at the adjacent upstream nucleotide [8,24]. Unfortunately, attempts at identifying consensus motifs beyond this one important nucleotide have met with little success [8,24,25]. The second predictive approach exploits the tendency of RNA editing to increase protein conservation across diverse taxa. Because of this "correcting" nature of RNA editing, it is possible to scan a protein sequence alignment for unconserved amino acids. Very often, when these unconserved amino acids have the potential to be corrected by RNA editing, they are actually edited. This approach has been shown to be very successful in predicting sites of RNA editing for several genes [6,25,26], and has also been used to infer the absence of RNA editing in the entire mitochondrial genomes of Marchantia polymorpha, a complex thalloid liverwort, and the green algae Chara vulgaris and Chaetosphaeridium globosum [27-29]. Limited experimental evidence has so far corroborated the lack of editing in these lineages [4-6,30].
In order to test the generality of the second predictive approach for any plant mitochondrial protein-coding gene, the PREP-Mt program was designed to predict editing sites using protein sequence comparisons and the correcting nature of RNA editing. Because the test results indicate that PREP-Mt is both fast and accurate, an online tool was also developed [31]. This resource should be useful now since editing sites have been experimentally determined for only a small percentage of plant mitochondrial genes available in the sequence databases, and it will become increasingly useful as more mitochondrial genomes get sequenced in the near future. PREP-Mt may also be effective in discriminating between functional genes and pseudogenes, as well as in elucidating mechanisms of RNA editing by exposing examples of genes that do not conform to the normal editing patterns in plant mitochondria.
Implementation
Construction of the Aligned Sequence Database
395 full-length or nearly full-length plant mitochondrial protein-coding genes, for which RNA editing sites have been experimentally determined or from organisms (Marchantia, Chara, and Chaetosphaeridium) inferred to lack RNA editing capability, were collected from Genbank. Gene sequences were extracted from each file and then edited according to Genbank annotations or literature sources. The edited gene sequences were translated into proteins according to the standard genetic code. Homologous proteins were aligned using ClustalW version 1.81 [32] and manually adjusted when necessary. The Aligned Sequence Database (ASD) consisted of these protein sequence alignments (see additional files 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42 for alignments).
In some cases, editing site annotations in the Genbank files were associated with nucleotides that were not a cytidine. These incorrect annotations were usually the result of obvious human errors and were corrected by referring to the literature sources prior to inclusion in the database. In addition, nine sequences from Marchantia (atp4, atp8, ccmFc-A, ccmFc-B, ccmFn, rpl2, rps1, rps3, rps4), eight from Chara (atp4, atp8, ccmF, rpl2, rpl5, rps1, rps3, rps4), and seven from Chaetosphaeridium (atp4, atp8, rpl2, rpl5, rps1, rps3, rps4) aligned poorly to the other ASD sequences. Because the PREP-Mt program relies on accurate alignments to determine RNA editing sites, these divergent proteins were not included. For the same reason, non-homologous 5' and 3' extensions present in some mitochondrial proteins (e.g., atp6 and rps2) were trimmed from the alignments. After removal of the divergent sequences, 371 sequences remained in the final alignments, spanning all 42 known protein-coding genes variably present in land plant mitochondria [33]. There were 8.8 sequences in each alignment on average, with the actual numbers ranging from 22 sequences for nad3 to only a single Marchantia sequence for rps8.
Algorithm for RNA editing site prediction
Given a protein-coding DNA sequence and its gene identity, PREP-Mt predicts sites of C to U RNA editing [31]. The input sequence is translated using the standard genetic code and then aligned to the homologous ASD alignment with ClustalW using default parameters and the quicktree option. Next, for each column in the protein alignment, the corresponding codon from the input DNA sequence is examined to determine whether editing is possible. If the codon contains one or more cytidines, then the set of all possible unedited and edited states for that codon is determined. For example, if the input DNA sequence contained the codon "CCG", then the set of possible states in the RNA transcript would be "CCG" (not edited), "UCG" (edited at first position), "CUG" (edited at second position), and "UUG" (edited at first and second position). The amino acid, i, encoded by each of the possible codon states is compared to the amino acid, j, from all N database sequences. The score for each state, Si, is then defined by the equation
where the match parameter, Mij, is determined by
Thus, the score for each possible state is a value that ranges from 0 to 1 and is simply the percentage of matches to the amino acids in the ASD sequences for that column. The state with the highest score is reported as the predicted state. In case of a tie, the state that requires the fewest number of edits is chosen as the predicted state, since the vast majority of cytidines in plant mitochondrial genes are not actually edited. Based on this rule, silent editing sites are always disfavored because, by definition, they do not affect the encoded amino acid and would therefore always tie a state that had fewer editing sites. If a tie occurs between states that require an equal number of edits, the state that is edited at the second codon position is chosen as the predicted state, since approximately 50% of all editing sites in plants occur at the second position [8,24]. An example of the scoring scheme is presented in Figure 1.
As an additional requirement, a cutoff value, C, can be enforced. If a cutoff value is specified, Si for an edited state must be greater than or equal to C in order to be reported as the predicted state. Thus, PREP-Mt would predict an unedited state for a particular codon if Si for the edited state is less than C, even if the edited state has a higher Si than the unedited state. C must be a value ranging from 0 to 1.
PREP-Mt performance analyses
To evaluate the predictive performance of PREP-Mt, each sequence in the ASD was used as a test case. First, the protein sequence of the test case was removed from the ASD so that it would not be tested against itself. Then, the full-length protein-coding region for the test case was collected from its Genbank file and this unedited DNA sequence was used as input into PREP-Mt. Sites of RNA editing predicted by PREP-Mt were scored as either correct (TP, true positive) or incorrect (FP, false positive) based on comparison to the known edited sites. Similarly, sites predicted to remain unedited were scored as either correct (TN, true negative) or incorrect (FN, false negative) after comparison to the known unedited sites. This process was repeated for each sequence in the ASD, except for the single rps8 sequence which could not be tested because there were no other rps8 sequences to test against. Using the above classification, several statistical measures of predictive performance were calculated:
Because the number of known edited sites was proportionately much lower than the number of known unedited sites, the accuracy value was highly dependent on the specificity value. To determine the expected accuracy if the number of known edited and unedited sites were in equal proportion, a balanced accuracy statistic was also calculated according to the formula
Finally, to evaluate the effect of the cutoff value on the predictive performance of PREP-Mt, the performance analysis described above was rerun with C values ranging from 0.1 to 1.0 at all increments of 0.1.
Results
Classification of known editing sites
There were 60,263 cytidines present in the 370 tested protein-coding sequences (the single rps8 gene could not be tested because there were no rps8 homologues to test against). For sequences from the complete mitochondrial genomes of Marchantia, Chara, and Chaetosphaeridium, RNA editing capability was assumed to be absent, as indicated by predictive and experimental analyses [4-6,27-30]. For the remaining sequences, sites of RNA editing had been determined experimentally and were reported in the Genbank files and/or the literature. Based on these sources, 3,698 (6.1%) of the cytidines were classified as known sites of RNA editing, while the remaining 56,565 (93.9%) were classified as known unedited cytidines.
PREP-Mt performance analyses
PREP-Mt was used to predict editing sites for all 370 sequences in the test set. PREP-Mt's predictive performance was measured by comparing the predicted state of each cytidine to its known state (Tables 1 and 2). Of the 3,698 known edited sites, PREP-Mt correctly identified 3,038 (TP) as edited sites and incorrectly predicted 660 (FN) as unedited sites (sensitivity = 82.2%). PREP-Mt also correctly identified 55,956 (TN) out of the 56,565 known unedited sites, while incorrectly predicting 609 (FP) to be edited (specificity = 98.9%). Altogether, PREP-Mt correctly classified 58,994 of the 60,263 cytidines in the test sequences as either an edited or unedited position (accuracy = 97.9%, balanced accuracy = 90.5%). Of the 660 false negatives, 489 occurred at first or third codon positions that did not change the amino acid encoded by the codon. Excluding these silent editing positions, which have no effect on protein sequence or function, sensitivity increased to 94.7%, accuracy increased to 98.7%, and balanced accuracy increased to 96.8%. Specificity was unaffected by the silent site adjustment. The speed of prediction was also very fast. For each of the 370 tested sequences, editing site prediction took less than one second on a 3.2 GHz Pentium IV computer running RedHat Linux 9 with 1 GB of RAM (data not shown).
To evaluate the predictive performance of PREP-Mt in more detail, the performance results were subdivided by gene (Table 1) and by genus (Table 2). After doing so, it was apparent that specificity remained very high for all treatments, never falling below 95% for any gene or genus. Accuracy was also consistently high, with only a single case lower than 90%. In contrast, sensitivity was dependent on the gene or genus analyzed. In particular, several genes (sdh3, sdh4, atp8, rpl2, rps1, rps3, and rps19) and genera (Gymnocladus, Nicotiana, Oxalis, Podophyllum, and Secale) exhibited low sensitivity scores. In some cases, the poor sensitivity scores were due to the fact that a large portion of the known edited sites were at silent positions and therefore could not be predicted by PREP-Mt. Calculating sensitivity after excluding the silent edited sites helped to alleviate many of these low scores, and in general greatly increased sensitivity scores for most genes and genera. In addition, low sensitivity may be the result of the small sample sizes found for most of the poorly performing examples, if by chance the small set of genes or genera sampled does not conform to the normal editing patterns of plant mitochondrial genes. In this regard, it is interesting to note that the editing data for Gymnocladus, Oxalis, and Podophyllum came from sdh3 and sdh4. It is possible that the low sensitivities seen for Gymnocladus, Oxalis, and Podophyllum were the result of sampling from these poorly performing genes, or, conversely, that the low scores for sdh3 and sdh4 were due to the inclusion of these poorly performing genera.
PREP-Mt predictive performance was also affected by the inclusion of sequences with unusually poor predictive results. 172 of the 609 false positives were found in genes from the three organisms (Marchantia, Chara, and Chaetosphaeridium) assumed to lack the ability to edit their transcripts (Table 2). The numbers of false positives across these three genomes were similar to those from the four complete genomes (Arabidopsis, Beta, Brassica, and Oryza) that do edit their transcripts, which is most likely reflective of an underlying rate of false positive prediction by PREP-Mt. However, if the assumption that Marchantia, Chara, and Chaetosphaeridium lack RNA editing is not correct, then some of these false positives sites may be true sites of editing. Further experimental analysis is needed to verify that RNA editing is absent in these three species. Another large portion of the false positives came from individual sequences that demonstrated poor predictive results (Table 3). Of the 58 predicted editing sites from 12 examples, 54 were false positives. On average, the false positives received moderately to very high predictive scores, making it unlikely that these examples are a collection of sequences that have a large number of spurious predictions by chance. These examples could simply be the result of incomplete experimental analysis, annotation errors, or pseudogenization. More interestingly, they may represent a small class of functional mitochondrial sequences that exhibit unusual editing properties for as yet undetermined reasons.
Effect of the cutoff score on editing site prediction
The predictive analyses presented above did not impose a cutoff value (i.e., C = 0). To determine the effect of imposing a minimum cutoff value on editing site prediction, the predictive analyses were reevaluated using C values ranging from 0.1 to 1.0 (Fig. 2). As expected, increasing the cutoff value led to a decrease in the number of false positives and an increase in the number of false negatives. Up to C = 0.6, this was a balanced trade-off, because the number of false positives that decreased was approximately equal to the number of increased false negatives. As C was increased from 0.6 to 1.0, however, false negatives accumulated much faster than false positives diminished. Consequently, prediction accuracy remained roughly constant as the C value was increased to 0.6, but then sharply fell upon further increases of C. Because specificity is inversely related to the number of false positives (and is not affected by the number of false negatives), the specificity and false positive curves are also inversely related. Likewise, the sensitivity and false negative curves are inversely proportional to one another.
Discussion
PREP-Mt performance analysis
The PREP-Mt program is very accurate, correctly classifying 98% of all cytidines as either an edited or unedited site from 370 sequences spanning 41 functionally distinct genes and 44 diverse genera. PREP-Mt is also extremely specific and highly sensitive (at least for non-silent editing sites), properly identifying 99% of the known unedited sites and 95% of the known edited sites that change the amino acid encoded by a codon. Because these results are consistent for almost all genes and genera, the PREP-Mt methodology appears to be generally applicable for plant mitochondrial protein-coding genes. Furthermore, predictions are made exceptionally fast, which makes this program appropriate as an online resource. The speed of PREP-Mt is primarily due to the facts that the ASD homologues are prealigned, the number of sequences in each alignment is low, and mitochondrial genes are less than 1,000 nucleotides in length on average.
The high specificity of PREP-Mt is not surprising, because the methodology takes advantage of the extraordinary conservation (with rare exception) of plant mitochondrial sequences, which have the lowest known substitution rates for any organism [[34,35]; but see ref. [36]], and the limited range of amino acids possibly produced after RNA editing of a specific codon, which is usually only one out of 20. Thus, it is very unlikely that a column in a protein alignment will lead to a spurious prediction by chance. However, in poorly conserved regions of a protein alignment, one or two sequences in the ASD could by chance have the specific amino acid encoded by the edited state of an input sequence codon. As an example, initial analyses of PREP-Mt performance did not trim the non-homologous 5' and 3' extensions found in several mitochondrial proteins. Because of this, numerous sites of RNA editing were incorrectly inferred in these regions. Interestingly, the actual number of true editing sites in these same regions is extremely low, with less than one site observed on average in extensions that can span up to 1,000 nucleotides in length. Therefore, these non-homologous extensions can be removed from the ASD with almost no negative consequences.
Accuracy is also consistently high, which is due to the strong influence of specificity on the accuracy score. Accuracy is a measure of overall performance, combining the individual performances of predicting the known unedited sites (measured by specificity) and the known edited sites (measured by sensitivity). However, since the number of known unedited sites vastly outnumbers the number of known edited sites regardless of gene or genus examined, PREP-Mt's ability to identify the known unedited sites is more heavily weighted in the accuracy score. To get an unbiased measure of overall performance, sensitivity and specificity can be averaged. This balanced accuracy value represents the (biologically unrealistic) scenario in which the numbers of known unedited and edited sites are equal. Since specificity is relatively constant across all treatments, the balanced accuracy statistic correlates most strongly with the fluctuations in sensitivity.
PREP-Mt is also very sensitive in identifying the known editing sites that change the encoded amino acid, finding 95% of these non-silent editing sites. Some of the missed non-silent editing sites are likely lineage-specific gains of RNA editing that do not get predicted because the lineage is underrepresented in the ASD. Others are probably due to errors in the data that defines the known edited and unedited sites (see below). Analysis of sensitivity reveals that the main limitation of the PREP-Mt methodology is its inability to identify silent editing sites. Currently, the tie-breaking rules used by PREP-Mt always select the codon state that requires the fewest number of edits. Since any codon with the potential for silent editing produces the same amino acid regardless of editing status, the scores for the edited and unedited states will be identical and the tie-breaking rule will always select the unedited state. It would be possible to change this tie-breaking rule so that the state with the most number of edits is preferentially chosen, but doing so would lead to an overwhelming number of false positives at these potential silent editing sites, since the vast majority of these sites are not actually edited. Aligning DNA sequences instead of proteins would help to identify a number of these silent editing sites, but the chances of identifying false positives would be much greater here as well since there are only four different nucleotides in DNA as opposed to 20 amino acids in proteins. Determination of sequence motifs that unambiguously specify a particular cytidine as a site of editing would also help to predict these problematic sites. However, even if such motifs exist and are discovered, silent editing sites may remain difficult to predict. Silent editing sites are often found only occasionally edited in experimental studies, suggesting that the putative motifs are more weakly conserved for many silent editing sites.
Unlike the problem of silent editing site prediction, which is due to a real limitation of the methodology, PREP-Mt predictive performance is also negatively affected by errors in the data used to define the known edited and unedited sites, which causes some correctly predicted sites to be incorrectly classified as false positives or false negatives. In experimental determination of RNA editing sites, several types of errors are produced. In some cases, edited sites are not detected, either because a particular site is only occasionally edited or because incompletely edited transcripts are preferentially amplified during PCR. Thus, some predicted editing sites may in fact be true editing sites, but since they were not detected in the experimental analysis they are identified as false positives. In other cases, errors introduced during reverse transcription or sequencing lead to identification of C to U changes that are not actual editing sites. Inclusion of these errors in the known set of editing sites causes predicted unedited sites to be incorrectly identified as false negatives. Other noise in the known editing data comes from the fact that the Genbank annotations are subject to human error. Numerous examples of annotation mistakes were encountered while constructing the Aligned Sequence Database. Most were easily corrected because they involved a nucleotide that was not a cytidine. While three cases were known sites of the rare U to C editing in angiosperms, the vast majority were clearly mistakes. It is expected that about 25% of annotation errors would by chance happen to annotate another cytidine, and these are undetectable except by cross-referencing every known edited site with the literature. Since this was not done, a small increase in both false positives and false negatives is expected. Including sequences from Marchantia, Chara, and Chaetosphaeridium, which may actually perform RNA editing to some extent, and from Table 3, which may simply be the result of incomplete experimental analysis, could also cause predicted editing sites to be incorrectly classified as false positives.
Improving predictive performance
Because the current prediction scheme calculates the score using all ASD homologues, lineage-specific gains or losses of editing may be missed due to the skewed phylogenetic distribution of sequences in the ASD (Fig. 3). Although mitochondrial RNA editing is known to occur in almost all major land plant lineages [4-6], 74% of the proteins in the ASD were from the angiosperms, while almost all of the other proteins came from organisms that most likely do not perform RNA editing. In contrast, only five sequences were available from the gymnosperms, and the other major plant lineages were not represented at all. Partial sequence data with editing information was available for a number of genes from the underrepresented groups; however, these sequences covered less than half of the gene in almost all cases and were therefore not included in the ASD.
One way to overcome the skewed sample of ASD sequences would be to modify the scoring scheme so that it approximates phylogenetic methods. Because RNA editing sites are heritable, more sites are shared between closely related species than between distantly related ones. Thus, instead of using all of the homologous ASD sequences, a subset of sequences that are closely related to the input sequence could be specified. Assigning sequence weights to the database homologues based on their overall similarity to the input sequence would achieve comparable results. The best solution to this problem would be to use actual phylogenetic methods to identify the most parsimonious or most likely state for a particular codon in the input sequence; however, the speed of prediction would certainly suffer if this strategy was implemented. An alternative approach to overcome the skewed distribution of ASD sequences would be to expand the database so that it consists of a more balanced diversity of species. This approach would also help to reduce the negative effects of small sample sizes on predictive performance. As additional mitochondrial genome and transcriptome data become available, it is expected that PREP-Mt performance will continue to increase.
The simplistic scoring scheme may also lead to spurious predictions in poorly conserved protein regions. To address this issue, a component that measures the conservation in the database sequences for each column in the alignment could be incorporated into the score. Regions of high conservation between homologues are often functionally important for that protein. Thus, the correcting nature of RNA editing is expected to be most valuable in these regions and the PREP-Mt prediction scores could be weighted based on the level of alignment conservation.
Comparison to other prediction methods
A recent paper used sequence context and the estimated folding energy of these regions to predict editing sites in plant mitochondrial genes [24]. The authors constructed a data set containing all of the known edited sites from the complete genomes of Arabidopsis thaliana, Brassica napus, and Oryza sativa and compared them to an equal number of known unedited sites randomly selected from these genomes. Using tree-based statistical models, the combined data set was partitioned based on the variables that showed the greatest ability to discriminate between edited and unedited sites. Editing sites were predicted with 70.5% accuracy using a simple single-tree approach and 84.8% accuracy using a "random forest" method that analyzed thousands of trees. These values are directly comparable to the balanced accuracy of 90.5% for PREP-Mt, since the balanced accuracy statistic is the accuracy expected for PREP-Mt if an equal number of known edited and unedited sites were used. The speed of the tree-based statistical models were not presented, but they are very unlikely (especially the random forest model) to be competitive with the nearly instantaneous predictive results from PREP-Mt.
It should be noted that Cummings and Myers [24] also calculated the sensitivity and specificity of their approach; however, they classified false positives and false negatives differently than what is reported for PREP-Mt. Cummings and Myers classified false negatives as the unedited sites that were incorrectly partitioned with the true positives (the correctly classified edited sites) and false positives as the edited sites that were incorrectly partitioned with the true negatives (the correctly classified unedited sites). This has no effect on the calculation of accuracy, but it has a major effect on sensitivity and specificity scores. For their single-tree approach, Cummings and Myers provided the raw numbers of correctly and incorrectly classified edited and unedited positions, making it possible to calculate specificity and sensitivity in the same way as for PREP-Mt. This method correctly classified 1,262 out of 1,347 known editing sites (sensitivity = 93.7%), but only 637 out of 1,347 known unedited sites (specificity = 47.3%). Thus, on the one hand, their approach is very good at identifying RNA edited sites. However, approximately half of all known unedited sites are also incorrectly classified as edited sites. Because their model uses the unrealistic scenario in which the numbers of edited sites and unedited sites are equal, this problem would be increased significantly when applied to biologically realistic situations. For the 1,347 known edited sites that the authors examined from Arabidopsis, Brassica, and Oryza, there are actually about 17,900 cytidines that are not edited (TN + FP from Table 2). Using their single-tree approach, over 9,000 of these unedited cytidines would be incorrectly classified as editing sites. For the random forest model, actual numbers of correctly and incorrectly classified sites were not provided, so similar calculations cannot be determined. However, because their value that incorporates the number of incorrectly partitioned unedited sites is lowest for both approaches, it seems likely that the random forest approach will also identify a substantial number of false positives in biologically realistic situations.
A second recent paper used homologous protein alignments (as does PREP-Mt) to predict C insertion editing sites in the slime mold Physarum polycephalum [37]. Bundschuh predicted RNA editing sites for six Physarum mitochondrial genes by determining their optimal states from a modified hidden markov model (HMM) that allows for cytidine insertions. Transition parameters in the HMM were defined for each gene using the position-specific scoring matrix (PSSM) from a PSI-BLAST alignment of all homologous proteins in Genbank's non-redundant database. Compared to PREP-Mt, the HMM approach was less sensitive in finding editing sites (71% on average vs. 82% for PREP-Mt) and less accurate in determining the correct amino acid sequences (92% on average vs. 99% for PREP-Mt [data not shown]). These comparisons of performance should be taken with caution since the two methods were used to predict different types of RNA editing in different organismal lineages. To allow for a more direct comparison, both approaches could be easily modified to predict the other type of editing. However, Bundschuh's use of all homologous proteins to define the PSSM may be problematic if applied to C to U editing in plant mitochondria. A significant percentage of homologues identified for a plant mitochondrial protein would be other plant mitochondrial proteins, but almost all of these sequences in Genbank are predictions based on genomic DNA. In other words, RNA editing information is not incorporated or not known and so the predicted protein sequences are not correct. Prediction of editing will get misled by inclusion of these incorrect proteins since many editing sites are shared between plants. Additionally, inclusion of non-plant homologues to define the PSSM will lead to more variable alignments, increasing the chances of spurious prediction. It will also make the beginning and end of genes more difficult to align, as already observed for Bundschuh's method which could not analyze approximately 10% of each Physarum gene. The PREP-Mt method gets around these problems by limiting the sequences to plants, so that the alignments are highly conserved and the protein extremities are easily aligned (except for atp6 and rps2 which have non-homologous 5' and 3' extensions for many species). Furthermore, only correct protein sequences are used in the alignments, since PREP-Mt limits the plant sequences to those with known editing information. Finally, because these alignments are predefined for PREP-Mt, Bundschuh's approach would need to use predefined HMMs for each gene to be competitive in terms of speed.
Applications
The main use of PREP-Mt will be to identify RNA editing sites in the thousands of plant mitochondrial gene sequences available in the sequence databases, since editing information is only known for a small percentage of these sequences. Furthermore, as many additional plant mitochondrial genome sequencing projects are planned or already underway, PREP-Mt could serve an important role by quickly and accurately determining most sites of RNA editing without the need for sequencing of mitochondrial transcripts.
Currently, RNA editing information for full-length plant mitochondrial genes is limited mostly to the angiosperms (Fig. 3), which almost always convert in the C to U direction. Because of this, the reverse U to C editing phenomenon was not considered here. However, reverse editing has been found to be much more common in fern and hornwort mitochondria and was shown to increase protein conservation as well [4-6]. It is likely, therefore, that modification of PREP-Mt to allow for U to C prediction would be successful in the species that regularly perform this type of editing. Similarly, the PREP-Mt methodology could be applied to the problem of RNA editing in plant chloroplasts, which also perform editing that leads to an increase in protein conservation across species [7,11-15]. More generally, the PREP-Mt methodology should be effective for any system where RNA editing increases protein conservation across species.
PREP-Mt could also be used as a biological tool beyond simply identifying sites of RNA editing. For example, PREP-Mt could be used as a determinant of gene functionality. Unlike animals and fungi, whose mitochondrial gene content has remained stable for tens to hundreds of millions of years, gene content in plant mitochondrial genomes is much more variable [33]. The lineage-specific differences in plant mitochondrial gene content are mostly due to the proclivity of some genes to be relocated to the nuclear genome [38]. Upon transfer to the nucleus, the mitochondrial gene copy often degrades into a pseudogene. Because these pseudogenes are often still transcribed and edited [8,9,19-21], non-functionality is ascribed based on the presence of internal stop codons or frameshifts. This could be problematic since a mitochondrial gene, which has been functionally replaced by a nuclear gene, may still be intact and in-frame. Using PREP-Mt, pseudogenes could also be identified based on the fact that their editing positions do not always lead to the increased protein sequence conservation across species that functional genes demonstrate [19-21]. Intact and in-frame genes that demonstrate unusual editing properties, such as those listed in Table 3, may indicate loss of functionality and the presence of a functional nuclear copy, as already suggested for rps1 from Oenothera [39] and rps14 from Brassica [10]. It is interesting to note that 10 of the 12 cases in Table 3 are from genes that are very often found transferred to the nucleus in plants [38].
For atp1 and cox3, pseudogenization is not a likely hypothesis because these genes have never been found transferred to the nucleus in plants [38]. These two examples (as well as the 10 discussed above) could represent bona fide cases of functional mitochondrial transcripts that do not get edited properly for some reason. Identification of genes with abnormal editing patterns and further analysis into the causes underlying these patterns could lead to an understanding of the mechanism of RNA editing in plants, which is still largely unknown [2,3].
Conclusion
PREP-Mt is available as an online resource that predicts sites of C to U RNA editing in plant mitochondrial protein-coding genes. It was tested on a comprehensive set of genes with known RNA editing information and was shown to be highly sensitive, specific, and accurate in most cases. The speed of prediction was also extremely fast. Thus, PREP-Mt is a substantial improvement over other RNA editing prediction methods, and its predictive performance is expected to continue to improve as more editing data become available. PREP-Mt may be useful for predicting protein sequences, for determining gene functionality, and for understanding the mechanism of RNA editing. The PREP-Mt methodology could be used to predict editing sites in any system where the effect of editing is to increase protein conservation across species, such as for reverse U to C editing in plants and for plastid RNA editing.
Availability and requirements
PREP-Mt is an online tool that is freely available for use at
Supplementary Material
Additional File 1
atp1 alignment in FASTA format
Click here for file
Additional File 2
atp4 alignment in FASTA format
Click here for file
Additional File 3
atp6 alignment in FASTA format
Click here for file
Additional File 4
atp8 alignment in FASTA format
Click here for file
Additional File 5
atp9 alignment in FASTA format
Click here for file
Additional File 6
ccmB alignment in FASTA format
Click here for file
Additional File 7
ccmC alignment in FASTA format
Click here for file
Additional File 8
ccmFc alignment in FASTA format
Click here for file
Additional File 9
ccmFn alignment in FASTA format
Click here for file
Additional File 10
cob alignment in FASTA format
Click here for file
Additional File 11
cox1 alignment in FASTA format
Click here for file
Additional File 12
cox2 alignment in FASTA format
Click here for file
Additional File 13
cox3 alignment in FASTA format
Click here for file
Additional File 14
matR alignment in FASTA format
Click here for file
Additional File 15
mttB alignment in FASTA format
Click here for file
Additional File 16
nad1 alignment in FASTA format
Click here for file
Additional File 17
nad2 alignment in FASTA format
Click here for file
Additional File 18
nad3 alignment in FASTA format
Click here for file
Additional File 19
nad4 alignment in FASTA format
Click here for file
Additional File 20
nad4L alignment in FASTA format
Click here for file
Additional File 21
nad5 alignment in FASTA format
Click here for file
Additional File 22
nad6 alignment in FASTA format
Click here for file
Additional File 23
nad7 alignment in FASTA format
Click here for file
Additional File 24
nad9 alignment in FASTA format
Click here for file
Additional File 25
rpl2 alignment in FASTA format
Click here for file
Additional File 26
rpl5 alignment in FASTA format
Click here for file
Additional File 27
rpl6 alignment in FASTA format
Click here for file
Additional File 28
rpl16 alignment in FASTA format
Click here for file
Additional File 29
rps1 alignment in FASTA format
Click here for file
Additional File 30
rps2 alignment in FASTA format
Click here for file
Additional File 31
rps3 alignment in FASTA format
Click here for file
Additional File 32
rps4 alignment in FASTA format
Click here for file
Additional File 33
rps7 alignment in FASTA format
Click here for file
Additional File 34
rps8 alignment in FASTA format
Click here for file
Additional File 35
rps10 alignment in FASTA format
Click here for file
Additional File 36
rps11 alignment in FASTA format
Click here for file
Additional File 37
rps12 alignment in FASTA format
Click here for file
Additional File 38
rps13 alignment in FASTA format
Click here for file
Additional File 39
rps14 alignment in FASTA format
Click here for file
Additional File 40
rps19 alignment in FASTA format
Click here for file
Additional File 41
sdh3 alignment in FASTA format
Click here for file
Additional File 42
sdh4 alignment in FASTA format
Click here for file
Acknowledgements
I thank Jeffrey Palmer, Danny Rice, Aaron Richardson, and Predrag Radivojac for helpful discussions and Alejandro Araya for providing the RNA editing positions in exon 5 of nad1 for Triticum aestivum. This research was supported by National Institutes of Health Grant R01-GM-35087 (to Jeffrey D. Palmer).
Figures and Tables
Figure 1 Score calculation and prediction of RNA editing. The hypothetical alignment shows an input sequence aligned to three homologues from the Aligned Sequence Database. The predicted state is boxed. Column 1 shows a case where RNA editing is not possible. Column 2 presents a case where the editing state is more strongly supported. Column 3 demonstrates the various rules placed on the algorithm. The CUG is taken as the correct state because it requires fewer edits than UUG and is edited at the second position, whereas UCG is edited at the first position.
Figure 2 Effect of the cutoff score on PREP-Mt predictive performance. Effect of the cutoff score as measured by (A) accuracy, (B) sensitivity, (C) specificity, and (D) the numbers of false positives (square), false negatives (triangle), and non-silent false negatives which change the encoded amino acid (diamond). Accuracy and sensitivity were calculated using either all false negatives (triangle) or only the false negatives that change the encoded amino acid (diamond). Cutoff values used for testing ranged from 0 to 1.0 at all increments of 0.1.
Figure 3 Skewed phylogenetic distribution of sequences in the Aligned Sequence Database. Shown are the number of sequences in the ASD for each lineage and the percentage of the total number of sequences. Editing information is available for several genes from the underrepresented groups, but they were not included in the ASD because only portions of the genes were analyzed. The dotted lines indicate lack of RNA editing based on previous predictive and experimental analyses (see main text). The marchantiid liverworts are thought to have secondarily lost the ability to perform RNA editing. The relationship between plant lineages was taken from Knoop [33].
Table 1 PREP-Mt performance results subdivided by gene
Gene No. of Species TP TN FP FN Specificity Sensitivity Accuracy Balanced Accuracy
Complex I
nad1 10 135 1632 14 33 (2) 99.1 80.4 (98.5) 97.4 (99.1) 89.8 (98.8)
nad2 9 140 2337 22 42 (9) 99.1 76.9 (94.0) 97.5 (98.8) 88.0 (96.5)
nad3 22 274 1217 8 47 (4) 99.3 85.4 (98.6) 96.4 (99.2) 92.4 (99.0)
nad4 9 133 2358 19 12 (1) 99.2 91.7 (99.3) 98.8 (99.2) 95.5 (99.2)
nad4L 8 49 303 5 0 (0) 98.4 100.0 (100.0) 98.6 (98.6) 99.2 (99.2)
nad5 10 122 3647 29 10 (4) 99.2 92.4 (96.8) 99.0 (99.1) 95.8 (98.0)
nad6 10 70 995 16 15 (3) 98.4 82.4 (95.9) 97.2 (98.2) 90.4 (97.2)
nad7 8 117 1655 5 22 (1) 99.7 84.2 (99.2) 98.5 (99.7) 91.9 (99.4)
nad9 12 75 1236 7 5 (3) 99.4 93.8 (96.2) 99.1 (99.2) 96.6 (97.8)
Complex II
sdh3 7 6 391 12 11 (2) 97.0 35.3 (75.0) 94.5 (96.6) 66.2 (86.0)
sdh4 8 10 361 14 16 (3) 96.3 38.5 (76.9) 92.5 (95.6) 67.4 (86.6)
Complex III
cob 13 117 2834 19 25 (8) 99.3 82.4 (93.6) 98.5 (99.1) 90.9 (96.5)
Complex IV
cox1 9 38 2936 7 7 (5) 99.8 84.4 (88.4) 99.5 (99.6) 92.1 (94.1)
cox2 19 193 2644 22 33 (11) 99.2 85.4 (94.6) 98.1 (98.9) 92.3 (96.9)
cox3 12 80 1793 21 11 (2) 98.8 87.9 (97.6) 98.3 (98.8) 93.4 (98.2)
Complex V
atp1 10 21 2926 10 8 (1) 99.7 72.4 (95.5) 99.4 (99.6) 86.0 (97.6)
atp4 7 51 730 21 18 (8) 97.2 73.9 (86.4) 95.2 (96.4) 85.6 (91.8)
atp6 12 88 1893 11 10 (2) 99.4 89.8 (97.8) 99.0 (99.3) 94.6 (98.6)
atp8 8 16 757 10 10 (2) 98.7 61.5 (88.9) 97.5 (98.5) 80.1 (93.8)
atp9 19 91 792 1 17 (0) 99.9 84.3 (100.0) 98.0 (99.9) 92.1 (99.9)
Cytochrome c biogenesis
ccmB 9 216 1013 14 63 (6) 98.6 77.4 (97.3) 94.1 (98.4) 88.0 (98.0)
ccmC 9 187 1212 24 26 (1) 98.1 87.8 (99.5) 96.5 (98.2) 92.9 (98.8)
ccmFc 6 84 1631 23 30 (10) 98.6 73.7 (89.4) 97.0 (98.1) 86.1 (94.0)
ccmFn 6 155 2249 39 38 (21) 98.3 80.3 (88.1) 96.9 (97.6) 89.3 (93.2)
Ribosomal proteins
rpl2 4 2 1057 14 6 (4) 98.7 25.0 (33.3) 98.1 (98.3) 61.8 (66.0)
rpl5 9 45 919 14 5 (3) 98.5 90.0 98.1 (98.3) 94.2 (96.1)
rpl6 3 0 138 3 0 (0) 97.9 - 97.9 (97.9) -
rpl16 9 33 692 6 13 (2) 99.1 71.7 (94.3) 97.4 (98.9) 85.4 (96.7)
rps1 3 2 295 4 5 (5) 98.7 28.6 (28.6) 97.1 (97.1) 63.6 (63.6)
rps2 7 25 1489 15 7 (5) 99.0 78.1 (83.3) 98.6 (98.7) 88.6 (91.2)
rps3 7 54 2183 19 33 (11) 99.1 62.1 (83.1) 97.7 (98.7) 80.6 (91.1)
rps4 6 83 1116 22 18 (6) 98.1 82.2 (93.3) 96.8 (97.7) 90.1 (95.7)
rps7 8 6 664 19 2 (1) 97.2 75.0 (85.7) 97.0 (97.1) 86.1 (91.5)
rps10 5 8 266 9 0 (0) 96.7 100.0 (100.0) 96.8 (96.8) 98.4 (98.4)
rps11 3 0 222 1 0 (0) 99.6 - 99.6 (99.6) -
rps12 17 87 1142 13 5 (1) 98.9 94.6 (98.9) 98.6 (98.9) 96.7 (98.9)
rps13 10 34 554 8 5 (2) 98.6 87.2 (94.4) 97.8 (98.3) 92.9 (96.5)
rps14 5 2 239 7 0 (0) 97.2 100.0 (100.0) (97.2) (97.2) 98.6 (98.6)
rps19 6 16 242 11 9 (3) 95.7 64.0 (84.2) 92.8 (94.9) 79.8 (89.9)
Other
matR 8 73 4109 30 20 (7) 99.3 78.5 (91.3) 98.8 (99.1) 88.9 (95.3)
mttB 8 100 1087 41 23 (12) 96.4 81.3 (89.3) 94.9 (95.7) 88.8 (92.8)
Overall 370 3,038 55,956 609 660 (171) 98.9 82.2 (94.7) 97.9 (98.7) 90.5 (96.8)
Values after exclusion of silent editing sites are shown in parentheses. In some cases (marked with a '-'), sensitivity and balanced accuracy could not be calculated because there were no known edited sites (i.e., TP + FN = 0).
Table 2 PREP-Mt performance results subdivided by genus
Genus No. of Genes TP TN FP FN Specificity Sensitivity Accuracy Balanced Accuracy
Angiosperms
Eudicots
Amphicarpaea 1 13 143 0 0 (0) 100.0 100.0 (100.0) 100.0 (100.0) 100.0 (100.0)
Arabidopsis 31 344 5880 58 89 (24) 99.0 79.4 (93.5) 97.7 (98.7) 89.2 (96.3)
Beta 30 299 5406 52 50 (11) 99.0 85.7 (96.5) 98.2 (98.9) 92.4 (97.7)
Brassica 32 362 5874 53 55 (15) 99.1 86.8 (96.0) 98.3 (98.9) 93.0 (97.6)
Cologania 1 13 138 1 1 (0) 99.3 92.9 (100.0) 98.7 (99.3) 96.1 (99.6)
Cucumis 2 19 332 7 0 (0) 97.9 100.0 (100.0) 98.0 (98.0) 99.0 (99.0)
Daucus 5 42 820 4 9 (4) 99.5 82.4 (91.3) 98.5 (99.1) 90.9 (95.4)
Dumasia 1 14 141 0 1 (0) 100.0 93.3 (100.0) 99.4 (100.0) 96.7 (100.0)
Euphorbia 1 1 53 0 0 (0) 100.0 100.0 (100.0) 100.0 (100.0) 100.0 (100.0)
Glycine 2 14 566 2 2 (0) 99.6 87.5 (100.0) 99.3 (99.7) 93.6 (99.8)
Gymnocladus 2 1 108 4 2 (1) 96.4 33.3 (50.0) 94.8 (95.6) 64.9 (73.2)
Helianthus 4 39 362 1 3 (1) 99.7 92.9 (97.5) 99.0 (99.5) 96.3 (98.6)
Lactuca 1 16 282 1 0 (0) 99.6 100.0 (100.0) 99.7 (99.7) 99.8 (99.8)
Lespedeza 1 13 142 0 0 (0) 100.0 100.0 (100.0) 100.0 (100.0) 100.0 (100.0)
Lupinus 6 68 484 3 3 (0) 99.4 95.8 (100.0) 98.9 (99.5) 97.6 (99.7)
Lycopersicon 4 52 550 1 21 (3) 99.8 71.2 (94.5) 96.5 (99.3) 85.5 (97.2)
Malus 1 8 47 0 0 (0) 100.0 100.0 (100.0) 100.0 (100.0) 100.0 (100.0)
Nicotiana 2 17 258 1 12 (3) 99.6 58.6 (85.0) 95.5 (98.6) 79.1 (92.3)
Oenothera 23 283 4395 58 95 (34) 98.7 74.9 (89.3) 96.8 (98.1) 86.8 (94.0)
Olea 2 11 249 0 1 (0) 100.0 91.7 (100.0) 99.6 (100.0) 95.8 (100.0)
Oxalis 1 2 51 0 3 (1) 100.0 40.0 (66.7) 94.6 (98.1) 70.0 (83.3)
Petunia 9 84 1132 8 24 (5) 99.3 77.8 (94.4) 97.4 (98.9) 88.5 (96.8)
Pisum 4 53 487 4 8 (0) 99.2 86.9 (100.0) 97.8 (99.3) 93.0 (99.6)
Podophyllum 2 3 96 5 8 (0) 95.0 27.3 (100.0) 88.4 (95.2) 61.2 (97.5)
Raphanus 5 31 453 3 6 (1) 99.3 83.8 (96.9) 98.2 (99.2) 91.6 (98.1)
Solanum 8 58 1259 16 21 (9) 98.7 73.4 (86.6) 97.3 (98.1) 86.1 (92.7)
Vitis 1 20 224 1 1 (0) 99.6 95.2 (100.0) 99.2 (99.6) 97.4 (99.8)
Magnoliids
Magnolia 5 79 527 2 12 (3) 99.6 86.8 (96.3) 97.7 (99.2) 93.2 (98.0)
Monocots
Acorus 1 8 143 2 0 (0) 98.6 100.0 (100.0) 98.7 (98.7) 99.3 (99.3)
Allium 2 24 131 2 0 (0) 98.5 100.0 (100.0) 98.7 (98.7) 99.2 (99.2)
Coix 2 23 121 0 3 (0) 100.0 88.5 (100.0) 98.0 (100.0) 94.2 (100.0)
Elymus 1 12 157 2 0 (0) 98.7 100.0 (100.0) 98.8 (98.8) 99.4 (99.4)
Hordeum 2 12 388 1 3 (1) 99.7 80.0 (92.3) 99.0 (99.5) 89.9 (96.0)
Oryza 34 395 5919 71 90 (28) 98.8 81.4 (93.4) 97.5 (98.5) 90.1 (96.1)
Ruscus 1 16 146 1 3 (2) 99.3 84.2 (88.9) 97.6 (98.2) 91.8 (94.1)
Secale 2 6 353 5 3 (0) 98.6 66.7 (100.0) 97.8 (98.6) 82.6 (99.3)
Sorghum 5 52 479 4 12 (1) 99.2 81.3 (98.1) 97.1 (99.1) 90.2 (98.6)
Triticum 26 360 4748 36 79 (13) 99.2 82.0 (96.5) 97.8 (99.0) 90.6 (97.9)
Zea 11 104 1933 16 23 (7) 99.2 81.9 (93.7) 98.1 (98.9) 90.5 (96.4)
Gymnosperms
Cycas 3 35 444 7 14 (4) 98.4 71.4 (89.7) 95.8 (97.8) 84.9 (94.1)
Pinus 2 32 123 5 3 (0) 96.1 91.4 (100.0) 95.1 (96.9) 93.8 (98.0)
Other
Chaetosphaeridium 29 0 3011 49 0 (0) 98.4 - 98.4 (98.4) -
Chara 30 0 3549 72 0 (0) 98.0 - 98.0 (98.0) -
Marchantia 32 0 3852 51 0 (0) 98.7 - 98.7 (98.7) -
Overall 370 3,038 55,956 609 660 (171) 98.9 82.2 (94.7) 97.9 (98.7) 90.5 (96.8)
Notes are the same as for Table 1.
Table 3 Examples of unusually poor predictive results
Gene Species TP FP FN Score
atp1 Secale cereale 0 5 1 (0) 0.96
cox3 Oryza sativa 1 13 0 (0) 0.93
rpl2 Arabidopsis thaliana 0 3 1 (0) 0.78
rpl2 Brassica napus 1 3 1 (0) 0.78
rpl2 Oenothera berteriana 1 5 3 (3) 0.67
rpl2 Oryza sativa 0 3 1 (1) 0.67
rps1 Oenothera berteriana 0 3 0 (0) 1.00
rps7 Arabidopsis thaliana 0 2 0 (0) 0.93
rps12 Oryza sativa 0 6 0 (0) 0.95
rps14 Brassica napus 0 3 0 (0) 0.67
sdh4 Gymnocladus dioica 0 4 0 (0) 0.68
sdh4 Podophyllum peltatum 1 4 4 (0) 0.68
Overall 4 54 11 (4) 0.83
Predicted score is averaged across all incorrectly predicted edited sites (FP). The numbers of FN after exclusion of silent editing sites are shown in parentheses.
==== Refs
Benne R Van den Burg J Brakenhoff JP Sloof P Van Boom JH Tromp MC Major transcript of the frameshifted coxII gene from trypanosome mitochondria contains four nucleotides that are not encoded in the DNA Cell 1986 46 819 826 3019552 10.1016/0092-8674(86)90063-2
Brennicke A Marchfelder A Binder S RNA editing FEMS Microbiol Rev 1999 23 297 316 10371035 10.1016/S0168-6445(99)00009-1
Mulligan RM Daniel H, Chase CD RNA editing in plant organelles Molecular Biology and Biotechnology of Plant Organelles 2004 Dordrecht: Kluwer 239 260
Hiesel R Combettes B Brennicke A Evidence for RNA editing in mitochondria of all major groups of land plants except the Bryophyta Proc Natl Acad Sci USA 1994 91 629 633 8290575
Malek O Lättig K Hiesel R Brennicke A Knoop V RNA editing in bryophytes and a molecular phylogeny of land plants EMBO J 1996 15 1403 1411 8635473
Steinhauser S Beckert S Capesius I Malek O Knoop V Plant Mitochondrial RNA Editing J Mol Evol 1999 48 303 312 10093219
Freyer R Kiefer-Meyer MC Kössel H Occurrence of plastid RNA editing in all major lineages of land plants Proc Natl Acad Sci USA 1997 94 6285 6290 9177209 10.1073/pnas.94.12.6285
Giegé P Brennicke A RNA editing in Arabidopsis mitochondria effects 441 C to U changes in ORFs Proc Natl Acad Sci USA 1999 96 15324 15329 10611383 10.1073/pnas.96.26.15324
Notsu Y Masood S Nishikawa T Kubo N Akiduki G Nakazono M Hirai A Kadowaki K The complete sequence of the rice (Oryza sativa. L.) mitochondrial genome: frequent DNA sequence acquisition and loss during the evolution of flowering plants Mol Genet Genomics 2002 268 434 445 12471441 10.1007/s00438-002-0767-1
Handa H The complete nucleotide sequence and RNA editing content of the mitochondrial genome of rapeseed (Brassica napus L.): comparative analysis of the mitochondrial genomes of rapeseed and Arabidopsis thaliana Nucleic Acids Res 2003 31 5907 5916 14530439 10.1093/nar/gkg795
Maier RM Neckermann K Igloi GL Kössel H Complete sequence of the maize chloroplast genome: gene content, hotspots of divergence and fine tuning of genetic information by transcript editing J Mol Biol 1995 251 614 628 7666415 10.1006/jmbi.1995.0460
Hirose T Kusumegi T Tsudzuki T Sugiura M RNA editing sites in tobacco chloroplast transcripts: editing as a possible regulator of chloroplast RNA polymerase activity Mol Gen Genet 1999 262 462 467 10589833 10.1007/s004380051106
Schmitz-Linneweber C Regel R Du TG Hupfer H Herrmann RG Maier RM The plastid chromosome of Atropa belladonna and its comparison with that of Nicotiana tabacum : the role of RNA editing in generating divergence in the process of plant speciation Mol Biol Evol 2002 19 1602 1612 12200487
Kugita M Yamamoto Y Fujikawa T Matsumoto T Yoshinaga K RNA editing in hornwort chloroplasts makes more than half the genes functional Nucleic Acids Res 2003 31 2417 2423 12711687 10.1093/nar/gkg327
Wolf PG Rowe CA Hasebe M High levels of RNA editing in a vascular plant chloroplast genome: analysis of transcripts from the fern Adiantum capillus-veneris Gene 2004 339 89 97 15363849 10.1016/j.gene.2004.06.018
Covello PS Gray MW RNA editing in plant mitochondria Nature 1989 341 662 666 2552326 10.1038/341662a0
Gualberto JM Lamattina L Bonnard G Weil JH Grienenberger JM RNA editing in wheat mitochondria results in the conservation of protein sequences Nature 1989 341 660 662 2552325 10.1038/341660a0
Hiesel R Wissinger B Schuster W Brennicke A RNA editing in plant mitochondria Science 1989 246 1632 1634 2480644
Schuster W Brennicke A RNA editing makes mistakes in plant mitochondria: editing loses sense in transcripts of a rps19 pseudogene and in creating stop codons in coxI and rps3 mRNAs of Oenothera Nucleic Acids Res 1991 19 6923 6928 1762921
Quiñones V Zanlungo S Moenne A Gómez I Holuigue L Litvak S Jordana X The rpl5-rps14-cob gene arrangement in Solanum tuberosum: rps14 is a transcribed and unedited pseudogene Plant Mol Biol 1996 31 937 943 8806426 10.1007/BF00019483
Grohmann L Brennicke A Schuster W The mitochondrial gene encoding ribosomal protein S12 has been translocated to the nuclear genome in Oenothera Nucleic Acids Res 1992 20 5641 5646 1454526
Choury D Farré JC Jordana X Araya A Different patterns in the recognition of editing sites in plant mitochondria Nucleic Acids Res 2004 32 6397 6406 15585663 10.1093/nar/gkh969
Takenaka M Neuwirt J Brennicke A Complex cis-elements determine an RNA editing site in pea mitochondria Nucleic Acids Res 2004 32 4137 4144 15295040 10.1093/nar/gkh763
Cummings MP Myers DS Simple statistical models predict C-to-U edited sites in plant mitochondrial RNA BMC Bioinformatics 2004 5 132 15373947 10.1186/1471-2105-5-132
Gray MW Covello PS RNA editing in plant mitochondria and chloroplasts FASEB J 1993 7 64 71 8422976
Glaubitz JC Carlson JE RNA editing in the mitochondria of a conifer Curr Genet 1992 22 163 165 1384996 10.1007/BF00351477
Oda K Yamato K Ohta E Nakamura Y Takemura M Nozato N Akashi K Kanegae T Ogura Y Kohchi T Ohyama K Gene organization deduced from the complete sequence of liverwort Marchantia polymorpha mitochondrial DNA: A primitive form of plant mitochondrial genome J Mol Biol 1992 223 1 7 1731062 10.1016/0022-2836(92)90708-R
Turmel M Otis C Lemieux C The chloroplast and mitochondrial genome sequences of the charophyte Chaetosphaeridium globosum : insights into the timing of the events that restructured organelle DNAs within the green algal lineage that led to land plants Proc Natl Acad Sci USA 2002 99 11275 11280 12161560 10.1073/pnas.162203299
Turmel M Otis C Lemieux C The mitochondrial genome of Chara vulgaris : insights into the mitochondrial DNA architecture of the last common ancestor of green algae and land plants Plant Cell 2003 15 1888 1903 12897260 10.1105/tpc.013169
Ohyama K Oda K Ohta E Takemura M Brennicke A, Kück U Gene organization and evolution of introns of a liverwort, Marchantia polymorpha, mitochondrial genome Plant Mitochondria 1993 Weinheim: Verlag Chemie 115 129
predictive RNA editor for plant mitochondrial genes
Thompson JD Higgins DG Gibson TJ CLUSTAL W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, positions-specific gap penalties and weight matrix choice Nucleic Acids Res 1994 22 4673 4680 7984417
Knoop V The mitochondrial DNA of land plants: peculiarities in phylogenetic perspective Curr Genet 2004 46 123 139 15300404 10.1007/s00294-004-0522-8
Wolfe KH Li WH Sharp PM Rates of nucleotide substitution vary greatly among plant mitochondrial, chloroplast, and nuclear DNAs Proc Natl Acad Sci USA 1987 84 9054 9058 3480529
Palmer JD Herbon LA Plant mitochondrial DNA evolves rapidly in structure, but slowly in sequence J Mol Evol 1988 28 87 97 3148746
Cho Y Mower JP Qiu YL Palmer JD Mitochondrial substitution rates are extraordinarily elevated and variable in a genus of flowering plants Proc Natl Acad Sci USA 2004 101 17741 17746 15598738 10.1073/pnas.0408302101
Bundschuh R Computational prediction of RNA editing sites Bioinformatics 2004 20 3214 3220 15231535 10.1093/bioinformatics/bth387
Adams KL Qiu YL Stoutemyer M Palmer JD Punctuated evolution of mitochondrial gene content: high and variable rates of mitochondrial gene loss and transfer to the nucleus during angiosperm evolution Proc Natl Acad Sci USA 2002 99 9905 9912 12119382 10.1073/pnas.042694899
Mundel C Schuster W Loss of RNA editing of rps1 sequences in Oenothera mitochondria Curr Genet 1996 30 455 460 8929399 10.1007/s002940050156
| 15826309 | PMC1087475 | CC BY | 2021-01-04 16:02:49 | no | BMC Bioinformatics. 2005 Apr 12; 6:96 | utf-8 | BMC Bioinformatics | 2,005 | 10.1186/1471-2105-6-96 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-981582900710.1186/1471-2105-6-98SoftwareProGenExpress: Visualization of quantitative data on prokaryotic genomes Watson Michael [email protected] Institute for Animal Health, Compton laboratory, High street, Compton, Newbury, RG20 7NN, UK2005 13 4 2005 6 98 98 9 2 2005 13 4 2005 Copyright © 2005 Watson; 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 integration of genomic information with quantitative experimental data is a key component of systems biology. An increasing number of microbial genomes are being sequenced, leading to an increasing amount of data from post-genomics technologies. The genomes of prokaryotes contain many structures of interest, such as operons, pathogenicity islands and prophage sequences, whose behaviour is of interest during infection and disease. There is a need for simple and novel tools to display and analyse data from these integrated datasets, and we have developed ProGenExpress as a tool for visualising arbitrarily complex numerical data in the context of prokaryotic genomes.
Results
Here we describe ProGenExpress, an R package that allows researchers to easily and quickly visualize quantitative measurements, such as those produced by microarray experiments, in the context of the genome organization of sequenced prokaryotes. Data from microarrays, proteomics or other whole-genome technologies can be accurately displayed on the genome. ProGenExpress can also search for novel regions of interest that consist of groups of adjacent genes that show similar patterns across the experimental data set. We demonstrate ProGenExpress with microarray data from a time-course experiment involving Salmonella typhimurium.
Conclusion
ProGenExpress can be used to visualize quantitative data from complex experiments in the context of the genome of sequenced prokaryotes, and to find novel regions of interest.
==== Body
Background
The genomes of prokaryotic organisms contain many structures that may be involved in pathogenicity, including a variety of operons, pathogenicity islands and prophage sequences. Operons are sets of adjacent genes in bacteria that form a single transcriptional unit, and many, such as those coding for flagella [1] or fimbriae [2], have been implicated in pathogenicity. Pathogenicity islands are distinct regions of the genome that confer virulence upon the host, and are found in many pathogens of humans, animals and plants, and at least ten pathogenicity islands have been identified in Salmonella alone [3]. Prophage sequences represent the chromosomes of bacteriophage integrated as part of the genome of the bacterial host, and have also been implicated in pathogenicity in several species [4].
In order to study the behaviour of these elements, it is essential to integrate information about the genome structure of an organism with quantitative measurements produced by post-genomic technologies, such as those from microarray or proteomics experiments. This integrative biology approach is a key feature of systems biology. Studying the behaviour of these genomic elements, and other groups of adjacent genes, during infection and disease may reveal important information about the molecular mechanisms underlying pathogenicity.
Several microbial genome viewers have been developed which allow quantitative data to be displayed on the genome. The Microbial Genomes Viewer [5] offers a good online solution, however users must install a browser plug-in and may not be comfortable transmitting data over the internet. GenoMap [6] can be used to create plots of microarray data on microbial genomes, and is available as Tcl/TK source code. Genome2D [7] also offers good visualisation of quantitative data on microbial genomes, but is limited to the Windows operating system. Finally, GenomeViz [8] has recently been released, which offers much functionality, including visualisation of quantitative data, genome alignments and GC content. However this software is currently limited to unix-based systems. All of the above solutions are limited in two respects. Firstly, the quantitative values are represented as a colour-scale, which reduces the accuracy of the data and which may present problems in comparing one colour to the next. Secondly, the above tools can only display a single value for each gene, which precludes the visualisation of more complex data, such as a time-course experiment.
Implementation
ProGenExpress is released as a package for R. R is a freely available, open-source statistical package [9] that is widely used in the biological community. R has very powerful statistical and graphical capabilities, and many add-on packages are freely available. The bioconductor project [10,11] provides a huge number of add-on packages for R, covering a wide range of biological data analysis applications, and the implementation of ProGenExpress in R provides seamless integration with many of these packages. ProGenExpress is written in the native R language and has been fully tested on both windows and linux. R is available for windows, linux, unix and MacOS (including MacOS X).
Results and discussion
ProGenExpress has been written to allow researchers to quickly and simply visualize the behaviour of bacterial genomic regions of any size during experiments using whole genome technologies, such as microarray or proteomics experiments. For information relating to the genome organisation of prokaryotes, ProGenExpress includes functions for downloading and reading both NCBI .ptt files, which describe the location of protein coding genes in bacteria in a tabular format, and include links to the COGs database [12], and whole genome RefSeq entries [13]. For the quantitative experimental data, ProGenExpress can use the objects created by many of the packages from the bioconductor project [10,11], or data imported into R from text files, SQL databases and Excel.
There are currently 225 completed prokaryotic genomes in RefSeq [15] that ProGenExpress can read, and though the utility of ProGenExpress is demonstrated here using microarray data, any kind of numerical data that can be linked to the genes of prokaryotic organisms can be displayed using ProGenExpress. Where measures of the statistical significance of the data points for each gene are available, these can be passed to the plotting functions of ProGenExpress, with the result that those genes that are not significant will be plotted in white and those that are significant will be plotted in their normal plotting colour.
The genome is represented as two barplots, one for each strand. Each gene has a number of bars equal to the number of experimental data sets included, allowing time-course or complex strain/treatment experiments to be plotted. Distance between the bars for each gene is representative of intergenic distance. Slices of the genome can be selected either by base range, gene synonym or gene name. Both horizontal and vertical plots are possible, and bars can be coloured either by numerical value or by COGs [12] functional category.
The software is demonstrated here using microarray data from Eriksson et al [14]. This data set consists of gene expression measurements from intracellular Salmonella typhimurium at 4, 8 and 12 hours post murine macrophage infection. Gene expression values were calculated as the relative expression level of test RNA to that of RNA from bacteria grown in vitro, and the data has been centred and normalised according to Eriksson et al [14]. Data from Erikson et al is available as a spreadsheet [14]. This spreadsheet was pre-processed to contain only columns for gene synonym, gene name and relative expression level of test RNA to control RNA on a log 2 scale for each of the three time points. The spreadsheet was saved as tab-delimited text and read in to R using the read.table() function. The S typhimurium genome and plasmid sequences were read in to R using the read.ptt() function, with RefSeq files NC_003197.ptt and NC_003277.ptt respectively. The microarray data was linked to the gene location data using the linkem.avg() function. Images of the microarray data on both the entire genome and the plasmid were then generated using the plotrange() and plotrange.vertical() functions in conjunction with jpeg(), an internal R function. The results were viewed in Internet Explorer. Finally, the find.region() function was used to find regions of interest as described below.
Figure 1 shows the expression of all genes on Salmonella typhimurium LT2 plasmid pSLT, coloured by COGs functional category. The majority of the genes on this plasmid are up-regulated at all three time-points, implying a role for this plasmid during macrophage infection. Figure 2 displays a smaller region of the genome containing the fli operon, with all genes in the operon displaying similar expression profiles. Erikksson et al [14] found 919 genes to be significantly differentially expressed, and that measure of statistical significance has been incorporated into Figure 2. Significant genes are coloured normally, whereas those that are not significant are white. All but three of the 14 genes in the operon are shown to be significantly differentially expressed, suggesting that the whole operon is differentially expressed and that perhaps the measure of statistical significance used is too stringent. Finally, Figure 3 is a vertical plot of Salmonella pathogenicity island II (SPI-II), showing that most genes on this island are up-regulated at all three time-points. This island encodes a type III secretion system, and has been shown to be required for systemic infection by facilitating replication of intracellular bacteria within membrane-bound Salmonella-containing vacuoles [3].
ProGenExpress can also search for operons and other regions of interest by looking for clusters of genes that are close together and which display similar patterns in the experimental data. Using this facility, we identified over 200 potential regions of interest in Salmonella typhimurium consisting of four genes or more, including several known operons and potential unannotated operons. Figure 4 shows a region of the genome containing a group of six genes that has been found using ProGenExpress. The genes have no assigned gene name, have either an unknown or putative/predicted function, are close together on the genome and have similar expression profiles across the three time-points. We believe these genes may represent an unannotated operon.
ProGenExpress has several advantages over existing software. The package seamlessly integrates with the bioconductor project and the many packages available in R for microarray analysis, including limma, marray and affy, and is available for both Windows and Linux, amongst others. Both horizontal and vertical plots are possible, and an unlimited number of data points for each gene can be plotted, allowing for the visualization and analysis of complex time course or strain/treatment experiments. Furthermore, the bar-plots display numerical data accurately, and do not rely on a colour-scale to depict values. Finally, the ability to search integrated genomic and post-genomic data sets for clusters of genes which behave similarly represents an opportunity for the discovery of novel genomic elements involved in pathogenicity.
Conclusion
We describe ProGenExpress, an open-source R package which allows researchers to quickly and easily visualise quantitative data from arbitrarily complex experiments in the context of the genome of sequenced prokaryotes. ProGenExpress can also be used to search for genomic regions which may represent coherent functional units. We show how ProGenExpress can be used to visualise microarray data from a time-course experiment on the genome of Salmonella typhimurium, and to find unannotated genomic regions that may be involved in pathogenicity. Future plans for the software include the ability to read data from ensembl databases, and the development of visualisation tools for eukaryotic genomes. Software updates and new releases will be available from the project home page.
Availability and requirements
• Project Name: ProGenExpress
• Project Home Page:
• Operating Systems: Windows, Linux, Unix
• Programming Language: R
• Other Requirements: R version 2.0 or above
• License: GNU GPL
Authors' contributions
MW developed and tested the software in full.
List of abbreviations
COG: Cluster of Orthologous Groups
SPI-II: Salmonella pathogenicity island II
Acknowledgements
This work was funded by the core strategic grant of the Institute for Animal Health, provided by the BBSRC.
Figures and Tables
Figure 1 Gene expression measurements for Salmonella typhimurium plasmid pSLT. Gene expression measurements for the pSLT plasmid of Salmonella typhimurium. The genome is represented as two barplots, one for each strand. Each gene has three bars representing expression at 4 h, 8 h and 12 h post macrophage infection. Gene expression measurements (y-axis) are ratios of test RNA to control RNA on the log 2 scale. Bars are colour-coded according to the COGs [12] functional category. Distance between genes is relative and representative of intergenic distance. This image clearly shows that the majority of genes on the plasmid are up-regulated during macrophage infection.
Figure 2 Gene expression measurements for flagella biosynthesis. Gene expression measurements for the fli operon from Salmonella typhimurium. The genome is represented as two barplots, one for each strand. Each gene has three bars representing expression at 4 h, 8 h and 12 h post macrophage infection. Gene expression measurements (y-axis) are ratios of test RNA to control RNA on the log 2 scale. Distance between genes is relative and representative of intergenic distance. The fli operon is involved in flagella biosynthesis and this image clearly shows that the entire operon is strongly down-regulated during macrophage infection. Significantly differentially expressed genes are coloured red or green, whereas non-significant genes are coloured white. All but three of the 14 genes in the operon are significantly differentially expressed.
Figure 3 Gene expression measurements for Salmonella pathogenicity island II. Gene expression measurements for Salmonella Pathogenicity Island II (SPI-II) from Salmonella typhimurium. The genome is represented as two barplots, one for each strand. Each gene has three bars representing expression at 4 h, 8 h and 12 h post macrophage infection. Gene expression measurements (y-axis) are ratios of test RNA to control RNA on the log 2 scale. Distance between genes is relative and representative of intergenic distance. This island has been linked to pathogenicity, and encodes a type III secretion system. It is required for systemic infection and intracellular pathogenesis by facilitating replication of intracellular bacteria within membrane-bound Salmonella-containing vacuoles [3]. Here we can clearly see that the majority of genes in the island are strongly up-regulated during macrophage infection.
Figure 4 A putative operon. Gene expression measurements from a genomic region of Salmonella typhimurium. The genome is represented as two barplots, one for each strand. Each gene has three bars representing expression at 4 h, 8 h and 12 h post macrophage infection. Gene expression measurements (y-axis) are ratios of test RNA to control RNA on the log 2 scale. Distance between genes is relative and representative of intergenic distance. The image shows six genes that have been identified by ProGenExpress as potentially interesting. The genes are very close together on the genome and display similar expression patterns, and therefore could represent an as yet unannotated operon. The genes have synonyms STM4257 – STM4262 and currently have no confirmed function.
==== Refs
Robertson JMC McKenzie NH Duncan M Allen-Vercoe E Woodward MJ Flint HJ Grant G Lack of flagella disadvantages Salmonella enterica serovar Enteritidis during the early stages of infection in the rat J Med Micro 2003 52 91 99 10.1099/jmm.0.04901-0
Baumler AJ Tsolis RM Heffron F Contribution of fimbrial operons to attachment to and invasion of epithelial cell lines by Salmonella typhimurium Infect Immun 1996 64 1862 65 8613405
Parkhill J Dougan G James KD Thomson NR Pickard D Wain J Churcher C Mungall KL Bentley SD Holden MT Sebaihia M Baker S Basham D Brooks K Chillingworth T Connerton P Cronin A Davis P Davies RM Dowd L White N Farrar J Feltwell T Hamlin N Haque A Hien TT Holroyd S Jagels K Krogh A Larsen TS Leather S Moule S O'Gaora P Parry C Quail M Rutherford K Simmonds M Skelton J Stevens K Whitehead S Barrell BG Complete genome sequence of a multiple drug resistant Salmonella enterica serovar Typhi CT18 Nature 2001 413 848 52 11677608 10.1038/35101607
Banks DJ Lei B Musser JM Prophage induction and expression of prophage-encoded virulence factors in group A Streptococcus serotype M3 strain MGAS315 Infect Immun 2003 71 7079 86 14638798 10.1128/IAI.71.12.7079-7086.2003
Kerkhoven R van Enckevort FH Boekhorst J Molenaar D Siezen RJ Visualization for genomics: the Microbial Genome Viewer Bioinformatics 2004 20 1812 14 14988111 10.1093/bioinformatics/bth159
Sato N Ehira S GenoMap, a circular genome data viewer Bioinformatics 2003 19 1583 84 12912843 10.1093/bioinformatics/btg195
Baerends RJ Smits WK de Jong A Hamoen LW Kok J Kuipers OP Genome2D: a visualization tool for the rapid analysis of bacterial transcriptome data Genome Biol 2004 5 R37 15128451 10.1186/gb-2004-5-5-r37
Ghai R Hain T Chakraborty T GenomeViz: visualizing microbial genomes BMC Bioinformatics 5 198 15601465 10.1186/1471-2105-5-198
R
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
Bioconductor
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
Pruitt KD Tatusova T Maglott DR NCBI Reference Sequence (RefSeq): a curated non-redundant sequence database of genomes, transcripts and proteins Nucleic Acids Res 2005 33 D501 4 15608248 10.1093/nar/gki025
Eriksson S Lucchini S Thompson A Rhen M Hinton JCD Unravelling the biology of macrophage infection by gene expression profiling of intracellular Salmonella enterica Molecular Microbiology 2003 47 103 118 12492857 10.1046/j.1365-2958.2003.03313.x
NCBI Refseq completed microbial genomes
| 15829007 | PMC1087476 | CC BY | 2021-01-04 16:02:50 | no | BMC Bioinformatics. 2005 Apr 13; 6:98 | utf-8 | BMC Bioinformatics | 2,005 | 10.1186/1471-2105-6-98 | oa_comm |
==== Front
BMC BiolBMC Biology1741-7007BioMed Central London 1741-7007-3-111583110010.1186/1741-7007-3-11Research ArticleMicrotubule plus-ends reveal essential links between intracellular polarization and localized modulation of endocytosis during division-plane establishment in plant cells Dhonukshe Pankaj [email protected] Jaideep [email protected]ülskamp Martin [email protected] Theodorus WJ [email protected] Section of Molecular Cytology, Swammerdam Institute for Life Sciences, University of Amsterdam, Kruislaan 316, 1098 SM Amsterdam, The Netherlands2 Department of Plant Agriculture, Molecular Cell Biology Laboratory, University of Guelph, Guelph, ON, N1G 2W1, Canada3 Botanical Institute III, University of Köln, Gyrhofstrasse 15, Köln, 50931, Germany4 Centre for Molecular Biology of Plants, University of Tübingen, Auf der Morgenstelle 3, 72076 Tübingen, Germany2005 14 4 2005 3 11 11 26 1 2005 14 4 2005 Copyright © 2005 Dhonukshe et al; licensee BioMed Central Ltd.2005Dhonukshe 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 key event in plant morphogenesis is the establishment of a division plane. A plant-specific microtubular preprophase band (PPB) accurately predicts the line of cell division, whereas the phragmoplast, another plant-specific array, executes cell division by maintaining this predicted line. Although establishment of these specific arrays apparently involves intracellular repolarization events that focus cellular resources to a division site, it still remains unclear how microtubules position the cell division planes. Here we study GFP-AtEB1 decorated microtubule plus-ends to dissect events at the division plane.
Results
Early mitotic events included guided growth of endoplasmic microtubules (EMTs) towards the PPB site and their coincident localization with endocytic vesicles. Consequently, an endosomal belt lay in close proximity to the microtubular PPB at its maturation and was maintained during spindle formation. During cytokinesis, EMTs radiated from the former spindle poles in a geometrical conformation correlating with cell-plate navigation and tilt-correction. Naphthylphtalamic acid (NPA), an inhibitor of polar auxin efflux, caused abnormal PPBs and shifted division planes.
Conclusion
Our observations reveal a spatio-temporal link between microtubules and intracellular polarization essential for localized endocytosis and precise establishment of the division plane in plants. Additionally, they implicate the growth regulator, auxin, in this important cellular event.
==== Body
Background
In cell wall-encased, immobile plant cells, tight regulation of the cell division plane plays a crucial role in tissue and organ morphogenesis [1,2]. At the onset of mitosis a plant-specific cortical microtubule array, the PPB [3], emerges. Although the PPB disassembles as the cell enters mitosis, it precisely predicts where the new cell plate attaches to the parental cell walls at the end of cytokinesis [1,2]. Cell division planes are drastically affected by the absence of PPB [4], which may be achieved through experimental obliteration [5] or through genetic defects [6]. Since the discovery of the PPB four decades ago, the mechanism governing its creation and effect on delineating the future cell division plane has remained a mystery.
Though initially considered to be a cortical process, PPB formation involves two coincident cytoplasmic events too, namely the migration of the nucleus towards the cell center [7], and recruitment of numerous EMTs within motile cytoplasmic strands [8,9]. In centrifuged protonemata (consisting of a single elongated cell), nuclear displacement prior to pre-prophase induces the formation of a PPB around the new nuclear position, suggesting a role for the nucleus in PPB placement [10]. However, in plant cells under normal gravity conditions, the nucleus is positioned by microtubules, since its random migration is prevented by microtubular depolymerization [11]. These observations point towards a functional connection between the nucleus, microtubules and cell cortex for marking the PPB. In budding yeast, EMTs play a role in bringing the nucleus to the division plane by a cortical 'search and capture' mechanism [12-14]; and in fission yeast, EMTs position the nucleus by pushing against the cortex [15]. Under in vitro, simulated conditions, the microtubule seeds also probe the metallic boundaries of artificial chambers with cellular dimensions to bring nuclei to the center [16]. Together, these studies on diverse systems corroborate the role of microtubules in probing the cellular geometry to settle the nucleus at the center of the cell. In the absence of information on the polarity and dynamics of plant EMTs, their precise role in association with the nucleus and the PPB is unclear.
In plants, in addition to the PPB, the other two mitotic microtubular arrays, the spindle and the phragmoplast microtubules, are also of endoplasmic nature. During the spindle stage, the perpendicularity of the spindle to the cell division plane is often lost because of its rotation. In mammalian and yeast cells the spindle is kept at the center of the cell or at the bud site by the centrosome-originating astral microtubules [17,18]. In addition, any error in spindle orientation in these cell types is corrected by the outgoing astral microtubules, which probe the cortex to ensure that the segregated chromosomes are sufficiently distant and are not cleaved by the constricting actomyosin ring [19,20]. In acentrosomal plant cells, after PPB breakdown and disappearance of the nuclear surface-bound EMT, the factors maintaining the spindle at the cell centre still remain unresolved. Paradoxically, in plant cells possessing disoriented spindles, the cell plates are still able to anchor properly to the division sites marked perpendicularly to the cell axis. A mechanism obviously exists in plant cells to correct spindle disorientation and reinforce the line of cell division, but it remains obscure.
Here, we investigated the role of EMTs during PPB formation and their subsequent behavior following karyokinesis. Because of their uniform cell size, continuous cell division activity and absence of background fluorescence, Tobacco BY-2 cell suspensions remain a system of choice for plant cell cycle studies [21]. Therefore, we used tobacco BY-2 cells stably transformed with different microtubular markers, and performed a live cell time-lapse analysis to elucidate the role of EMTs in reinforcing the lines of the cell division planes in plants.
Results
GFP-AtEB1 labeled plant microtubule plus ends exhibit guided growth to create bundles
Microtubule plus-end labeling has been achieved using a GFP-AtEB1 fusion protein and permits observations on microtubule growth directionality and dynamics [22]. In interphase cells, GFP-AtEB1 highlighted the bidirectional movement of comet-like structures, suggesting plus end growth of cortical microtubules of opposite growth polarities (Figure 1A, B; see also additional file 1: Movie 1). In co-transformed cells, GFP-AtEB1 labeled the growing ends of YFP-MAP4 labeled microtubules (Figure 1C; see also additional file 2: Movie 2) and co-localized with YFP-CLIP170 (Figure 1D; see also additional file 3: Movie 3) on the growing microtubular plus ends, reconfirming the plus end-specific localization of GFP-AtEB1 in BY-2 cells.
Figure 1 Mechanisms for microtubule guidance and bundling. Green: GFP-AtEB1 (in A-H and N) and GFP-MAP4 (in I-M). Red: YFP-MAP4 (in C, E-G and N) and YFP-CLIP170 (in D). (A) 3-D maximum projection of interphase BY-2 cell, highlighting punctuate GFP-AtEB1 labeling in the cortex. (B) In interphase cells, GFP-AtEB1 comets at the cortex move in the same or opposite direction on the same track (arrowheads), and sometimes together (arrow) (See additional file 1: Movie 1). (C) In co-transformed cells, GFP-EB1 labels the growing ends of microtubules labeled by YFP-MAP4 (See additional file 2: Movie 2). (D) Co-localization of GFP-AtEB1 and YFP-CLIP170 on growing microtubule plus ends (See additional file 3: Movie 3). (E) GFP-AtEB1 labeled growing microtubule changing from one microtubule track (labeled with YFP-MAP4) to another. [58] Growth of two separate unbundled microtubules (arrowheads) transiently meeting (arrow) and afterwards separating without inducing catastrophe. (G) Microtubules growing in opposite directions on the same track with similar speed (arrowheads) meet (arrow) and continue growing in opposite directions without inducing catastrophe. (H) Microtubule nucleation and growth (arrowheads) on an already existing microtubule (arrow). (I) Situation where attachment of a microtubule plus end (yellow arrowhead) to an existing microtubule induces a translocation of the minus end (red arrowhead) from one microtubule to another. (J) Microtubule plus end (yellow arrowhead) growing towards an existing microtubule, followed by guided growth in a new direction, causes bending at the point of previous attachment (red arrowhead). (K) Treadmilling microtubule (red arrowhead) with its plus end (red arrow) and minus end (yellow arrowhead) moving in the same direction (yellow arrowheads and red arrows) initiates bundling by bridging two other separate preexisting microtubules. (L) Long microtubule growing (arrowhead) and interacting with an existing microtubule induces reorientation of growth and bending, resulting in bundling with a shorter growing microtubule. (M) Depolymerization of one microtubule partially associated with a bundle (yellow arrowhead) causes bending of the remaining structure (red arrowhead). (N) Microtubule minus end detachment and subsequent movement (arrowheads) induces loss of GFP-AtEB1 from its plus end (yellow arrows); it recovers GFP-AtEB1 labeling and plus end growth once its minus end acquires new support on another polymer (arrowhead). Note that the same microtubule bends (red arrow) when its minus end is fixed and the plus end (yellow arrow) hits another microtubule. Time is indicated in seconds and bars represent 5 μm in A-D, 3 μm in K-M, 2 μm in E, I, J and 1 μm in F-H, N.
General observations on microtubule guidance and bundling formed the basis of our subsequent experiments. Tracking the GFP-AtEB1 comets, it was found that >80 % of freshly polymerizing microtubules exhibited guided growth on tracks established by existing microtubules, thereby creating bundles of microtubules (Figure 1C; see also additional file 2: Movie 2). Preexisting microtubule-guided oriented bundling often involved i) two or more microtubules with apparently similar polarity moving one after the other, ii) microtubules with presumably opposite polarities moving in opposite directions to each other or iii) independent microtubules moving together as a pair with similar velocities (4.15 ± 0.41 μm/min, n = 29) (Figure 1B; see also additional file 1: Movie 1). Although mammalian EB1 often induces microtubule bundling upon overexpression, AtEB1 did not cause the observed bundling in the plant cells since it lacks the microtubule bundling domain present in the mammalian ortholog [23]. Instead, intermicrotubular bridges [24] might be responsible for the observed bundle formation. In addition, when two microtubules grew together side-by-side on the same track, it was frequently observed that one of them shrank while the other continued growing (see additional file 1: Movie 1). Further information on microtubular guidance and bundling mechanisms came from analysis of cells coexpressing GFP-AtEB1 and YFP-MAP4. A growing microtubule (green arrow, Figure 1E) could detach from an existing track and move on to another track where its growth became guided in another direction (Figure 1E). Interphase microtubules in mammalian cells show a similar guidance mechanism [25]. We also observed individual microtubule plus ends approaching each other from a similar (Figure 1F) or opposite (Figure 1G) direction and meeting without inducing catastrophe. Strikingly, microtubular nucleation was sometimes initiated on an existing microtubule (Figure 1H), an observation consistent with the plant-specific localization of gamma tubulin along the microtubule length [26] and with a study reporting microtubule nucleation from stable tubulin oligomers [27]. Microtubules also changed trajectories by reorienting their minus or plus ends when one of the ends was supported on the other polymer (Figure 1I, J). It was also observed that motile polymers exhibiting specialized treadmilling [28] initiated bundling by bridging the two preexisting and separate polymers (Figure 1K). In cases of guided growth-induced bundling, a shorter microtubule could adopt the growth direction of a preexisting longer microtubule and vice versa (Figure 1L). When one of the microtubules in a bundle retracted, it frequently caused the other to bend, implying the exertion of a pulling force (Figure 1M). We also observed that upon release of a minus end from a nucleation site, the opposite plus end depolymerized (with concomitant loss of GFP-AtEB1), whereas upon acquisition of a new support by its minus end the microtubule retained growth (and regained GFP-AtEB1) (Figure 1N). These observations indicate a provision for new nucleation on existing polymers, while suggesting that in certain cases the plus end somehow 'senses' the physical state of the opposite minus end. Together, these findings implicate intermicrotubular affinities and the capacity of the polymers to nucleate new or detached microtubules as a general mode of microtubule survival, reorientation and bundle creation.
Emergence, polarity and dynamics of EMTs at the onset of cell division
Equipped with information on the general polar behavior of microtubules in interphase cells we approached the questions of appearance, polarity and dynamics of the EMTs, specifically at the onset of cell division. During preprophase, more dynamic EMTs emerged, bridging the nucleus to the cortex and exhibiting considerable bidirectionality (Figure 2A–F; see also additional file 4: Movie 4) with outgoing (from the nucleus towards the cortex) and incoming (from the cortex towards the nucleus) EMTs. More outgoing (80%) than incoming microtubules were observed though their growth rates were similar (5.86 ± 0.82 μm/min, n = 15 – outgoing; and 5.56 ± 0.47, n = 15 – incoming). Like cortical microtubules, EMTs also exhibited bundling and guidance characteristics indicating the existence of intermicrotubular affinities in the cytosol even in the absence of a cortical support. In contrast to yeast preprophase cells, where unidirectional microtubules (growing from the center towards the cell periphery) position the nucleus by pushing or pulling forces [12,15], our observations together with others [11] on the requirement of EMTs for nuclear displacement and their bidirectionality suggest that plant cells can utilize both outgoing and incoming EMTs for positioning the premitotic nucleus. Conversely, with bundling and track follow-up, the incoming microtubules might guide the outgoing ones to achieve selective cortical targeting. This may be an efficient mechanism for their navigation of intracellular space, since when many microtubules grow simultaneously in a bundle, the chance that microtubules will reach the cortical target(s) without becoming depolymerized in the process is expected to be substantially higher. Interestingly, EMTs maintained continuous contact with the cortical areas occupied by the developing PPB. At PPB maturation, the EMTs between the PPB and nuclear envelope (NE) remained bidirectional while those connecting to more distal cortical areas became unidirectional, displaying a radiating comet-like spectacular firework (Figure 2G–L; see also additional file 5: Movie 5). At this stage, kymographs generated by tracking the GFP-AtEB1 comets clearly illustrate accelerated growth for both outgoing (8.33 ± 0.83 μm/min, n = 30) and incoming (8.02 ± 1.15 μm/min, n = 12) EMTs. Microtubule growth was maintained at ca. 6.78 ± 0.89 μm/min (n = 50) at the PPB [9] (Figure 2M–O). Consequently, the microtubule density on the NE gradually increased (see Figure 2G–K), confirming earlier observations on microtubular growth and stabilization on the NE in plant [9] and mammalian [29] cells. Our observations suggest that at the onset of mitosis, the outwardly-radiating EMTs position the nucleus in the center of the cell by pushing/pulling forces, while the bidirectional EMTs connecting the NE to the PPB position the nucleus at the centre of the PPB. Moreover, during PPB maturation, the change from bidirectional growth (from the distal cortex towards the NE and from the NE towards the distal cortex) to unidirectional growth (from the NE towards the distal cortex) of EMTs severely reduces their chance of survival and thereby causes their detachment and collapse.
Figure 2 Polarity and growth speed of EMTs bridging nucleus and cortex. Green: GFP-AtEB1 (in A-L). (A-F) EMTs exhibit bidirectional growth and microtubule bundling. Note that that the microtubule originating from the nuclear surface (outgoing) and the one coming from the cortex (incoming) cross each other (arrow) and, as in the cortical array, grow with similar speeds without interfering each other (arrowheads) (see additional file 4: Movie 4). (G-L) EMT plus ends radiating mainly in an outward direction from the NE during PPB maturation (see additional file 5: Movie 5). Kymograph projection of microtubule plus ends in the interphase cortex (M), PPB cortex (N) and preprophase cytoplasm (O) showing sustained polymerization. The horizontal axis, d, represents distance (18 μm in M, 13 μm in N and 20 μm in O), and the vertical axis, t, represents time (290 s in M, 140 s in N and 390 s in O). Note that for each of the 3 cases (M-O), the microtubules follow the tracks, exhibit bi-directionality and grow with the same speeds. By comparing the slopes between images M-O, it becomes evident that the microtubule growth speed increases from interphase to the PPB stage, as previously reported [9]. Note that the arrowhead in M shows the crossing of two EMTs growing on the same path at the same time but in opposite directions. Nucleus is marked by 'N', time is indicated in seconds and bars represent 8 μm.
Role of EMTs in premitotic cytoplasmic organization
The implications of the EMT configuration for the organization of the premitotic cytoplasm were now investigated using GFP-MAP4 transformed BY-2 cells [9] together with various organelle markers. During G2-M transition, FM4-64-labeled endosomes (Figure 3A), Alexa 633-labeled pinocytic vesicles (Figure 3B), ST-YFP-labeled Golgi bodies (GA) (Figure 3C) and Mitotracker-labeled mitochondria (Figure 3D) all localized along the EMTs. In contrast to interphase, when the microtubules remain at the cortex and large vacuoles occupy the cell space (Figure 3E), the vacuoles appeared fragmented by intersecting EMTs during preprophase (Figure 3F). Previous studies have shown that the motility of cytoplasmic organelles in plants is mainly actin-based [30] and that the actin cytoskeleton co-exists with the mitotic microtubular arrays [31]. To further investigate the respective roles of microtubules and actin filaments in premitotic cytoplasmic organization, we treated the cells with latrunculin B (an actin polymerization inhibitor) and oryzalin (a microtubule depolymerizing herbicide). In latrunculin B-treated cells, the EMTs appeared stabilized and more intense with a normal cytoplasmic configuration (Figure 3G), whereas oryzalin destroyed the EMTs and caused cytoplasmic disorganization and nuclear displacement (Figure 3H). After combined oryzalin and latrunculin B treatment the nucleus completely lost its central position and the cytoplasm (stained with unbound GFP-MAP4) became completely disorganized (Figure 3I). After the oryzalin was washed out, the EMTs gradually reappeared and the cytoplasm regained its normal configuration with the nucleus replaced at the cell centre (data not shown). Together, these results suggest that EMTs have a major role in organizing the premitotic cytoplasm, but they do not discount the role of actin in mediating organelle motility.
Figure 3 EMTs configure the premitotic cytoplasm. Green: GFP-MAP4 (in A-I). Red: FM4-64 (in A, E and F), Alexa 633 (in B), ST-YFP (in C), Mitotracker (in D) FM4-64 labeled endosomes (A), Alexa 633 labeled pinocytic vesicles (B), ST-YFP labeled GA (C) and Mitotracker labeled mitochondria (D) all remain in the vicinity of GFP-MAP4 marked EMT tracks. (E) At interphase, GFP-MAP4 labeled microtubules remain in the cortex and FM4-64 labeled vacuoles occupy most of the endoplasmic space. [F] At preprophase, GFP-MAP4 labeled EMTs intersect the vacuoles labeled by FM4-64. Premitotic cells treated with latrunculin B (G), oryzalin (H) or both (I) show cytoplasmic disorganization in the presence of oryzalin (H-I). Bars represent 8 μm.
Guided growth of EMTs towards the PPB site and their coincident localization with endocytic vesicles
The implications of the observation, which differentiated between intracellular motility and intracellular compartmentalization, became apparent when we investigated the localization pattern for FM4-64 labeled endocytic vesicles in relation to the microtubules. During interphase, FM4-64 labeled endocytic vesicles were randomly localized in the cell (data not shown), but early in the G2-M transition they started coaligning with emerging EMTs (Figure 4A, B). In GFP-AtEB1 transformed cells, these endocytic vesicles displayed internalization paths along the EMT trajectories and their appearance coincided with the cortical sites approached by the EMT plus ends (Figure 4C–F; see also additional file 6: Movie 6). Oryzalin-induced microtubule depolymerization immediately affected endocytic vesicular internalization, with complete disruption of their internalization routes (Figure 4G–I; see also additional file 7: Movie 7). When the oryzalin was washed out, the reformed EMTs again approached the cortex and the internalization of the endocytic vesicle traffic resumed (Figure 4J). Most importantly, during PPB maturation the endocytic material aggregated at the cortical areas occupied by the PPB and approached by the radiating EMT plus ends (Figure 4K; see also additional file 8: Movie 8). Consequently, the endocytic vesicles formed a cortical belt loosely co-localizing with the microtubular PPB (Figure 4L, M). Support for this observation comes from a recent electron microscope study analyzing the membrane architecture of the PPB, which revealed an accumulation of clathrin-coated and non-coated pits specifically in the PPB regions [32]. Moreover, the activity of an endosomal marker protein Ara7 (Rab5 homologue from Arabidopsis), is known to be up-regulated during mitosis [33], and a similar PPB belt was observed using GFP-Ara7 labeled endosomes (Figure 4N). Furthermore, the endosomal band we observed co-localized with a band comprising Golgi bodies (Figure 4O) [34]. It is noteworthy that during PPB maturation, the EMTs, which connect the nucleus to the cortical PPB, prohibit a continuous vacuolar structure and thereby create a cytoplasmic area proximal to the PPB (Figure 4P). This cytoplasmic area occupied by the EMTs at PPB maturation is still maintained at the spindle stage (Figure 5A). It has been proposed that the actin-depleted zone (ADZ), which appears during PPB breakdown and is also maintained throughout cytokinesis, participates in regulating the division plane, since actin disruption before ADZ formation affects the cell division planes [35]. We speculate that the lack of actin prohibits further transport of continuously endocytosed material, contributing to the formation of a coherent endosomal belt, for it has been shown that plant endosomal trafficking is mostly actin-based [36].
Figure 4 Endosomal belt co-localizes with microtubular PPB during preprophase. Green: GFP-MAP4 (in A, B, L, M and P), GFP-AtEB1 (in C-K), GFP-Ara7 (in N) and ST-YFP (in O). Red: FM4-64 (in A-M and O-P). Early in the G2-M transition, FM4-64 labeled endocytic vesicles follow the emerging EMTs labeled with GFP-MAP4, as shown in single median section (A) and 3D-projection (B). The marked rectangle in (C) is zoomed in for (D-I). FM4-64 labeled endocytic vesicles preferentially internalize from the cortical areas approached by the GFP-AtEB1 labeled EMT plus ends (D-F) (see additional file 6: Movie 6), and oryzalin-induced microtubule depolymerization disrupts their internalization routes (G-I) (see additional file 7: Movie 7) whereas the internalization paths are recovered after oryzalin removal (J). (K) Close-up of GFP-AtEB1 marked EMT plus ends bridging the NE and PPB. Note that during PPB narrowing, FM4-64 labeled endocytic vesicles preferentially internalize from the cortical areas approached by the GFP-AtEB1 (see additional file 8: Movie 8). Formation of an FM4-64 labeled cortical belt at the PPB site (labeled with GFP-MAP4) is shown in a single median section (L) and in 3-D projection (M). (N) 3-D projection of GFP-Ara7 labeled endosomes exhibiting an endosomal belt at preprophase. (O) Both FM4-64 labeled endosomes and ST-YFP labeled GAs form a cortical belt at the PPB site. (P) GFP-MAP4 labeled EMTs connecting the nucleus to the PPB intersect FM4-64 labeled vacuoles. Time is indicated in minutes. Bars represent 7 μm in A, B, J, L, M, 10 μm in C, N-P and 5 μm in K.
Figure 5 PM targeted EMT plus ends probe the areas occupied by the preceding PPB and align the cell plates for proper docking at the parental walls. Green: GFP-MAP4 (in A, F-I), GFP-AtEB1 (in B-E, J-O and P). Red: FM4-64 (in A, F-O), YFP-MAP4 (in P). (A) Discontinuity of the vacuolar structures in the preceding PPB site (arrowheads) is maintained at the spindle stage, as visualized with FM4-64 labeled vacuoles and GFP-MAP4 labeled microtubules.(B-C) At the onset of the phragmoplast stage, GFP-AtEB1 labeled EMT plus ends (red arrowheads) originating from the former spindle poles grow towards the cortex (see additional file 9: Movie 9). Occasionally, they grow towards the polar areas (yellow arrowhead). (D) GFP-AtEB1 labeled EMT plus ends (arrowheads) are attracted to the cortical areas marked by the preceding PPB. At late telophase, the distance through which GFP-AtEB1 labeled EMT plus ends reach towards the cortex is reduced. (E) 3-D projection showing GFP-AtEB1 labeled EMT plus end trajectories directed towards the cortex, which are different from the main phragmoplast structure. GFP-MAP4 labeled EMTs (F-I) or GFP-AtEB1 labeled EMT plus ends (J-M) continue to reach the cortex at the former PPB site and display close proximity to FM4-64 labeled endosomes (red arrow and arrowheads). These endosomes display movement towards the minus end of these EMTs. (N-O) GFP-AtEB1 labeled plus end growth of EMTs (arrowheads) towards opposite sides of the cortex is maintained during cell plate and phragmoplast tilting (see additional file 10: Movie 10). (P) Enrichment of GFP-AtEB1 labeled microtubule plus ends (arrowhead) but not of YFP-MAP4 labeled microtubular parts at the phragmoplast midline. Time in F-I is given in seconds while that in J-O is indicated in minutes. Bars in A-O represent 8 μm while that in P represents 10 μm.
EMTs radiating from the former spindle poles attain a geometrical conformation correlating with cell-plate navigation and tilt-correction
Following observations on microtubule and organelle behavior during the early stages of mitosis, we analyzed events at later stages. Immediately after chromosomal separation at anaphase, the EMTs emanating from the region occupied by spindle poles 'probed' the cell cortex (Figure 5B, C; see also additional file 9: Movie 9) while exhibiting unidirectional growth at speeds of 8.52 ± 1.23 μm/min (n = 45). During late telophase, these EMTs mainly probed the cortical areas previously occupied by the PPB (Figure 5D). In addition, the EMTs originating from the non-facing surfaces of the daughter nuclei appeared fewer than the phragmoplast microtubules and occasionally exhibited growth trajectories towards the cell poles (Figure 5E). Because of the remarkable co-incidence between the EMTs and endocytosed material during preprophase, we investigated whether these EMTs approaching the cortex co-localized with FM4-64-labeled endocytic material during telophase. Such co-localization was observed both for GFP-MAP4 (Figure 5F–I) and GFP-AtEB1 (Figure 5J–M) labeled EMTs which were approaching the former PPB sites. During cell plate expansion, the EMTs appeared to navigate the cell plate and to align it to establish a clear line of division (Figure 5N, O; see also additional file 10: Movie 10). EMT plus ends within the phragmoplast midline (where the cell plate subsequently formed) were labeled strongly with GFP-AtEB1 but less strongly with YFP-MAP4, indicating the polarity of the phragmoplast microtubules, oriented with their plus ends towards the developing cell plate (Figure 5P).
In budding yeast, cytokinesis is delayed until the spindle is properly positioned, but in a mutant for Bim1 (EB1 homologue) this delay is abolished, resulting in abnormal cell division [37]. This suggests that Bim1 has an important role in sensing and positioning the spindle. Cytokinesis in plants often begins with tilted spindles, indicating the absence of a spindle alignment checkpoint in plant cells. However, this spindle tilting is corrected later during the progression of cytokinesis, which may indicate the involvement of a positioning sensor and correction mechanism in plant cells. We propose that the EMTs initiating from the spindle poles and approaching the cortex (Figure 5B–O) are involved in this sensor and tilting mechanism. Though different scenarios can be evoked for the correction mechanism, in each case a specialized cortical reference site would be required. We therefore investigated the effect of an auxin efflux inhibitor, NPA, which has been shown to block both vesicular trafficking and internalization of plasma membrane-localized proteins [36] without directly affecting microtubules or actin filaments [38].
Polarity inhibitor induces abnormal PPBs and shifts cell division planes
Prolonged NPA treatment in BY-2 cells caused inclined and periclinal cell divisions (Figure 6C) instead of normal anticlinal cell divisions (Figure 6A), as in tobacco VBI-0 cells [39]. Cells with aberrant division planes also exhibited major alterations in interphase cortical microtubule alignments in the daughter cells (Compare Figure 6A, B with Figure 6C, D). NPA treatment caused formation of abnormal PPBs (Figure 6E) that resulted in inclined spindles (Figure 6E) and phragmoplasts (Figure 6F), resulting in shifted division planes. In some NPA-treated cells, two separate PPBs were observed and the inclined cell plate was attached to the parental cell walls, with either end docking at one of the places marked by these two PPBs (Figure 6G–L; see also additional file 11: Movie 11). Closer observation of this two-PPB situation revealed a preferential attachment of the cell plate at the PPB sites connecting with the largest number of EMTs (Figure 6H). These results, together with observations from other laboratories, implicate a link between the intracellular establishment of polarity, endocytosis, the placement of initial PPBs and the final cell division planes.
Figure 6 NPA induces abnormal PPBs and altered cell divisions. (A-B) Normal anticlinal cell divisions (arrowheads in A) and transverse organization of cortical microtubules in NPA untreated cells. (C-D) Inclined and periclinal cell divisions (red arrowheads in C) with altered organization of GFP-MAP4 labeled cortical microtubules (red arrowheads in D) in NPA treated cells. A, C show single median sections and B, D show 3-D projections. Note that the first round of cell division (yellow arrowheads) is normal and a shift in the cell division planes occurs in the second round. (E-F) Formation of periclinal PPBs and spindles (arrowheads in E) and periclinal phragmoplasts (arrowhead in F). Bidirectional arrows in A, C, E and F show the long axes of the cells. (G-L) NPA treatment sometimes causes formation of two separate PPBs (arrowheads in G) equidistant from the nucleus, which results in tilted spindle formation (I) and phragmoplast initiation (J), phragmoplast growth (K) and cell plate docking (L) at sites marked by either of the PPBs (arrowheads) (see additional file 11: Movie 11). G shows 3-D projection and H-L show single median sections. Bars represent 10 μm and time is indicated in minutes.
Discussion
In mammalian and yeast cells, a microtubule plus end-mediated 'search and capture' mechanism has been credited with positioning and aligning the spindles [17,19] and determining the plane of cell division [40]. Our observations suggest that EMTs in plant cells may behave similarly to establish and regulate the cell division plane. Interestingly, it has been shown previously that injuries caused by inserting microneedles at these specific cortical sites, probed by EMTs during cytokinesis, affect cell plate alignment [41].
In mammalian and yeast cells, EB1 binds to adenomatous polyposis coli (APC). In epithelial cells, APC is mainly found at specialized cortical sites [42], providing a planar cue [43]. In mammalian cells, microtubule-mediated APC delivery to specialized cortical sites has been demonstrated [44,45]. A mechanism for attaining polarity cues is proposed for mammalian cells, according to which the EB1-labeled microtubule plus ends target to the specialized APC-marked cortical sites [46]. A similar role in attracting EB1-labeled microtubular plus ends towards the cortex has been attributed to Kar9p in budding yeast [17] and to Moe1 in fission yeast [47]. Moreover, it has been suggested that LIS1 is a regulated adapter between CLIP170 and cytoplasmic dynein at sites involved in cargo-microtubule loading and/or the control of microtubule dynamics [48].
Interestingly, plants seem to possess homologues for APC, Kar9p and Moe1, and the plant cytoskeletal-related Tonneau2 [49] protein contains a LisH domain present in LIS1 (our unpublished results based on the NCBI search engine). In maize, the Tangled1 (which is distantly related to the APC) mutant displays altered PPBs, spindles and phragmoplasts and shifted cell division planes. Furthermore, Tangled1 expression and its microtubule localization correlate with the cell division stage [50]. In mammalian cells, EB1 is required for microtubule tip-specific localization of APC but not vice versa. In the absence of EB1, APC localizes all along the microtubule lengths [50,51]. In addition, APC assembly in the cortical clusters is EB1-independent but depends on the existence of the armadillo domain [52]. In plants, it remains to be determined how EB1 localizes in the Tangled1 mutant and vice versa. From parallels in the mammalian cell literature one might hypothesize that Tangled1-mediated EB1 targeting to specialized cortical sites regulates the cell division planes in plants. However, there are several problems with this hypothesis. For instance, Tangled1 contains only the microtubule binding domain and lacks both the EB1 binding and the armadillo domains of APC [52]. Inversely, AtEB1 possesses an APC interaction domain [23] and a unique C-terminal acidic tail [22], and armadillo domain-containing proteins exist in Arabidopsis [53]. This leaves open the possibility that EB1 interacts with other PM-localized basic protein(s). In addition, plants lacking Tonneau2 fail to assemble PPB [4] and Tonneau2 possesses a LisH domain. From the parallel mammalian cell literature one might hypothesize that the LisH domain in Tonneau2 may mediate EMT plus end cortical interactions. Hence it will be interesting to determine how EB1 localizes in Tangled1 and Tonneau2 mutants and vice versa.
In mammalian cells, many proteins have been shown to associate in a microtubule plus end complex, which has been described as a plus end raft [54]. As with lipid rafts, protein concentration at the distal ends may allow a cascade of interactions in the restricted area of a microtubule plus end. This may, in turn, control the dynamic behavior of this cytoskeletal network and its anchoring to other structures [54]. An alternative to this would be that EMTs, by interpreting cell geometry and polarity cues, deposit protein(s) at the PPB, which subsequently attracts the phragmoplast microtubules. Conversely, the connection between EMT plus ends and endocytosis may indicate a role for localized endocytosis in modifying the PM architecture, which may transmit the memory for re-attracting EMT plus ends during cytokinesis. We consider that the polarity inhibitor NPA affects PPB formation by modulating endocytosis, as it does not affect microtubules but interferes with endocytosis and with polarity-based EMT plus end targeting to specialized areas of the PM. Thus, feedback loops, comprising polarity establishment-endocytosis-microtubule plus end guidance-further endocytosis, appear to be essential for defining and creating planes of cell division in plant cells.
Conclusion
In conclusion, our results suggest that in mitotic plant cells, EMT plus ends may act as cell shape/polarity sensing and orienting machines by their sustained cortical targeting, as shown for yeast [15,55]. EMTs in premitotic plant cells are bundled and bidirectional, as reported very recently in fission yeast by analysis of the EB1 homologue mal3p [56], indicating evolutionary conservation of the processes involved in defining cell division planes. Importantly, we show that at preprophase the targeting of EMT plus ends coincides with endocytosis events to establish a plant-specific cortical endocytic belt. During cytokinesis, this same belt again interacts with the EMT plus ends of the expanding phragmoplast to ensure proper cell plate navigation and docking. Our results reveal a link between the position of EMT plus ends, the establishment of intracellular polarity and the localization of endocytosis that is essential for the regulation of cell division planes in plants.
Methods
Plant material and growth conditions
Tobacco BY-2 cells were cultured and transformed as reported previously [9].
Construction of reporter genes
The construction of GFP-MAP4, YFP-MAP4, YFP-CLIP170 and GFP-AtEB1 was described previously [9,22]. GFP-Ara7 in vector pBSIIKS+ [33] was excised with HindIII-XbaI and sub-cloned into the binary vector pBINPLUS. STtmd-YFP in vector pMON [57] was digested with PstI -SmaI and cloned into the binary vector pCAMBIA 1390.
Fluorescent dyes and drugs
FM4-64 (Molecular Probes) dissolved in water was applied to the BY-2 cells at 2 μM final concentration for 5 min. The cells were washed with BY-2 medium to remove excess dye and were observed immediately. Alexa 633 (Molecular probes) and Mitotracker (Molecular probes) dissolved in water were also applied at 2 μM final concentration and the cells were observed immediately. Stock solutions of taxol (Sigma-Aldrich), latrunculin B (Sigma-Aldrich) and NPA (Sigma-Aldrich) in DMSO were applied to the cells to give final concentrations of 10 μM, 10 μM and 50 μM respectively. Stock solutions of oryzalin (Greyhound Chromatography and Allied Chemicals, Merseyside, UK) were prepared in ethanol and used at 10 μM final concentration.
Live cell analysis
For live cell analysis, the Zeiss CLSM510 system implemented on an inverted (Axiovert 100) microscope was used. The microscopy system, sample preparation, single wavelength scanning, image processing and movie generation were as previously described [9]. Dual color imaging was performed using dual excitation/emission scanning in multitracking mode. For GFP /YFP dual scanning, we used excitation/emission combinations of 458 nm/ BP 475–525 for GFP and 514 nm/ BP 530–600 for YFP, in combination with the HFT 458/514 primary and NFT515 secondary dichroic splitters. For GFP/ FM4-64 dual scanning, we used excitation/emission combinations of 488 nm/ BP 505–550 for GFP and 543 nm/ LP585 for FM4-64, in combination with the HFT 488/543 primary and NFT545 secondary dichroic splitters. For GFP/ Alexa dual scanning, we used excitation/emission combinations of 488 nm/ BP 505–550 for GFP and 633 nm/ LP650 for Alexa, in combination with the HFT UV/488/543/633 primary and NFT545 secondary dichroic splitters. All filters were from Zeiss. For time-lapse analysis, images were obtained at 1–10 s time intervals. All experiments were repeated 3–5 times. Acquired images were processed using LSM510 Image Browser version 3.0 (Zeiss Corp.). Maximum projections were obtained from 0.5 μm spaced serial optical sections and were exported as TIFF files. For time-series scans, all the images were exported as time-series TIFF files. The exported images were processed with Adobe Photoshop version 5.0 (Adobe Systems Inc.). For individual plant microtubule growth measurements, all time scans were analyzed in the animation mode of LSM510 Image Browser 3.2 (Zeiss Corp.) by marking the single ends of individual microtubules in each image by a zoom function and tracking them for several minutes. The shortest displacement of the plus ends resolvable in this analysis was 0.1 μm. The time values were obtained from the respective frame times in the time-lapse. Thereafter, the data were manually transferred into Excel and processed. The microtubule growth histories (kymographs) were obtained by processing the raw data in LSM510 Image Browser 3.2 (Zeiss Corp.).
Authors' contributions
PD designed the experiments, acquired the data, analyzed and interpreted them and drafted the manuscript. TWJG supervised the research. PD, JM, MH and TWJG participated in manuscript designing, coordination and editing. All authors read and approved the final manuscript.
Supplementary Material
Additional File 1
GFP-AtEB1 labeled cortical microtubule plus-end dynamics exhibiting bidirectional growth during interphase. The movie is accelerated 80 times. The real time is 4 min.
Click here for file
Additional File 2
GFP-AtEB1 (green) labels growing plus ends of YFP-MAP4 (red) decorated microtubules exhibiting microtubule guidance and bundling. The movie is accelerated 90 times. The real time is 3.75 min.
Click here for file
Additional File 3
GFP-AtEB1 (green) colocalizes with YFP-CLIP170 (red) on growing microtubule plus ends. The movie is accelerated 60 times. The real time is 6 min.
Click here for file
Additional File 4
GFP-AtEB1 labeled endoplasmic microtubule plus ends display bidirectional (incoming and outgoing) growth polarity and bundling. The movie is accelerated 110 times. The real time is 6.5 min.
Click here for file
Additional File 5
During PPB maturation, GFP-AtEB1 labeled endoplasmic microtubule plus ends radiate symmetrically, mainly in an outward direction from the NE. The movie is accelerated 100 times. The real time is 12 min.
Click here for file
Additional File 6
FM4-64 labeled endocytic vesicles (red) preferentially internalize from the cortical areas approached by the GFP-AtEB1 labeled EMT plus ends (green). The movie is accelerated 80 times. The real time is 20 min.
Click here for file
Additional File 7
Oryzalin-induced microtubule depolymerization disrupts internalization routes of FM4-64 labeled endosomes. The movie is accelerated at 80 times. The real time is 7 min.
Click here for file
Additional File 8
GFP-AtEB1 marked EMT plus ends bridging NE and PPB. Note that during PPB narrowing, FM4-64 labeled endocytic vesicles preferentially internalize from the cortical areas approached by the GFP-AtEB1. The movie is accelerated at 120 times. The real is 18 min.
Click here for file
Additional File 9
At the onset of the phragmoplast stage, GFP-AtEB1 labeled EMT plus ends originating from the former spindle poles grow towards the cortex. The movie is accelerated 60 times. The real time is 3 min.
Click here for file
Additional File 10
GFP-AtEB1 labeled plus end growth of EMTs towards opposite sides of the cortex is maintained during cell plate and phragmoplast tilting. The movie is accelerated 60 times. The real time is 2 min.
Click here for file
Additional File 11
NPA treatment sometimes causes formation of two separate PPBs equidistant from the nucleus, resulting in tilted spindle formation and phragmoplast initiation, growth and cell plate docking at sites marked by either of the PPBs. The movie is accelerated 550 times. The real time is 120 min.
Click here for file
Acknowledgements
We acknowledge F. Baluska for stimulating discussions. We are grateful to A. Nakano (RIKEN, Saitama, Japan) for the GFP-Ara7 construct and J. Carette (Wageningen Univ., the Netherlands) for the ST-YFP construct. P.D. and T.W.J.G. were supported by NWO FOM-ALW 805.47.012 and by NWO van der Leeuw 835.25.004. J. M. was supported by a Volkswagon Stiftung grant to M.H.
==== Refs
Smith L Plant cell division: building walls in the right places Nat Rev Mol Cell Biol 2001 2 33 40 11413463 10.1038/35048050
Mathur J Plant cytoskeleton: reinforcing lines of division in plant cells Curr Biol 2004 14 R287 289 15062125 10.1016/j.cub.2004.03.027
Pickett-Heaps JD Northcote DH Organization of microtubules and endoplasmic reticulum during mitosis and cytokinesis in wheat meristems J Cell Sci 1966 1 109 120 5929804
Traas J Bellini C Nacry P Kronenberger J Bouchez D Caboche M Normal differentiation patterns in plants lacking microtubular preprophase bands Nature 1995 375 676 677 10.1038/375676a0
Mineyuki Y The Preprophase Band of Microtubules: Its Function as a Cytokinetic Apparatus in Higher Plants Int Rev Cytol 1999 187 1 49
Cleary AL Smith LG The Tangled1 gene is required for spatial control of cytoskeletal arrays associated with cell division during maize leaf development Plant Cell 1998 10 1875 1888 9811795 10.1105/tpc.10.11.1875
Sinnott EW Bloch R Cytoplasmic behaviour during division of vacuolate plant cells Proc Natl Acad Sci U S A 1940 26 223 227 16588342
Flanders DJ Rawlins DJ Shaw PJ Lloyd CW Nucleus-associated microtubules help determine the division plane of plant epidermal cells: avoidance of four-way junctions and the role of cell geometry J Cell Biol 1990 110 1111 1122 2324196 10.1083/jcb.110.4.1111
Dhonukshe P Gadella TW Jr Alteration of microtubule dynamic instability during preprophase band formation revealed by yellow fluorescent protein-CLIP170 microtubule plus-end labeling Plant Cell 2003 15 597 611 12615935 10.1105/tpc.008961
Murata T Wada M Effects of centrifugation on preprophase-band formation in Adiantum protonemata Planta 1991 183 391 398 10.1007/BF00197738
Katsuta J Hashiguchi Y Shibaoka H The role of the cytoskeleton in positioning of the nucleus in premitotic tobacco BY-2 cells J Cell Sci 1990 95 413 422
Adames NR Cooper JA Microtubule interactions with the cell cortex causing nuclear movements in Saccharomyces cerevisiae J Cell Biol 2000 149 863 874 10811827 10.1083/jcb.149.4.863
Gundersen GG Evolutionary conservation of microtubule-capture mechanisms Nat Rev Mol Cell Biol 2002 3 296 304 11994749 10.1038/nrm777
Mimori-Kiyosue Y Tsukita S "Search-and-capture" of microtubules through plus-end-binding proteins (+TIPs) J Biochem (Tokyo) 2003 134 321 326 14561716
Tran PT Marsh L Doye V Inoue S Chang F A mechanism for nuclear positioning in fission yeast based on microtubule pushing J Cell Biol 2001 153 397 411 11309419 10.1083/jcb.153.2.397
Faivre-Moskalenko C Dogterom M Dynamics of microtubule asters in microfabricated chambers: the role of catastrophes Proc Natl Acad Sci U S A 2002 99 16788 16793 12486218 10.1073/pnas.252407099
Lee L Tirnauer JS Li J Schuyler SC Liu JY Pellman D Positioning of the mitotic spindle by a cortical-microtubule capture mechanism Science 2000 287 2260 2262 10731147 10.1126/science.287.5461.2260
Ahringer J Control of cell polarity and mitotic spindle positioning in animal cells Curr Opin Cell Biol 2003 15 73 81 12517707 10.1016/S0955-0674(02)00018-2
Oliferenko S Balasubramanian MK Astral microtubules monitor metaphase spindle alignment in fission yeast Nat Cell Biol 2002 4 816 820 12360293 10.1038/ncb861
Wang H Oliferenko S Balasubramanian MK Cytokinesis: relative alignment of the cell division apparatus and the mitotic spindle Curr Opin Cell Biol 2003 15 82 87 12517708 10.1016/S0955-0674(02)00006-6
Geelen DN Inze DG A bright future for the bright yellow-2 cell culture Plant Physiol 2001 127 1375 1379 11743076 10.1104/pp.127.4.1375
Mathur J Mathur N Kernebeck B Srinivas BP Hulskamp M A novel localization pattern for an EB1-like protein links microtubule dynamics to endomembrane organization Curr Biol 2003 13 1991 1997 14614826 10.1016/j.cub.2003.10.033
Bu W Su LK Characterization of functional domains of human EB1 family proteins J Biol Chem 2003 278 49721 49731 14514668 10.1074/jbc.M306194200
Chan J Jensen CG Jensen LC Bush M Lloyd CW The 65-kDa carrot microtubule-associated protein forms regularly arranged filamentous cross-bridges between microtubules Proc Natl Acad Sci U S A 1999 96 14931 14936 10611315 10.1073/pnas.96.26.14931
Krylyshkina O Anderson KI Kaverina I Upmann I Manstein DJ Small JV Toomre DK Nanometer targeting of microtubules to focal adhesions J Cell Biol 2003 161 853 859 12782685 10.1083/jcb.200301102
Drykova D Cenklova V Sulimenko V Volc J Draber P Binarova P Plant gamma-tubulin interacts with alphabeta-tubulin dimers and forms membrane-associated complexes Plant Cell 2003 15 465 480 12566585 10.1105/tpc.007005
Caudron N Arnal I Buhler E Job D Valiron O Microtubule nucleation from stable tubulin oligomers J Biol Chem 2002 277 50973 50979 12393880 10.1074/jbc.M209753200
Shaw SL Kamyar R Ehrhardt DW Sustained microtubule treadmilling in Arabidopsis cortical arrays Science 2003 300 1715 1718 12714675 10.1126/science.1083529
Piehl M Cassimeris L Organization and dynamics of growing microtubule plus ends during early mitosis Mol Biol Cell 2003 14 916 925 12631713 10.1091/mbc.E02-09-0607
Boevink P Oparka K Santa Cruz S Martin B Betteridge A Hawes C Stacks on tracks: the plant Golgi apparatus traffics on an actin/ER network Plant J 1998 15 441 447 9750355 10.1046/j.1365-313X.1998.00208.x
Schmit AC Lambert AM Characterization and dynamics of cytoplasmic F-actin in higher plant endosperm cells during interphase, mitosis, and cytokinesis J Cell Biol 1987 105 2157 2166 3680376 10.1083/jcb.105.5.2157
Mineyuki Y Karahara I Staehelin LA Quantitative analysis of cytoskeletal arrays and endocytic vesicles in the cortex of dividing plant cells by dual-axis EM tomography Procceedings of the 30th NIPS International Symposium, March 12–15, 2003 2003 Okazaki, Japan 120 121
Ueda T Yamaguchi M Uchimiya H Nakano A Ara6, a plant-unique novel type Rab GTPase, functions in the endocytic pathway of Arabidopsis thaliana Embo J 2001 20 4730 4741 11532937 10.1093/emboj/20.17.4730
Nebenfuhr A Frohlick JA Staehelin LA Redistribution of Golgi stacks and other organelles during mitosis and cytokinesis in plant cells Plant Physiol 2000 124 135 151 10982429 10.1104/pp.124.1.135
Hoshino H Yoneda A Kumagai F Hasezawa S Roles of actin-depleted zone and preprophase band in determining the division site of higher-plant cells, a tobacco BY-2 cell line expressing GFP-tubulin Protoplasma 2003 222 157 165 14714204 10.1007/s00709-003-0012-8
Geldner N Friml J Stierhof YD Jurgens G Palme K Auxin transport inhibitors block PIN1 cycling and vesicle trafficking Nature 2001 413 425 428 11574889 10.1038/35096571
Muhua L Adames NR Murphy MD Shields CR Cooper JA A cytokinesis checkpoint requiring the yeast homologue of an APC-binding protein Nature 1998 393 487 491 9624007 10.1038/31014
Petrasek J Cerna A Schwarzerova K Elckner M Morris DA Zazimalova E Do phytotropins inhibit auxin efflux by impairing vesicle traffic? Plant Physiol 2003 131 254 263 12529533 10.1104/pp.012740
Petrasek J Elckner M Morris DA Zazimalova E Auxin efflux carrier activity and auxin accumulation regulate cell division and polarity in tobacco cells Planta 2002 216 302 308 12447544 10.1007/s00425-002-0845-y
Canman JC Cameron LA Maddox PS Straight A Tirnauer JS Mitchison TJ Fang G Kapoor TM Salmon ED Determining the position of the cell division plane Nature 2003 424 1074 1078 12904818 10.1038/nature01860
Gunning BE Wick SM Preprophase bands, phragmoplasts, and spatial control of cytokinesis J Cell Sci Suppl 1985 2 157 179 3867671
Nathke IS Adams CL Polakis P Sellin JH Nelson WJ The adenomatous polyposis coli tumor suppressor protein localizes to plasma membrane sites involved in active cell migration J Cell Biol 1996 134 165 179 8698812 10.1083/jcb.134.1.165
Bienz M Spindles cotton on to junctions, APC and EB1 Nat Cell Biol 2001 3 E67 68 11231588 10.1038/35060140
Askham JM Moncur P Markham AF Morrison EE Regulation and function of the interaction between the APC tumour suppressor protein and EB1 Oncogene 2000 19 1950 1958 10773885 10.1038/sj.onc.1203498
Mimori-Kiyosue Y Shiina N Tsukita S Adenomatous polyposis coli (APC) protein moves along microtubules and concentrates at their growing ends in epithelial cells J Cell Biol 2000 148 505 518 10662776 10.1083/jcb.148.3.505
Lu B Roegiers F Jan LY Jan YN Adherens junctions inhibit asymmetric division in the Drosophila epithelium Nature 2001 409 522 525 11206549 10.1038/35054077
Chen CR Chen J Chang EC A conserved interaction between Moe1 and Mal3 is important for proper spindle formation in Schizosaccharomyces pombe Mol Biol Cell 2000 11 4067 4077 11102508
Coquelle FM Caspi M Cordelieres FP Dompierre JP Dujardin DL Koifman C Martin P Hoogenraad CC Akhmanova A Galjart N LIS1, CLIP-170's key to the dynein/dynactin pathway Mol Cell Biol 2002 22 3089 3102 11940666 10.1128/MCB.22.9.3089-3102.2002
Camilleri C Azimzadeh J Pastuglia M Bellini C Grandjean O Bouchez D The Arabidopsis TONNEAU2 gene encodes a putative novel protein phosphatase 2A regulatory subunit essential for the control of the cortical cytoskeleton Plant Cell 2002 14 833 845 11971138 10.1105/tpc.010402
Smith LG Gerttula SM Han S Levy J Tangled1: a microtubule binding protein required for the spatial control of cytokinesis in maize J Cell Biol 2001 152 231 236 11149933 10.1083/jcb.152.1.231
Tirnauer JS Bierer BE EB1 proteins regulate microtubule dynamics, cell polarity, and chromosome stability J Cell Biol 2000 149 761 766 10811817 10.1083/jcb.149.4.761
Barth AI Siemers KA Nelson WJ Dissecting interactions between EB1, microtubules and APC in cortical clusters at the plasma membrane J Cell Sci 2002 115 1583 1590 11950877
Coates JC Armadillo repeat proteins: beyond the animal kingdom Trends Cell Biol 2003 13 463 471 12946625 10.1016/S0962-8924(03)00167-3
Galjart N Perez F A plus-end raft to control microtubule dynamics and function Curr Opin Cell Biol 2003 15 48 53 12517703 10.1016/S0955-0674(02)00007-8
Korinek WS Copeland MJ Chaudhuri A Chant J Molecular linkage underlying microtubule orientation toward cortical sites in yeast Science 2000 287 2257 2259 10731146 10.1126/science.287.5461.2257
Busch KE Brunner D The microtubule plus end-tracking proteins mal3p and tip1p cooperate for cell-end targeting of interphase microtubules Curr Biol 2004 14 548 559 15062095 10.1016/j.cub.2004.03.029
Carette JE Stuiver M Van Lent J Wellink J Van Kammen A Cowpea mosaic virus infection induces a massive proliferation of endoplasmic reticulum but not Golgi membranes and is dependent on de novo membrane synthesis J Virol 2000 74 6556 6563 10864669 10.1128/JVI.74.14.6556-6563.2000
| 15831100 | PMC1087477 | CC BY | 2021-01-04 16:39:15 | no | BMC Biol. 2005 Apr 14; 3:11 | utf-8 | BMC Biol | 2,005 | 10.1186/1741-7007-3-11 | oa_comm |
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-5-351581118710.1186/1471-2407-5-35Research ArticleBreast conserving surgery versus mastectomy: cancer practice by general surgeons in Iran Najafi Massoome [email protected] Mandana [email protected] Ahmad [email protected] Esmat [email protected] Ali [email protected] Iranian Center for Breast Cancer (ICBC), Tehran, Iran2 Department of Surgery, Faculty of Medicine, Tehran University of Medical Sciences, Tehran, Iran2005 5 4 2005 5 35 35 29 9 2004 5 4 2005 Copyright © 2005 Najafi 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 appear to be geographical differences in decisions to perform mastectomy or breast conserving surgery for early-stage breast cancer. This study was carried out to evaluate general surgeons' preferences in breast cancer surgery and to assess the factors predicting cancer practice in Iran.
Methods
A structured questionnaire was mailed to 235 general surgeons chosen from the address list of the Iranian Medical Council. The questionnaire elicited information about the general surgeons' characteristics and about their work experience, posts they have held, number of breast cancer operations performed per year, preferences for mastectomy or breast conserving surgery, and the reasons for these preferences.
Results
In all, 83 surgeons returned the completed questionnaire. The results indicated that only 19% of the surgeons routinely performed breast conserving surgery (BCS) and this was significantly associated with their breast cancer case load (P < 0.01). There were no associations between BCS practice and the other variables studied. The most frequent reasons for not performing BCS were uncertainty about conservative therapy results (46%), uncertainty about the quality of available radiotherapy services (32%), and the probability of patients' non-compliance in radiotherapy (32%).
Conclusion
The findings indicate that Iranian surgeons do not routinely perform BCS as the first and the best treatment modality. Further research is recommended to evaluate patients' outcomes after BCS treatment in Iran, with regard to available radiotherapy facilities and cultural factors (patients' compliance).
==== Body
Background
Many randomized clinical trials have demonstrated that patient survival rates are similar after treatment by mastectomy or by conservative surgery and radiotherapy [1-4]. However, the National Institute of Health Consensus Conference concluded that for most women with early-stage breast cancers (stages I and II), breast conservation surgery (BCS) is an appropriate method of treatment [5]. Despite these findings, mastectomy remains the most prevalent surgical treatment for early-stage breast cancer in many parts of the world. In Iran, BCS is an uncommon modality for this condition. For example, in one study in Isfahan University of Medical Sciences, 386 breast cancer patients were reviewed and mastectomy was found to be the most prevalent surgical treatment [6]. Similarly, in another study in Tehran University of Medical Sciences, 348 breast cancer patients of whom 50% had stage I or stage II breast cancer were evaluated, and it was found that only 22 patients underwent BCS [7].
It is argued that one factor influencing choice of treatment in breast cancer patients is the attitude of the operating surgeon. There is evidence that most women follow surgeons' recommendations, their primary source of information about treatment options. Studies have shown that surgeons recommended a specific treatment option in 69% of cases; in 89% of these cases they offered mastectomy and in 11% BCS. The rates of compliance with these recommendations were reported to be 93% and 89% respectively [8].
The present study was designed to evaluate the preferences of surgeons in Iran regarding breast cancer treatment and the factors influencing the type of treatment recommended.
Methods
A descriptive study was conducted to investigate cancer practice among general surgeons in Iran. A specially designed questionnaire was mailed to a sample of 235 general surgeons, selected randomly from the Iranian Medical Council's address list for general surgeons and including a few surgical oncologists. All general surgeons (n = 1977, 1831 male and 146 female) belong to the Council. It is worth noting that in Iran, gynecologists and plastic surgeons do not perform breast cancer surgery. The sample size was determined by expectations of the rate of BCS treatment by the general surgeons: it was estimated that at best 15% and at worst 5% of the surgeons might perform BCS. Based on this assumption, a sample of 81 surgeons would give results with a 99% confidence level. To achieve this sample size it was decided to mail the questionnaire to 235 general surgeons, nearly three times more than needed. The questionnaire addressed the surgeons' characteristics including age, gender, work experience, posts held (teaching hospitals or private and public institutions), and number of breast cancer patients treated per year. In addition, respondents were asked whether they routinely performed BCS, and if the answer was "no" the reasons were solicited. Data were analyzed in a descriptive fashion, reporting numbers and percentages and using the chi-square test (Fisher's exact test) where necessary.
Results
In all, 83 completed questionnaires were returned, giving a response rate of 33%. The characteristics of the study sample are shown in Table 1. The mean age of the respondents was 55.2 years (SD = 12.3; range 32 to 80). Most surgeons were male (n = 75, 90%), had more than 10 years work experience (76%), and were working at public or private centers (75%). For 80% of the surgeons surveyed, the average number of breast cancer patients treated per year was less than 20; 20% reported that their annual case-load was more than 20.
Only 19% of the surgeons said that they performed BCS routinely. Table 2 shows the reasons given for avoiding BCS. The most frequent reason was the surgeons' uncertainty about the results of breast conserving surgery (46.3%). The other most frequently-stated barriers to performing BCS were uncertainty about radiotherapy services and patients' non-compliance in radiotherapy (32.8%). Of the factors studied, only breast cancer case-load had a statistically significant association with performance of BCS (P = 0.007). The results are shown in Table 3.
Discussion
Breast-conserving surgery is a treatment modality for early-stage breast cancer that causes less physical disfigurement and psychological trauma to the patient. Many prospective randomized trials have demonstrated that overall and disease-free survival rates for early-stage breast cancer are equivalent after mastectomy or BCS with postoperative radiotherapy [2-4], [8,9]. There appear to be geographical differences in the surgical treatment of early-stage breast cancer. Evidence suggests that physician recommendation exerts the most significant influence on surgical treatment selection for the majority of women [10].
In this study only 19% of surgeons used BCS in their practice while the remaining 81% performed mastectomy. This concurs with two other studies, performed in Tehran and Isfahan University of Medical Sciences, showing that mastectomy was the most prevalent surgical treatment for breast cancer patients [6,7]. However, it is arguable that BCS rates would rise in Iran if general surgeons, for example, could benefit from continuing medical education, attend scientific meetings, and have access to modern textbooks and current journals. Discussion of these topics lies beyond the scope of this study. Briefly, it should be noted that there is a National Society of Surgeons in Iran. In addition to other scientific meetings that address oncology issues, members of this society have their own annual and occasional scientific meetings, and usually receive printed educational materials including issues on current technology. At present all surgeons and radiation therapists in teaching hospitals and many working in the private sector are Board-certified. Some modern radiation therapy facilities in Iran are very similar to existing facilities in developed countries. Furthermore, we suspect that predicting patient non-compliance with radiation therapy might be subjective impressions on the part of the surgeons, since there is no evidence or published data to indicate that patient compliance with radiation therapy is low.
The findings indicated that gender had no effect on the performance of BCS. This may be due to the small number of female surgeons in this study. Similarly, in a study comparing rates of BCS between male and female surgeons, no statistically significant difference was found between the two groups [11]. However, it is interesting that in our study, 14 out of the 75 male surgeons (19%) indicated that they routinely perform BCS, while 3 out of the 8 female surgeons (38%) stated that they performed BCS.
We found that the surgeons' age and work experience was not significantly associated with performing BCS; other studies have shown that younger surgeons are more likely to perform BCS. In one retrospective study that reviewed the medical records of 952 patients, it was found that surgeons who graduated after 1960 performed BCS more frequently than those who graduated before 1960 [12]. Another study showed that surgeons who graduated after 1980 performed BCS at a significantly higher rate than those who graduated before 1961 [13]. However, the mean age of our study sample was 55.2 (SD = 12.3) years and the mean duration of work experience was 23.2 years (SD = 12.7). In other words, most of our participants were old; only 17% were relatively young. This might explain why most surgeons in this study stated that they performed mastectomy. This is also consistent with the fact that mastectomy is the most prevalent surgical therapy for breast cancer in our teaching hospitals [6,7]. However, the findings from the present study indicate that there is a trend towards more BCS at teaching hospitals, and it is very likely that a bigger sample would show a significant difference between teaching hospitals and private institutes.
There was a statistically significant association between the experience of the surgeons (measured by the number of beast cancer patients treated per year) and performance of BCS. Surgeons treating more than 20 breast cancer patients per year had a statistically significant higher BCS rate (P = 0.007). Few studies have reported a correlation between the surgeons' experience and their breast conservation surgery rates. A population-based study found that patients treated by surgeons with higher breast cancer case-loads were more likely to receive breast conservation therapy [14]. A survey of surgeons carried out in 1986 showed no evidence of a relationship between the number of breast surgery operations performed and the rate of conservative surgery; but in a repeat survey in 1990, when the overall conservative surgery rate had risen to 42%, a significant association was found for surgeons treating 20 or more patients per year [13].
In this study, the most common reasons given for avoiding BCS were uncertainty about the results of breast conservation therapy (46.3%), uncertainty about the quality of available radiotherapy services (32%), and probability of patient non-compliance (32%). Other reasons noted by the surgeons were the higher cost of breast conserving therapy (17.9%), non-availability of radiotherapy facilities (25.4%), and insufficiency of experience with the BCS technique (14.9%). A few studies have reported surgeons' reasons for avoiding BCS. In one study, the majority of surgeons believed that long-term disease-free survival was equal for mastectomy and BCS, and the preference for mastectomy was mostly due to the inconvenience of radiotherapy [15].
Conclusion
In conclusion, considering the study limitations (e.g. the sample size was small, the surgeons were predominantly old and only eight of them were female), the findings suggest that the overall performance of BCS in Iran is low. In order to increase the rate of conservation therapy there is need to study the outcomes of breast conservation therapy in Iran to determine overall and disease-free survival and local recurrence rates. At the same time the quality of available radiotherapy services should be evaluated. Perhaps this information might help to improve breast cancer outcomes.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
MN contributed to the study design, data collection and analysis, and wrote the first draft. ME contributed to the study design and data analysis. AK contributed to the study design and data collection. EH contributed to the data collection. AM contributed to the data analysis and wrote the final draft.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Figures and Tables
Table 1 The characteristics of the study sample
Number %
Age (years)
<40 14 17
40–59 37 45
≥60 32 38
Mean (SD) 55.2 (12.3)
Gender
Female 8 10
Male 75 90
City of practice
Tehran (the capital) 20 24
Other cities 63 76
Work experience (years)
≤10 20 24
>10 63 76
Mean (SD) 23.2 (12.7)
Occupation status
Working at a teaching hospital. 21 25
Working at a public/private hospital 62 75
Breast cancer case load (patient/per year, n = 70)
≤20 56 80
>20 14 20
Preferred type of surgery for breast cancer
Mastectomy 67 81
Breast conserving surgery 16 19
Table 2 Reasons stated by the surgeons for avoiding conservative therapy*
Number % of the study sample
Uncertainty about results of conservative therapy 32 46.3
Uncertainty about radiotherapy services 22 32.8
Probability of patient incompliance 22 32.8
Unavailability of radiotherapy services 17 25.4
Higher cost of conservative therapy 12 17.9
Lack of enough experience for performing BCS 10 14.9
* The total exceeds the sample size since each respondent could state several reasons.
Table 3 Association between breast cancer practice and the respondents' characteristics
Breast conservation No. (%) Mastectomy No. (%) P value
Gender 0.35
Male 14 (82) 61 (92)
Female 3 (18) 5 (8)
Age (years) 0.60
<40 4 (24) 10 (15)
40–59 6 (35) 31 (47)
≥60 7 (41) 25 (38)
Work experience 1.0
≤10 4 (24) 16 (24)
>10 13 (76) 50 (76)
Working at a teaching hospital 0.12
Yes 7 (41) 14 (21)
No 10 (59) 52 (79)
Breast cancer case load of the surgeon (n = 70) 0.007
≤20 8 (53) 48 (87)
>20 7 (47) 7 (13)
==== Refs
Veronesi U Salvadori B Luini A Greco M Saccozzi R del Vecchio M Mariani L Zurrida S Rilke F Breast conservation is a safe method in patients with small cancer of the breast. Long term results of three randomized trials on 1,973 patients Eur J Cancer 1995 31A 1574 1579 7488404 10.1016/0959-8049(95)00271-J
Arriagada R Le MG Rochard F Contesso G Conservative treatment versus mastectomy in early breast cancer: patterns of failure with 15 years of follow-up data. Institute Gustave-Roussy Breast Cancer Group J Clin Oncol 1996 14 1558 1564 8622072
Fisher B Anderson S Redmond CK Wolmark N Wickerham DL Cronin WM Reanalysis and results after 12 years of follow-up in a randomized clinical trial comparing total mastectomy with lumpectomy with or without irradiation in the treatment of breast cancer N Engl J Med 1995 333 1456 1461 7477145 10.1056/NEJM199511303332203
Jacobson JA Danforth DN Cowan KH d'Angelo T Steinberg SM Pierce L Lippman ME Lichter AS Glatstein E Okunieff P Ten-year results of a comparison of conservation with mastectomy in the treatment of stage I and II breast cancer N Engl J Med 1995 332 907 911 7877647 10.1056/NEJM199504063321402
NIH Consensus Conference. Treatment of early-stage breast cancer JAMA 1991 265 391 395 1984541 10.1001/jama.265.3.391
Mousavi R Akbari ME Epidemiologic study of breast diseases in Isfahan PhD thesis 1993 Isfahan University of Medical Sciences, Department of Surgery
Atai Fashtami H Atri M A retrospective study of 350 cases of breast cancer: histological type, sex, stage, treatment and the results of 10 year follow-up PhD thesis 1994 Tehran University of Medical Sciences, Department of Surgery
Kotwall CA Maxwell JG Covington DL Churchill P Smith SE Covan EK Clinicopathologic factors and patient perceptions associated with surgical breast-conserving treatment Ann Surg Oncol 1996 3 169 175 8646518
Van Donegan JA Bartelink H Fentiman IS Lerut T Mignolet F Olthuis G van der Schueren E Sylvester R Winter J van Zijl K Randomized clinical trial to assess the value of breast-conserving therapy in stage I and II breast cancer: EORTC 10801 trial J Natl Cancer Inst Monogr 1992 11 15 18 1627421
Harris JR Lippman ME Morrow M Osborne CK Disease of the breast 2000 2 Lippincot Williams & Wilkins
Weinberg E Woods S Grannan K Hendy MP The influence of the gender of the surgeon on surgical procedure preference for breast cancer Am Surg 2002 68 398 400 11952257
Kotwall C Covington D Churchill P Brinker C Weintritt D Maxwell JG Breast conservation surgery for breast cancer at a regional medical center Am J Surg 1998 176 510 514 9926780 10.1016/S0002-9610(98)00254-2
Kotwall CA Covington DL Rutledge R Churchill MP Meyer AA Patient, hospital and surgeon factors associated with breast conservation surgery. A statewide analysis in North Carolina Ann Surg 1996 224 419 426 8857847 10.1097/00000658-199610000-00001
Taylor R Stubbs JM Langlands AO Boyages J Predictors of mastectomy for women with breast cancer in greater western region of Sydney Breast J 1999 5 116 121 11348270 10.1046/j.1524-4741.1999.00135.x
Stafford D Szczys R Becker R Anderson J Bushfield S How breast cancer treatment decisions are made by women in North Dakota Am J Surg 1997 176 515 519 10.1016/S0002-9610(98)00257-8
| 15811187 | PMC1087478 | CC BY | 2021-01-04 16:03:03 | no | BMC Cancer. 2005 Apr 5; 5:35 | utf-8 | BMC Cancer | 2,005 | 10.1186/1471-2407-5-35 | oa_comm |
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-5-371582631610.1186/1471-2407-5-37Study ProtocolPhase III trial of postoperative cisplatin, interferon alpha-2b, and 5-FU combined with external radiation treatment versus 5-FU alone for patients with resected pancreatic adenocarcinoma – CapRI: study protocol [ISRCTN62866759] Knaebel HP [email protected]ärten A [email protected] J [email protected] K [email protected] C [email protected] K [email protected] H [email protected] S [email protected] T [email protected] H [email protected] U [email protected] J [email protected] V [email protected]üchler MW [email protected] University of Heidelberg, Department of Surgery, Im Neuenheimer Feld 110,69120 Heidelberg, Germany2 University of Heidelberg, Department of Radiology, Im Neuenheimer Feld 400,69120 Heidelberg, Germany3 University of Heidelberg, Department of Medicine, Im Neuenheimer Feld 410,69120 Heidelberg, Germany4 LMU München, IBE Medical School, Marchioninistr. 15, 81377 München, Germany5 National Centre for Tumor Diseases, Im Neuenheimer Feld 350, 69120 Heidelberg, Germany2005 12 4 2005 5 37 37 3 3 2005 12 4 2005 Copyright © 2005 Knaebel 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.
After surgical intervention with curative intention in specialised centres the five-year survival of patients with carcinoma of the exocrine pancreas is only 15%. The ESPAC-1 trial showed an increased five-year survival of 21% achieved with adjuvant chemotherapy. Investigators from the Virginia Mason Clinic have reported a 5-year survival rate of 55% in a phase II trial evaluating adjuvant chemotherapy, immunotherapy and external-beam radiation.
Design
The CapRI study is an open, controlled, prospective, randomised multi-centre phase III trial. Patients in study arm A will be treated as outpatients with 5-Fluorouracil; Cisplatin and 3 million units Interferon alpha-2b for 5 1/2 weeks combined with external beam radiation. After chemo-radiation the patients receive continuous 5-FU infusions for two more cycles. Patients in study arm B will be treated as outpatients with intravenous bolus injections of folinic acid, followed by intravenous bolus injections of 5-FU given on 5 consecutive days every 28 days for 6 cycles. A total of 110 patients with specimen-proven R0 or R1 resected pancreatic adenocarcinoma will be enrolled. An interim analysis for patient safety reasons will be done one year after start of recruitment. Evaluation of the primary endpoint will be performed two years after the last patients' enrolment.
Discussion
The aim of this study is to evaluate the overall survival period attained by chemo-radiotherapy including interferon alpha 2b administration with adjuvant chemotherapy. The influence of interferon alpha on the effectiveness of the patients' chemoradiation regimen, the toxicity, the disease-free interval and the quality of life are analysed. Different factors are tested in terms of their potential role as predictive markers.
==== Body
Background
Only 10–20% of patients with pancreatic cancer can be resected with curative intent at the time of diagnosis. Loco-regional recurrence and/or metastatic disease develop in the majority of patients who undergo pancreatic resection. Relapse occurs within 9–15 months after initial presentation and patients have median life expectancies of only 12–15 months without adjuvant therapy. The 5-year survival rate of patients with resected pancreatic adenocarinoma is approximately 20% [1]. The statistics for the 80 to 90 % of patients who present with locally advanced and metastatic pancreatic cancer are even more dismal. Rarely do such patients achieve a complete response to treatment; median survival is 5–10 months and 5-year survival is near zero [2].
There is very little randomised data on adjuvant therapy for pancreatic carcinoma. The ESPAC-1 trial assessed the role of adjuvant therapy in a randomized study. 548 patients were enrolled in Europe in this first large randomized trial. There was evidence that adjuvant chemotherapy brought survival benefits. The five-year survival rate was 21 percent among patients who received chemotherapy [3].
In 1995 a phase II trial was initiated by the Virginia Mason Clinic, Seattle, USA combining pancreaticoduodenectomy and adjuvant therapy with 5-FU, cisplatin, interferon alpha and radiation therapy. Picozzi et al. reported treatment and results in a series of 43 patients with high-risk resected pancreatic adenocarcinoma (84% node positive, 19% margin positive). After a median follow-up period of 32 months the 2-year survival rate was 64% and the 5-year survival rate was 55% [4]. The overall recurrence rate was 12% of which 80% occurred within 2 years after surgery [5].
Since the Virginia Mason group reported such encouraging results in improving survival periods, there has been considerable interest in gaining increased experience with this therapy combination in an effort to confirm results and evaluate the toxicity profile of a modified version of this regimen.
Radio-sensitising properties of 5-FU (5-fluorouracil) and cisplatin are well known. The incorporation of interferon alpha-2b into a combined modality treatment program seems to offer a number of theoretical advantages. These include: 1) the radio-sensitisation effects of interferon alpha-2b and 5-FU perhaps synergistically [6]; 2) enhanced 5-FU based bio-availability; 3) a synergistic inhibition of pyrimidine metabolism with 5-FU; and 4) an independent immunomodulatory effect of interferon alpha-2b [7]. 5-FU and interferon alpha-2b have been used together to advantage in several cancer settings, but not as part of a combined modality program [8]. Cisplatin also has a radio-sensitising effect and shares similar properties of cytotoxic synergy with interferon alpha-2b and 5-FU in both experimental and clinical cancer systems [9]. In a previous study we were able to demonstrate in vitro that interferon alpha has direct inhibitory properties and that it reduces the enhanced proliferation rate and VEGF secretion after a single treatment with cisplatin [10]. The use of interferon alpha-2b, 5-FU and cisplatin together thus might represent a kind of "combination radiosensitizer" analogous to combination chemotherapy potentially useful in the treatment of pancreaticobiliary cancers, especially from the standpoint of local control.
Design
Trial organization
CapRI has been designed by the Department of Surgery, University of Heidelberg. The trial is carried out by the National Centre for Tumourdiseases (NCT). The NCT is a multidisciplinary clinical and basic research group as well as comprehensive cancer centre of the Medical School of the University of Heidelberg that has been founded to improve care and research for cancer patients. The trial is sponsored by the Manfred-Lautenschläger-Foundation. The sponsor is not involved in the database management and has no access to the randomisation code.
Coordination
The trial is co-ordinated by the Surgical Department in cooperation with the National Centre for Tumourdiseases (NCT) at the University of Heidelberg. The Dept. of Surgery is responsible for overall trial management, trial registration (International Standard Randomised Controlled Trial Number (ISRCTN 62866759), ), database management, quality assurance including monitoring, reporting and for the scientific program of all trial related meetings.
Investigators
Patients will be recruited by the Department of Surgery at the University of Heidelberg. Due to the multi-modal nature of the trial, all investigators are experienced oncologists from the fields of hematology/oncology, radiation oncology and general surgery at the University of Heidelberg co-operating in this trial.
Adverse events committee
This committee consists of 3 independent physicians (medical oncologist, radiation oncologist and surgeon) and decides on the final diagnostic classification of critical clinical events. For all serious adverse events the documentation and relevant patient data are verified by the co-ordinating personnel before submitting the data to the Adverse Events Committee for diagnostic classification.
Analysis of safety related data is performed with respect to frequency of:
• Serious Adverse Events and Adverse Events stratified by body-system
• Adverse Events stratified by severity
• Adverse Events stratified by causality.
Patient toxicities will be assessed using the NCI Common Toxicity Criteria (CTC). Toxicity will be evaluated pretreatment, weekly during chemoradiation/ chemotherapy, prior to each course of infusional 5-FU and at follow-up. Unacceptable toxicity is defined as unpredictable, or irreversible Grade 4 toxicity. Decisions regarding weekly chemoradiation treatment and chemotherapy dose-adjustment will be made using the guidelines below and based on hematological parameters (ANC and platelets) monitored weekly during chemoradiation before each dose of cisplatin.
Medication supply
All chemotherapeutic and immunotherapeutic agents are prepared and provided by the pharmacy of the University Hospital Heidelberg. Medication will be prepared for each patient specifically and delivered just prior to administration to the NCT.
On-site monitoring
During recruitment of patients monitoring on site is performed according to good clinical practice (GCP) guidelines. The monitoring for this trial will be performed by a monitor from a Clinical Research Organisation (CRO) who is not involved in the trial or in completion of the case report form (CRF). The data management will be performed by an independent medical documentalist from the Institute for Medical Bioinformatics, University of Heidelberg. The medical monitoring will be done by two independent oncologists not involved in conducting this trial.
Ethics, informed consent and safety
The final protocol was approved by the ethics committee of the University of Heidelberg, Medical School (L-042/2003). This study complies with the Helsinki Declaration in its recent German version, the Medical Association's professional code of conduct, the principles of Good Clinical Practice (GCP) guidelines and the Federal Data Protection Act. The trial will also be carried out in keeping with local legal and regulatory requirements. The medical secrecy and the Federal Data Protection Act will be followed.
Written informed consent is obtained from each patient in oral and written form before inclusion in the trial and the nature, scope, and possible consequences of the trial have been explained by a physician. The investigator will not undertake any measures specifically required only for the clinical trial until valid consent has been obtained.
Patient selection
CapRI focuses on hospitalised patients over 18 years of age treated with pancreatic head resection for pancreatic adenocarcinoma during an 18-month period started in August 2004. Men and women over eighteen years of age with biopsy-proven completely resected (R0 or R1) pancreatic adenocarcinoma of the pancreatic head or uncinate process will be screened for participation in the study. A detailed overview of all eligibility criteria is given in Table 2.
Study design
The CapRI study is designed as an open, controlled, prospective, randomised potentially multi-centre trial meant to evaluate the post-operative overall survival of patients with pancreatic adenocarcinoma receiving chemo-radiotherapy including interferon alpha 2b administration compared with adjuvant chemotherapy. The treatment is offered to a heterogeneous group of people under clinical circumstances, covering a wide age range, for both sexes and with heterogeneous characteristics / co-morbidities.
One year after inclusion of the first patient an interim analysis will be performed. If at this time point there is already a significant difference between both groups favouring the Virginia Mason scheme (p < 0.001) group B (5-FU/Folinic acid) will be stopped and replaced by group C (chemoradiation without interferon alpha). This group will help to define the possibly deciding influence of interferon alpha on the effectiveness of the Virginia Mason treatment protocol. Addition of group C to the study would be the matter of a separate study amendment. The study design will not be changed prior to agreement of the ethics committee.
Study objectives
The primary objective is to compare the overall survival at two years postoperatively between two different methods of adjuvant treatment: therapy with 5-FU, cisplatin, interferon alpha 2b combined with radiation and the standard treatment from the ESPAC-1-Trial with 5-FU plus folinic acid.
Secondary objectives are to determine the role and the mechanism of interferon alpha 2b in patient's chemoradiation regimen, the toxicity, the disease-free interval and the quality of life. Different factors are tested in terms of their potential role as predictive markers.
Randomisation and standardised treatment scheme
A block-randomisation-list is generated via computer system (SAS Version 8.2, SAS Institute Inc., Cary, USA). The sealed randomisation list is stored in the investigator file. Patients are randomised using sealed opaque envelopes in the independent study centre at the Department of Surgery (Clinical Study Centre – "Klinisches Studienzentrum Chirurgie" – KSC) until informed consent is attained and diagnostic procedures rule out any contra-indication for participation in this trial.
After randomisation and pre-treatment evaluation treatment must begin within 12 weeks of surgery. A port catheter will be placed after informed consent and randomisation will be done within one to two weeks after pancreatoduodenectomy for patients in study arm A. These patients will be treated with 200 mg/m2/day 5-FU by continuous intravenous infusion at days 1–38.
Cisplatin – 30 mg/m2 (maximum single cisplatin dose of 60 mg) iv over 60 minutes on days 1, 8, 15, 22, 29, 36 (6 doses). Two to three hours before and after cisplatin administration the patients will receive hydration of at least 2 litres.
Interferon alpha-2b (Intron A) will be administered at a dose of 3 million units subcutaneously three times weekly (Monday, Wednesday, and Friday) for 17 doses. Non-steroidal anti-inflammatory drugs and steroids should be avoided if possible during Intron A treatment.
External beam radiation is to be given concurrently with chemotherapy with a total dose of 50.4 Gy in 28 fractions over 5.5 weeks (1.8 Gy/day). The pancreatic bed (i. e. resection margin) will be covered with a minimum margin of 2 cm. The hepatoduodenal ligament, origins of the celiac axis and superior mesenteric artery will be included. The AP/PA fields must include the entire duodenal C-loop as seen on the pre-operative CT scan. All patients will receive XRT per informed consent. A four-field technique with AP/PA and lateral fields and customised blocking is required. The beams will be weighted more heavily in the AP/PA fields (usually 2:1 compared to the lateral fields). The dose contribution from the lateral fields should be restricted to 20 Gy. Simulation should be done with the patient in the supine position with "arms up" position. A dosage greater than or equal to 10 MV photons should be used.
In study arm A the patients receive post-chemoradiation 5-FU infusions of 200/mg/m2 /day by continuous intravenous infusion on days 64–101 and 120–161. The timing of these courses of infusional 5-FU will be adjusted in patients who have treatment interruptions.
Patients in study arm B will be treated with 20 mg/m2 intravenous bolus injection of Folinic acid, D-L form, followed by 425 mg/m2/day intravenous bolus injection of 5-FU given on 5 consecutive days every 28 days for 6 cycles, i.e. 24 weeks.
The treatment protocol is outlined in figure 1.
Investigation schedule and follow-up
Pre-treatment evaluation for patients who are enrolled in study arm A includes a single low-dose (3 Mio U) injection of INTRON A prior to therapy (LDI). Blood will be drawn for intensive immunological studies. All patients (study arm A or B) must have appropriate lab and radiographic studies (CXR; CT abdomen [done post-operatively]; CBC; platelet count; BUN; creatinine; bilirubin, CA 19-9, and CEA) conducted prior to study enrolment to meet eligibility criteria.
During days 1–38 [study arm A] or during chemotherapy [study arm B] patients will be assessed with laboratory evaluation: complete blood count and blood chemistries weekly. Laboratory parameters in study arm A will be evaluated before each dose of cisplatin. Creatinine will be determined also one day after administration of cisplatin. Blood from members of study arm A will be investigated weekly for immunological markers.
Vital signs (blood pressure and pulse rate) and temperature are controlled daily during treatment. Patients are evaluated prior to receiving chemoradiation or chemotherapy. Patients enrolled in study arm A are evaluated weekly by the medical oncology and radiation oncology team during treatment. The team will check patients at each visit for symptoms due to therapy; a physical examination and complete safety labs should be performed. The patients' mental state will be investigated weekly. The questionnaire CES-D will be used as support. The three module questionnaire will be filled out during weeks 3 and 6 (study arm A) or week 3 and week 24 for study arm B.
During post-chemoradiation, infusional 5-FU (Study Arm A) patients will be evaluated by a physician prior to treatment and every 2 to 3 weeks with clinical assessment and laboratory parameters including a CBC, electrolytes, BUN, and creatinine. If patients undergo post chemoradiotherapy infusional 5-FU coordinated by an outside medical oncologist, a study nurse will contact the patient at least once weekly by telephone.
In the post-treatment period patients will be seen every 3 months by the surgical service for the first 2 years, every 4 months for the third year, and every 6 months during the 4th and 5th post-treatment years.
The aggregate clinical, laboratory, and imaging evaluations required per protocol as well as the timing of the optional three module questionnaire are outlined in table 1.
The follow-up will be continued for two years. Follow-up data of overall survival will be evaluated annually.
Assessment of quality of life
Measurement of quality of life is one of the secondary objectives of the trial. Overall survival, return to previous employment as well as persistence of symptoms, the ability to perform appropriate activities and to care for oneself are criteria applied in the three questionnaires used in this study.
EORTC QLQ-C30 is a general measure of qualitiy of life in cancer patients. It incorporates nine multi-item scales: five functional scales (physical, role, cognitive, emotional, and social); three symptom scales (fatigue, pain, and nausea and vomiting); and a global health and quality-of-life scale [17]. Specific symptoms (dyspnoea, insomnia, anorexia, constipation, diarrhoea, and financial impact) are measured as six single items. This instrument has been used extensively with a variety of cancer patients and was able to discriminate between individuals with metastatic and non-metastatic disease, as well as between patients at different stages of illness. The scale has good internal consistency (alpha > 0.70), and good test re-test reliability (0.80 to 0.90) [18].
To assess disease-specific symptoms for patients with pancreatic cancer the pancreatic specific module (QLQ-PAN26) [19] that has been designed to use long with the general measure is used in this study.
The CES-D [20] is a 20-item self-report measure of depression that emphasises the emotional dimension. Its emphasis on the affective components of depression and is preferable for use in medical populations. It has demonstrated high internal consistency in both the general population and in patient populations and convergent validity with other measures of depression [21].
Evaluation of the role of interferon alpha
In addition to the Virginia Mason and the MD Anderson study, an investigation of the effects and mechanism of INTRON A will be performed. IFN alpha seems to have the greatest activity of all interferons in malignancies [21-23] but its mode of action is poorly understood.
It probably consists of a combination of stimulation of cell-mediated cytotoxicity [24,25], direct antiproliferative anti-tumour activity, and an anti-angiogenic effect [15,26-31]. The known molecular and cellular effects of IFN-alpha appear to complement the mechanism of action of other therapies [28].
In view of the significant improvement in survival rates achieved by using interferon alpha in the Virginia Mason study, studies are now needed to define the molecular and immunologic mechanism(s) of this modality. If the mechanism of action of IFN-alpha is more clearly understood it can be applied more selectively and its therapeutic index will be enhanced.
Patients in study arm A at the CapRI trial will receive a single low-dose injection of 1 million units (LDI) prior to therapy. Accompanying immunological analysis should help to define predictive response markers. During therapy patients blood samples of approx. 20 ml will be drawn once a week for immunological studies. The panel will include immunophenotypings, detection of cytokine mRNA in lymphocytes, determination of cytokine patterns in peripheral blood, cellular cytotoxicity assays against autologous tumor cells and determination of apoptotic effects of interferon against autologous tumor cells.
Statistical considerations and sample size estimation
The primary endpoint in this study is the overall survival period, measured from the date of resection. The sample size calculation is based on the assumption of a constant monthly hazard rate of 0.044 in the standard group and a constant monthly hazard rate of 0.021 in the experimental group. The hazard values are derived from two-year survival rates in the study groups (using the formula λgroup = -1 × ln(P24,group)/24) with a two-year survival rate of 35% in the control group [11,12]and a conservative two-year survival rate of 60% in study arm A[13].
Assuming an accrual period of 18 months and a follow-up of 42 months [14,15], testing for a difference in hazard (hazard ratio ≠ 1) on level α = 0.05 and with a power of 80% a study sample size of 96 patients is needed. Taking into consideration the estimate of approximately 14 patients which will not complete the treatment, a total number of 110 patients should be randomised.
The overall survival will be summarised by Kaplan-Meier estimate and differences in therapy protocols will be analysed by univariate Cox-regression.
Various secondary endpoints will be evaluated in this study as well. Disease free survival, measured from date of operation, will be summarised by Kaplan-Meier estimate.
One year after inclusion of the first patient an interim analysis will be carried out which could result in stopping arm B. The planning of the study is based on a fix sample approach and not on groups sequential. Because the alpha level of the interim analysis is set to α = 0.001, the sample size calculated from a group sequential approach will be similar to the sample size of the fixed study approach.
The main evaluation will be performed two years after the last patient's enrolment.
There will be explicit stopping rules in place to terminate the trial early in the unlikely event that an unacceptably high rate of treatment related deaths (TRD) is observed. TRD will be monitored using the design of Thall and Simon [16]. A non-informative Beta prior distribution (i.e., B (0.015, 0.085) for TRD rate is assumed. The trial will be stopped if at any point during the trial there is a greater than 90% probability that the true TRD rate is greater than 0.05. Each patient will subsequently be evaluated and, an independent safety board will be consulted in making decision.
In view of the poor prognosis of the patient group, there will be no explicit stopping rules based on the overall number of toxicities, since even high rates of reversible toxicities seem acceptable if there is a large survival gain. Patients can withdraw from study participation at any time. Patients are taken off the study if unacceptable toxicity appears. Unacceptable toxicity is defined as unexpected serious side effects or irreversible Grade 4 toxicity. Patients who withdraw from the study may be treated with 5-FU and folinic acid or with gemcitabine. The decision will be based on the individual reasons for withdrawing from the study.
One year after inclusion of the first patient an interim analysis will be done. If at this time point there already is a significant difference between both groups favouring the Virginia Mason scheme (p < 0.001) group B (5-FU/Folinic acid) will be stopped and replaced by group C (chemoradiation without interferon alpha).
Discussion
The role of adjuvant therapy in potentially curatively resected adenocarcinoma of the pancreas remains a matter of debate. As the first, large multi-centre randomised controlled trial (RCT), the ESPAC-1-Trial clearly favoured adjuvant chemotherapy over postoperative chemoradiation [3]. However the quality of the radio-chemotherapeutic regimen in the ESPAC-1-Trial has been disputed. The Virgina Mason study group in Seattle, USA, published very promising data in phase-II-study involving immunotherapy and chemoradiation in the adjuvant setting [4]. The reliability of the data has been intensively discussed and there even a source-data-verification was performed by the National Cancer Institute (NCI). In the current setting with a reference adjuvant treatment from an RCT and very promising data from a phase-II-trial, there is the ideal basis for a controlled trial comparing the two most current and successful regimen. Being a centre focusing on pancreatic diseases and especially malignancies we therefore planned and conduct such a trial.
The CapRI study is an open, randomised controlled trial investigating the survival of patients after potentially curative resection of a pancreatic adenocarcinoma treated adjuvant with 5-FU, cisplatin, INTRON A combined with radiation or 5-FU plus folinic acid. The role and the mechanism of interferon alpha 2b in patient's chemoradiation regimen are evaluated. The toxicity, the disease-free interval and the quality of life are assessed. Different factors are tested for a potential role as predictive marker.
The results of the CapRI trial will definitely advance clinical and scientific knowledge on the adjuvant treatment of pancreatic carcinoma as it may confirm or de-mystify the remarkable results from the Virginia Mason study group.
Abbreviations
CCC Comprehensive Cancer Centre
CRF Case Report Form
CRO Clinical Research Organisation
GCP Good Clinical Practice
ICH International Conference on Harmonisation of Technical
Requirements for Registration of Pharmaceuticals for
Human Use
IEC Independent Ethics Committee
ITT Intention-to-treat-analysis
NCT National Centre for Tumourdiseases
RCT Randomised Controlled Trial
SDGC Study Centre of the German Surgical Society
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
The first three authors contributed equally. AM, JS, MWB planned, coordinated and conducted the study. Medical care is covered by KL, KH, HSW, SF, HT and TH. HPK and CS recruited patients and provided randomization. HG, VD and DJ took part in conducting the study. Scientific program is planned by AM and carried out by AM and KH. All authors read and approved the final manuscript.
Note
Table 1: Investigation schedule
Table 2: Eligibility Criteria
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The trial is supported by the Manfred-Lautenschläger-Foundation, Gaiberg, Germany.
Figures and Tables
Figure 1 CapRI treatment scheme.
==== Refs
Wagner M Redaelli C Lietz M Seiler CA Friess H Büchler MW Cuartive resection is the single most important factor determining outcome in patients with pancreatic adenocarcinoma Br J Surg 2004 91 586 594 15122610 10.1002/bjs.4484
Evans D Table 33.4–8, Principles and Practices of Oncology, 2000 1145 6 Lippincott, Williams and Wilkins, Inc
Neoptolemos JP Stocken DD Friess H Bassi C Dunn JA Hickey H Beger H Fernandez-Cruz L Dervenis C Lacaine F Falconi M Pederzoli P Pap A Spooner D Kerr DJ Buchler MW European Study Group for Pancreatic Cancer A randomized trial of chemoradiotherapy and chemotherapy after resection of pancreatic cancer N Engl J Med 2004 350 1200 10 15028824 10.1056/NEJMoa032295
Picozzi VJ Kozarek RA Traverso LW Interferon-based adjuvant chemoradiation therapy after pancreaticoduodenectomy for pancreatic adenocarcinoma Am J Surg 2003 185 476 80 12727570 10.1016/S0002-9610(03)00051-5
Picozzi VJ Traverso LW The Virginia Mason approach to localized pancreatic cancer Surg Oncol Clin N Am 2004 13 663 74 15350940 10.1016/j.soc.2004.06.002
Holsti LR Mattson K Niiranen A Standertskiold-Nordenstam CG Stenman S Sovijarvi A Cantell K Enhancement of radiation effects by alpha interferon in the treatment of small cell carcinoma of the lung Int J Radiat Oncol Biol Phys 1987 13 1161 6 3038802
Pazdur R Fluorouracil and recombinant interferon alpha-2a in advanced gastrointestinal neoplasms Br J Haematol 1991 79 56 9 1931711
Pazdur R Ajani JA Patt YZ Winn R Jackson D Shepard B DuBrow R Campos L Quaraishi M Faintuch J Phase II study of fluorouracil and recombinant interferon alpha-2a in previously untreated advanced colorectal carcinoma J Clin Oncol 1990 8 2027 31 2230894
Etienne MC Bernard S Fischel JL Formento P Gioanni J Santini J Demard F Schneider M Milano G Dose reduction without loss of efficacy for 5-fluorouracil and cisplatin combined with folinic acid. In vitro study on human head and neck carcinoma cell lines Br J Cancer 1991 63 372 7 2003979
Ma J Patrut E Schmidt J Knaebel H-P Buchler M-W Maerten A A In vitro analysis of the role of interferon-alpha in combined chemoradiationtherapy in the adjuvant treatment of pancreatic carcinoma (CapRI) World J Gastroenterol
Neoptolemos JP Dunn JA Stocken DD Almond J Link K Beger H Bassi C Falconi M Pederzoli P Dervenis C Fernandez-Cruz L Lacaine F Pap A Spooner D Kerr DJ Friess H Buchler MW Adjuvant chemoradiotherapy and chemotherapy in resectable pancreatic cancer: a randomised controlled trial Lancet 2001 358 1576 85 11716884 10.1016/S0140-6736(01)06651-X
Sohn TA Yeo CJ Cameron JL Koniaris L Kaushal S Abrams RA Sauter PK Coleman J Hruban RH Lillemoe KD Resected adenocarcinoma of the pancreas-616 patients: results, outcomes, and prognostic indicators J Gastrointest Surg 2000 4 567 79 11307091 10.1016/S1091-255X(00)80105-5
Picozzi V Adjuvant chemoradiation for resected pancreas cancer using alpha-interferon, 5-FU and cisplatin as radiosensitizers: update of a phase II trial Proc ASCO #1032 2000
Kirkwood J Cancer immunotherapy: the interferon-alpha experience Semin Oncol 2002 29 18 26 12068384
Xu D Erickson S Szeps M Gruber A Sangfelt O Einhorn S Pisa P Grander D Interferon alpha down-regulates telomerase reverse transcriptase and telomerase activity in human malignant and nonmalignant hematopoietic cells, Blood 2000 96 4313 8 11110707
Thall PF Simon R Practical Bayesian guidelines for phase IIB clinical trials Biometrics 1994 50 337 49 7980801
Aaronson NK Ahmedzai S Bergman B Bullinger M Cull A Duez NJ Filiberti A Flechtner H Fleishman SB de Haes J C The European Organization for Research and Treatment of Cancer QLQ-C30: a quality-of-life instrument for use in international clinical trials in oncology J Natl Cancer Inst 1993 85 365 76 8433390
Hjermstad MJ Fossa SD Bjordal K Kaasa S Test/retest study of the European Organization for Research and Treatment of Cancer Core Quality-of-Life Questionnaire J Clin Oncol 1995 13 1249 54 7738629
Fitzsimmons D Johnson CD George S Payne S Sandberg AA Bassi C Beger HG Birk D Buchler MW Dervenis C Fernandez Cruz L Friess H Grahm AL Jeekel J Laugier R Meyer D Singer MW Tihanyi T Development of a disease specific quality of life (QoL) questionnaire module to supplement the EORTC core cancer QoL questionnaire, the QLQ-C30 in patients with pancreatic cancer EORTC Study Group on Quality of Life Eur J Cancer 1999 35 939 41 10533475 10.1016/S0959-8049(99)00047-7
Radloff L The CES-D scale A new self report depression scale for research in the general population. Applied Psychological Measurement 1:385-401 The CES-D scale: A new self report depression scale for research in the general population Applied Psychological Measurement 1977 1 385 401
Kirkwood J Pharmacology of cancer biotherapeutics: Interferons Cancer: Principles and Practices of Oncology 2000 6 Philadelphia, PA: Lippincott Williams & Wilkins 461 471
Kerr C Adjuvant interferon-alpha improves skin cancer survival Lancet Oncol 2002 3 262 12067791 10.1016/S1470-2045(02)00746-5
Soori G Dillman RO Wiemann MC Stark JJ Tai F DePriest CB Church CK Schulof R Phase II trial of subcutaneous interleukin-2, subcutaneous interferon-alpha, 5-fluorouracil and cis-retinoic acid in the treatment of renal cell carcinoma: final results of cancer biotherapy research group 94-10 Cancer Biother Radiopharm 2002 17 165 73 12030110 10.1089/108497802753773784
Chen X Regn S Raffegerst S Kolb HJ Roskrow M Interferon alpha in combination with GM-CSF induces the differentiation of leukaemic antigen-presenting cells that have the capacity to stimulate a specific anti-leukaemic cytotoxic T-cell response from patients with chronic myeloid leukaemia Br J Haematol 2000 111 596 607 11122108 10.1046/j.1365-2141.2000.02361.x
Grimm EA Smid CM Lee JJ Tseng CH Eton O Buzaid AC Unexpected cytokines in serum of malignant melanoma patients during sequential biochemotherapy Clin Cancer Res 2000 6 3895 903 11051235
Kirkwood J Cancer immunotherapy: the interferon-alpha experience Semin Oncol 2002 29 18 26 12068384
Nathan PD Eisen TG The biological treatment of renal-cell carcinoma and melanoma Lancet Oncol 2002 3 89 96 11902528 10.1016/S1470-2045(02)00650-2
Eguchi H Nagano H Yamamoto H Miyamoto A Kondo M Dono K Nakamori S Umeshita K Sakon M Monden M Augmentation of antitumor activity of 5-fluorouracil by interferon alpha is associated with up-regulation of p27Kip1 in human hepatocellular carcinoma cells Clin Cancer Res 2000 6 2881 90 10914738
Grander D How does interferon-alpha exert its antitumour activity in multiple myeloma? Acta Oncol 2000 39 801 5 11145436 10.1080/028418600750063532
Fishman M Seigne J Immunotherapy of metastatic renal cell cancer, Cancer Control 2002 9 293 304 12228755
Thyrell L Erickson S Zhivotovsky B Pokrovskaja K Sangfelt O Castro J Einhorn S Grander D Mechanisms of Interferon-alpha induced apoptosis in malignant cells Oncogene 2002 21 1251 62 11850845 10.1038/sj.onc.1205179
| 15826316 | PMC1087479 | CC BY | 2021-01-04 16:03:04 | no | BMC Cancer. 2005 Apr 12; 5:37 | utf-8 | BMC Cancer | 2,005 | 10.1186/1471-2407-5-37 | oa_comm |
==== Front
BMC Cell BiolBMC Cell Biology1471-2121BioMed Central London 1471-2121-6-161580789710.1186/1471-2121-6-16Research ArticleThe C-terminal subunit of artificially truncated human cathepsin B mediates its nuclear targeting and contributes to cell viability Bestvater Felix [email protected] Claudia [email protected] Eberhard [email protected] Deutsches Krebsforschungszentrum, PO Box 101949, D-69009 Heidelberg, Germany2005 4 4 2005 6 16 16 4 11 2004 4 4 2005 Copyright © 2005 Bestvater et al; licensee BioMed Central Ltd.2005Bestvater 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
Splicing variants of human cathepsinB primary transcripts (CB(-2,3)) result in an expression product product which lacks the signal peptide and parts of the propeptide. This naturally truncated Δ51CB is thus unable to follow the regular CB processing and sorting pathway. It is addressed to the mitochondria through an activated N-terminal mitochondrial targeting signal instead. Although Δ51CB is supposed to be devoid of the typical CB enzymatic activity, it might play a role in malignancies and trigger cell death/apoptosis independent from the function of the regular enzyme. Cytoplasmic presence of the mature CB might occur as a result of lysosomal damage.
Results
We investigated such "aberrant" proteins by artificial CB-GFP chimeras covering various sequence parts in respect to their enzymatic activity, their localization in different cell types, and the effects on the cell viability. Unlike the entire full length CB form, the artificial single chain form was not processed and did not reveal typical enzymatic CB activity during transient overexpression in large cell lung carcinoma cells. Δ51CB was found predominantly in mitochondria. In contrast, the shorter artificial CB constructs localized in the cytoplasm, inside the cell nucleus, and in the midbodies of dividing cells. Bleaching experiments revealed both mobile and immobile fractions of these constructs in the nucleus. Nuclear accumulation of artificially truncated CB variants led to disintegration of nuclei, followed by cell death.
Conclusion
We propose that cell death associated with CB is not necessarily triggered by its regular enzymatic activity but alternatively by a yet unknown activity profile of truncated CB. Cytoplasmic CB might be able to enter the cell nucleus. According to a mutational analysis, the part of CB that mediates its nuclear import is a signal patch within its heavy chain domain. The results suggest that besides the N-terminal signal peptide also other CB domains contain patterns which are responsible for a differentiated targeting of the molecule, e.g. to the mitochondria, to the nucleus, or to vesicles. We propose a hierarchy of targeting signals depending on their strength and availability. This implies other possible transport mechanisms besides the usual trafficking via the mannose-6-℗ pathway.
==== Body
Background
Lysosomal cysteine peptidases play an important part in intra- and extracellular protein degradation. Their primarily assumed function has changed: they "can no longer be considered as simple garbage disposers" [1], but do also function as key enzymes in cardinal processes of homeostasis and cell demise. This is particularly valid for the ubiquitous peptidase cathepsinB (CB, E.C.3.4.22.1). In higher organisms, this enzyme is present and active in almost all tissue types. For a long time it was therefore considered as an unspecifically degrading peptidase. Research of recent years has brought up specificity [2] and its implication in pathologic processes as arthritis [3] or cancer [4-6]. Furthermore, these investigations have revealed the pivotal role of CB in a number of apoptotic pathways [7-18].
The human CB gene is composed of 12 or 14 exons [19,20] (Fig 1A, top panel); its promoter is assumed to be regulative [21,22]. The regular mRNA population encodes a 48 kDa polypeptide which contains pre- (signal), pro-, and two functional domains (CB(FLM); Fig 1A, bottom panel). The signal peptide and glycosylated residues target the protein via ER and Golgi into the lysosomes by the mannose-6-℗ pathway. During this process a 31 kDa single chain or 25/5 kDa double chain glycosylated polypeptides are generated. Both forms exhibit enzymatic activity, albeit with different efficiency [19].
Figure 1 Expression and processing of CB. A. Chromosomal location (adapted from NCBI/NIH) and exon-intron organization of the hCB gene (top panel, modified from [19]), alternative splicing variants of CB primary transcripts (centre panel, modified from [56]), and domain organization of the entire translation product (bottom panel). Lightened sectors: non-coding exons (5'-UTR and 3'-UTR); darkened sectors: translated CB regions. The transcript population CB(-2) encodes the entire CB sequence, the alternative transcript population CB(-2, 3) leads to the truncated translation product Δ51CB which lacks the complete signal sequence (pre) as well as parts of the proregion (N'pro). In vivo, the single chain form of the native CB is mainly split into the light and the heavy chain while the C'pro-region is cut off during secretion. B. Processing of intrinsic CB (C, control) and of transiently overexpressed Δ72CB-EGFP (1) and CB(SC)-EGFP (2) in LCLC-103H cells (on the left: α-CB-Ab; on the right: α-GFP-Ab). The immunoblot reveals distinct native CB fragments at similar expression levels of ~31 kDa (single chain form), ~26 kDa (heavy chain), and several isoforms thereof indicated by weaker adjacent bands. Unspecific low molecular mass protein bands below signify degradation products. Additional expression products of ~62 kDa (α-CB, α-GFP) in the samples from the transfected cells represent entire unprocessed sequences. The Δ72CB-EGFP band runs slightly slower than the CB(SC)-EGFP band. Weak bands at ~22 kDa (α-GFP) indicate that a small amount of the EGFP tag is removed from the fusion proteins.
A number of mRNA variants may be generated by gene splicing (exon skipping) (Fig 1A, centre panel). The regulation of the splicing process remains unclear. All splicing variants might be expressed concomitantly [23]. They can be subdivided into two subpopulations which result in two distinct translation products [19,20,23,24]. The first species lacks exon2 (CB(-2)), which does not affect translated regions of the entire CB and appears to be a more easily transcribed message [19]. The second one lacks exons2 and 3 (CB(-2,3)). As a result of an additional initiation codon at position 53 within exon4, this message can give rise to the naturally truncated translation product Δ51CB (the original term from the literature is used here). Recent publications prove that Δ51CB has no regular CB enzymatic activity [25,26]. Such truncations are not exceptional among cysteine peptidases. Recently, there have been discovered other truncated cathepsins like a cathepsinH progeny lacking parts of the signal peptide with unknown functions and differentiated intracellular distribution [27] and a truncated form of cathepsinL devoid of the signal peptide with nuclear targeting and a specific cleaving activity [28]. According to present knowledge and interpretations, the truncated message of CB is linked to pathological findings, in so far as the respective expression product was found more prominently expressed in tumours [24] or arthritic tissues/cells [23]. However, it is not clear whether it promotes malignancies or is a means of defence.
Although the detection of CB outside the lysosomes was primarily considered atypical, its extracellular and plasma membrane associated appearance is now well documented [4,29]. Release of functional CB from lysosomes is connected with apoptosis [13,18]. It is uncertain, whether this can be attributed to a particular fraction of the enzyme.
Earlier on, CB was also found to be associated with the nucleus by immuno- and enzyme cytochemistry [30,31] and by biochemical methods [16,32,33]. The truncated Δ51CB was identified in association with the cytoplasmic side of the nuclear envelope [24]. Nuclear fractions of CB could be linked with cell death [7,12,14,33,34]. It remains unclear, whether they represent the mature CB released from the lysosomes or the truncated Δ51CB. However, detection methods based on polyclonal antibodies do not discriminate between different variants of CB, just as little as biochemical assays of cellular organelles do between inside and attached (outside) activity. Therefore, the results mentioned above should be considered with care. Both constraints can be overcome by living cell microscopy of genetically labelled samples. Δ51CB tagged by GFP appears accumulated in mitochondria; its overexpression provokes nuclear fragmentation and cell death [36]. Mitochondrial localization is not surprising since the first 20 amino acids of the residual proregion contain a mitochondrial targeting signal (MTS) [25]. These results point out a specific role of Δ51CB in cell death pathways besides those, for which the lysosome-released canonical forms of the peptidase might be responsible.
The aim of our studies was to characterize cytoplasmic CB forms lacking the ER signal peptide within living cells. Basically, there are two situations in which cytoplasmic CB can occur in vivo: (i) as a mature message product resulting from lysosomal leakage [37-40] and (ii) as naturally truncated Δ51CB resulting from alternative splicing [19,23-25,36]. We investigated the intracellular transport of various recombinant artificial CB forms in living cells by GFP labelling and by spectroscopic methods. Living cell fluorescence microscopy now offers a wide range of techniques to collect information about cellular processes and will surpass indirect labelling and studies on fixed material. Our findings support a nuclear targeting of CB produced from incomplete messages and they hence partly go together with the data mentioned before [24,25,36]. We could further elucidate the nuclear transport and the stability of nuclear binding by mutational analysis and light microscopy. Finally, we found a link to cell death for the truncated polypeptides synthetically generated from incomplete messages.
Results
Posttranslational cleavage and enzymatic activity
Artificially truncated CB constructs such as Δ72CB or the single chain form of CB, CB(SC), were obtained from an established human cDNA full-length message (FLM) by PCR, cloned into a eukaryotic expression vector, and tagged by distinct fluorescent protein (FP) variants. These genetic chimeras were then used for mutational analysis, localization, mobility, and functional studies.
Light and heavy chain (LC and HC) are linked by two residues, which are removed in vivo during the zymogene processing. Δ72CB resulted from an internal KpnI-site upstream the CB(SC) sequence and revealed a fluorescent expression product in cells implying a regular in-frame translation. The expected alternative translation origin lies seven residues before the first codon of CB(SC) at the M73 position. The polypeptide that would result from this message lacks the complete signal peptide and that part of the N-terminal proregion which encodes the MTS. It is the most similar in size to the 21 amino acids longer native Δ51CB encoded by the CB(-2,3) transcript [19] in respect to all the other constructs used.
Neither Δ72CB-EGFP nor CB(SC)-EGFP was posttranslationally cleaved in LCLC-103H cells during their transient expression as specified by SDS-PAGE and Western Blot analysis (Fig 1B). In addition, only a small fraction of the C-terminal fluorescent protein tag appeared to be cleaved off confirming the high stability of the constructs, which was important for the further studies. Both had similar apparent molecular masses of ~62 kDa. The native CB revealed a typical banding pattern primarily representing the single and the heavy chain forms. The amount of native CB remained unaffected by the transient overexpression of the recombinant products.
Both in vivo and in vitro studies did not reveal any CB specific enzymatic activity during transient overexpression of CB(SC) variants. In contrast, permanently expressed CB(FLM)-FP was properly processed and exhibited significantly elevated levels of enzymatic activity: In LCLC-103H-cells we measured 250 μEU/(μg protein). The cell clone with the highest CB(FLM)-FP expression revealed 2570 μEU/(μg protein). Nuclei were isolated from this cell clone and non-transfected control cells. We found 1150 and 8.4 μEU/(μg protein), respectively. Microscopy revealed that this activity is located in the nuclear envelope.
Naturally truncated Δ51CB
The intracellular localization of CB was first studied by a CB(FLM)-EGFP construct (Fig 2). As expected, the expression product was directed into the ER and was mainly found within the Golgi network and the vesicles. Furthermore, we investigated the alternative splicing variant Δ51CB tagged by EGFP [25,36]. Expression of this construct in the LCLC-103H cells resulted in strong fluorescence signals consistent with mitochondrial staining (Fig. 3). Their intensity was well above that of the cytosolic background. However, also nuclei exhibited fluorescence, whereas nucleoli were devoid of signal. Quantification of the fluorescence revealed the following order of intensity: mitochondria > nucleus > cytoplasm. The signal ratios are illustrated by an intensity profile of a ROI across an expressing cell (Fig. 3, inset).
Figure 2 Intracellular localization of recombinant CB(FLM). LCLC-103H cells transiently expressing CB(FLM)-EGFP revealed a strongly developed reticular signal distribution. Besides the ER, the nuclear envelope, the Golgi apparatus, and perinuclear granules were labelled (A: phase contrast; B: fluorescence channel). Obj. 40×/1.30 Oil; post deconvolution.
Figure 3 Intracellular localization of recombinant Δ51CB. LCLC-103H cells with transient Δ51CB-EGFP expression revealed an intensive mitochondrial distribution of the fluorescence signals. Beside a slight cytoplasmic background there was also a signal level in the nucleoplasm, which reached ~40% of the mitochondrial fluorescence intensity. The nucleoli were lacking any fluorescence. A profile cut across the cell elucidates the intensity proportions. The arrow tips mark the nuclear boundary. The intensity gap within this region represents a transected nucleolus. For deconvolution, 23 optical sections were used, which have been acquired with a distance of 200 nm. WFM; obj. 63×/1.4 Oil.
Intracellular localization of the single chain form
A number of established cell lines and freshly isolated endothelial cells were transfected with the chimeric single chain form CB(SC)-EGFP. The transient expression was analysed by fluorescence microscopy the following days. Pure EGFP, which lacks any targeting sequences, was used as control. In all cases, the artificially truncated CB without its signal peptide localized significantly different from the entire CB(FLM)-EGFP construct and Δ51CB.
The CB(SC)-EGFP construct appeared in the cytoplasm, but not in the ER or the Golgi of LCLC-103H cells (Fig 4A, B). Apart from that, it was enriched in perinuclear granules and prominently in cell nuclei showing either a homogeneous or a patterned distribution. The nucleoli showed a speckled staining pattern (Fig 4G, H). Fluorescence was also detected in the matrix of the midbody (Fleming-body) which is a persistent remnant of the spindle apparatus of mitotic cells and contains nuclear material among others (Fig 4C–G).
Figure 4 Intracellular localization of recombinant CB(SC). Transient overexpression in LCLC-103H cells visualized by WFM (A-F) or OPM (G, H). A, B. Cells expressing CB(SC)-EGFP (green) were counterstained with Hoechst33342 (blue) and LysoTracker Red (red). Diffuse, vesicular, and granular EGFP fluorescence signals were found in the cytoplasm and highly enriched inside the nucleus sparing out the nucleoli as indicated by the DNA staining. The reticular and vesicular staining of the lysosomal marker adjacent to the nuclear indentation did not overlap with EGFP signals. Obj. 40×/1.30 Oil. Processing of recombinant CB(SC) and influence of the fluorescent protein marker or the total molecule size on its localization were proven by differential tagging (N- or C-terminus, respectively) as well as by means of immunocytochemistry using a N-terminal myc-epitope. C, D. Double-tagged ECFP-CB(SC)-EYFP was distributed mainly in the nucleus analogously to CB(SC), which was marked at its C-terminus only. Accumulates were found within the nucleus and adjacent to it. Fluorescence also appeared in the ring shaped midbody matrix (enlarged region in the upper right corner). Obj. 40×/0.60; processed by deconvolution. E, F. Cells coexpressing myc-CB(SC) and CB(SC)-EGFP were fixed by acetone/methanol and immunostained against myc and GFP. A tight colocalization of both was found (E and F); the constructs were found mainly in the nucleus and stained the midbody (marked by the arrowhead; see inset). Obj. 40×/0.60. G, H. Optical sections of CB(SC)-EYFP expressing cells were subjected to spatial reconstruction (G: 3D-visualisation; H: orthogonal projection). Isosurfaces obtained by arbitrary fluorescence intensity thresholds represent distinct compartments (granules: opaque; nucleus and midbody: transparent). In the given cellular state, only a weak expression in the cytoplasm is found. The main signals arise from granular inclusions within the nucleus including distinct regions inside the nucleoli (marked by dashed lines in H) as well as from enrichment in the midbody (arrow). Obj. 63×/1.32 Oil. ROI: 72 × 59 × 11 μm3 (= 464 × 380 × 25 voxels).
Transient transfection of other cell lines than LCLC-103H resulted in similar localization of CB(SC)
HeLa cells and Hep-G2 hepatocytes produced higher numbers of perinuclear granules. Wi-38 fibroblasts showed vesicular structures even in peripheral regions of the cells (pseudopodia), but lower accumulations in the nuclei. The fluorescence distribution was strongly associated with the nucleus in MDCK cells. The localization in the COS-7 cells was more homogeneous in the cytoplasm with no particular accumulations and only low signals in the nucleus. Primary endothelial HDMEC contained intensive, punctual signals in the cytoplasm; nuclei were not involved. Hence, with the exception of HDMEC cells all cell types investigated revealed no principle qualitative differences. The absence of nuclear fluorescence in the primary endothelial cells might depend on a retarded or dysfunctional transport mechanism.
The nuclear accumulation of the CB(SC)-EGFP polypeptide, which is consistent for different cell types from various species, suggests the presence of a nuclear targeting mechanism. We did not find any common nuclear localization signal (NLS) within the CB sequence by appropriate computational analysis (PSORT v6.4, © K. NAKAI, University of Tokyo, Japan). In case of a signal patch, the abundance of nuclear targeting would ask for a specific folding of the polypeptide exposing the patch on its surface. The conformation of the CB(SC)-EGFP polypeptide is difficult to predict. In contrast to CB(FLM)-EGFP, which is folded cotranslationally within the ER, it must be determined by the conditions prevailing in the cytoplasm. To exclude a possible impact of the fluorescent protein sequence on the localization and to prove the integrity of the single chain form (in case of degradation or processing, respectively), further constructs were produced: (i) FP-CB(SC)-FP is flanked by two distinct FPs, (ii) myc-CB(SC)-FP includes a N-terminal myc-Ab epitope (EQKLISEEDL), and (iii) myc-CB(SC) contains this N-terminal myc tag only. FP-CB(SC)-FP showed the same distribution as the single-tagged construct in both fluorescence channels (Fig 4C, D). Application of α-myc and α-GFP Abs in immunocytochemical experiments to fixed LCLC-103H cells verified a strong colocalization of both signals. The product of myc-CB(SC) colocalized with CB(SC)-EGFP in the nucleus, in cytoplasmic granules, as well as in the midbody (Fig 4E, F). The dual-marker-construct myc-CB(SC)-EGFP revealed the same signal overlap. These results suggest that in most cases (i) neither the type (ii) nor the size of the tag (iii) nor the tagging site do exhibit significant impact onto the structure and the localization of CB(SC). In addition, (iv) degradation or processing of the CB single chain form into light and heavy chains under transient expression conditions is unlikely. This was already exposed by the Western Blot analysis (Fig 1B).
Nuclear binding studies
The mobility of GFP-tagged CB(SC) was studied by in vivo photobleaching experiments. Two-photon laser scanning microscopy (TPM) was applied to register loss of fluorescence through continuous irradiation or fluorescence recovery after photobleaching (FRAP), respectively. TPM confers less damage to the fluorochrome as well as to cells and thus prolongs the effective observation time. For comparative purposes, EGFP and two EGFP chimeras – the ribosomal transcription initiation factor TIF1A and the histone H2A – were taken as controls. Their diffusion and binding characteristics can be deduced from their known localization and functional properties. All controls appear in the nucleus to some extent and represent different types of nuclear binding: EGFP is known to have no binding capacity and to diffuse freely [41]. The transcription factor has a mobile (nucleoplasm) and an immobile (nucleoli) fraction. Histones are tightly bound in the nucleus [42]. Constitutively (H2A-EGFP) or transiently (other constructs) expressing LCLC-103H cells were examined. The obtained serial scans were evaluated in respect to their mean grey values by automated image processing routines.
Continuous photobleaching
A nuclear region was illuminated repeatedly and the fluorescence depletion was monitored simultaneously within this ROI in a series of images and evaluated as described in the methods section. The measurements are compiled in (Fig 5B). Decrease of EGFP fluorescence was low, after 1 min of irradiation only 10% of the original intensity have disappeared. This indicates the absence of an immobile fraction. The bleaching characteristics of TIF1A-EGFP were comparable to those of EGFP except of its drastic depletion to a third of its original value. In contrast, the H2A-EGFP curve shows an extremely steep initial decrease – within the first ~10s the equilibrium is reached. This finding suggests that most of the GFP-tagged histone is immobile; a supplementary pool of mobile H2A-EGFP does exist, though. The plots for both CB(SC) constructs (CB(SC)-EGFP and ECFP-CB(SC)-EYFP) show a strong initial loss and lie between the two control constructs EGFP and H2A-EGFP. Hence, we conclude that a significant fraction is not diffusing freely, but is in an associated state.
Figure 5 Nuclear diffusion and binding of recombinant CB(SC). Fluorescence was extinguished by TPM within distinct nuclear regions and the diffusion and binding characteristics of GFP-tagged constructs were determined. A. The approach is described by means of the GFP-tagged histone-construct H2A-EGFP. In the continuous photobleaching experiment a 2 × 2 μm2 nuclear region was scanned consecutively and the loss of fluorescence caused by the irradiation was monitored simultaneously (see enlarged section). In the FRAP experiment, a region of same dimensions was bleached by continuous irradiation. A time series was grabbed subsequentially from a larger detail of the nucleus (first and last scan are shown). The fluorescence within the bleached region as well as in an untreated control region was measured and normalized according to equation (1). B. Continuous bleaching curves of CB(SC)-EGFP, ECFP-CB(SC)-EYFP as well as of the control proteins EGFP, H2A-EGFP, and TIF1A-EGFP (n = 2 or 3) are plotted as a function of time. The fitting function is composed of two partial terms and matches the values sufficiently enough and more precise than a simple exponential function. The term meets the fact that there are both bound and freely diffusing fluorochrome labelled fractions. The first subterm describes the bleaching of the bound and the second one the bleaching of the diffusible component. While the graphs for EGFP and TIF1A-EGFP support free diffusion, the H2A-EGFP-population exists mostly in a bound state. Both CB(SC) constructs show an intermediate behaviour which points to bound as well as mobile fractions. C. FRAP curves of the same set of fusion proteins corroborate the findings above.
FRAP
The recovery of fluorescence after complete photobleaching within a ROI of the sample was analysed in analogy to the continuous bleaching experiment (Fig 5C). Back diffusion of fluorochrome into the bleached region was monitored continuously. Minor differences in the starting points are inherent to the procedure. The recovery of both EGFP and TIF1A-EGFP intensity was complete within milliseconds. Whereas only mobile EGFP was observed, there was a very small fraction (~3%) of immobile TIF1A-EGFP as indicated by its lower plateau. In contrast, the fluorescence of the tightly bound histone-EGFP chimera did hardly recover and revealed an immobile fraction comprising at least 60% of the material. Again, both CB(SC) variants showed an intermediate behaviour in respect to the controls which confirmed the notion of their binding to nuclear components. The immobile fractions were low and diffusion occurred retarded in both cases. In spite of the higher molecular mass, the double-tagged CB(SC) diffused faster than the single tagged variant. This can be explained by conformational changes in the CB protein caused by the additional N-terminal fluorescent protein, which might affect affinity characteristics. Furthermore, one has to consider the different bleaching characteristics of EGFP (~8% of original intensity) and ECFP/EYFP (~30% of original intensity), which could distort the results.
Mutational analysis
The accumulation of CB(SC) in the nucleus was fatal to the cells as will be shown later. It is therefore interesting to find out which parts of the protein are involved in the nuclear localization. We tackled this subject by a genetic mutational analysis of the single chain CB, CB(SC) (Fig 6). For comparative localization studies, the mutated constructs were tagged by distinct GFP variants, which differ in spectral characteristics (EGFP, EYFP, ECFP, and EBFP). The number and the site of tagging were also varied. The findings are compiled in Fig 6; Figs 3 and 7 illustrate the intracellular localization of the most relevant examples.
Figure 6 Mutational analysis. Besides the recombinant GFP-chimeras of the entire CB zymogene (top panel) and the splicing variant Δ51CB (#0), the mutated artificial CB-constructs used in this study (below) and their intracellular localization in LCLC-103H cells (right) are compiled. The active site amino acids in the FLM sequence are indicated by red asterisks. Deleted regions are represented by lines connecting the expressed regions (bars). The N-terminal fluorescent protein tags are displayed in a spliced way. HC: heavy chain; LC: light chain; Δ: deletion; *: site of amino acid exchange; Np: nucleoplasm; Mi: mitochondria; Mb: midbody; CG: cytoplasmic granules; NG: nuclear granules.
Figure 7 Localization of mutated CB-FP chimaeras in transiently expressing LCLC-103H cells. Individual expression with corresponding constructs (above: EYFP-tagged; below: ECFP-tagged) and false colour superposition of both fluorescence channels (YFP: red; CFP: green) are shown. Double-transfection of Δ72CB (A) or CB([C211_I243del]SC) (B) with CB(SC), respectively, resulted in a complete colocalization of both constructs. The heavy chain domain still colocalized with the single form to a large extent (C). Compared with the heavy chain, the light chain of CB was distributed homogeneously throughout the cell as pure GFP (D). In contrast to the CB(HCN') constructs, CB(HCC') expressed hotspots around the Golgi and inside the nucleus and stained the midbody (E). Transient coexpression of CB(SC) and its increasingly truncated C-terminal fragments (F-J) revealed decreased nuclear fluorescence with smaller fragment size. Already in case of CB(HCC'1) (F), the tendency to accumulate was reduced considerably and the accumulations disappeared completely in case of CB(HCC'4) transfection (I). The latter showed an almost complete homogeneous signal distribution without any accumulation in the nucleus, the granules, or the midbody, respectively. The elimination of 9 residues in the C-terminal part of CB (Q268_G277del) had no significant influence on the localization; the construct still colocalized completely with CB(SC) (J; the inset is a superposition of the midbody fluorescence with the corresponding phase contrast image). A-C, G-I. OPM; λEx.(ECFP): 430 nm; λEx.(EYFP): 500 nm; obj. 60×/1.2 Plan Apo water. D-F, J. WFM, in part after deconvolution (D, J); obj. 63×/1.25 Oil (D, F, J); obj. 40×/0.60 (E); bars 10 μm.
A construct encoding the natural Δ51CB (#0) localized to mitochondria and nuclei. Two constructs originate from the sequence inherent restriction sites: A KpnI deletion led to (i) a seven residues longer construct in respect to the CB(SC) (#3, 4) and contained parts of the N-terminal propeptide (Δ72CB, #1, 2). (ii) The BglII-generated mutant CB([C211_I243del]SC) (#8) lacked 33 residues, which contain a possible disulfide bridge. In the intact polypeptide, this sequence connects the two globular units of the heavy chain. Both mutants localized in cells exactly as CB(SC). The deletion of the 6 residues long C-terminal propeptide in Δ72CB (compare #1 and #2) did not exhibit any influence onto the signal distribution. This corresponds to other observations [36]. As constructs containing or lacking the C-terminal propeptide localized identically, we conclude that the chimeras were not processed at this site and that these results are reliable.
In the absence of a canonical NLS, one might suppose that conformation plays the essential part for the localization of artificially truncated CB. Our efforts were therefore concentrated on globular protein domains the destruction or deletion of which should less affect the folding of the remaining polypeptide. Two GFP-tagged constructs, which correspond to the heavy and the light chain of CB, were produced. While both CB(LC) constructs (#9, 10) revealed a diffuse distribution in LCLC-103H cells comparable to that of pure EGFP, the single-tagged CB(HC) construct (#11) was distributed identically to CB(SC) in the cytoplasm, the nucleoplasm, the granules, and the midbody. This leads to the assumption that the nuclear localization patch really exists and that it is confined within the heavy chain. Double-tagged constructs led to a less differentiated signal distribution in some cases (e.g. #12, 16).
To further narrow down the region of interest, the heavy chain sequence was subdivided into two parts of comparable size CB(HCN') (V129_N228; #13, 14) and CB(HCC') (N228_D333; #15, 16), which encode globular units in the native protein (Fig 8). CB(HCN') was found in the cytoplasm and in low amounts in the nucleus. In contrast, CB(HCC') was highly enriched in the nucleoplasm including granules and also in cytoplasmic granules as well as in the midbodies.
Figure 8 Spatial representation of the mutated sections. The positions of key fragments are indicated within the native mature CB protein. These are the globular domains of the light chain, the heavy chain, and two subunits CB(HCN') and CB(HCC'). The figure serves for illustrational purposes only and does not represent the authentic conformation of the fragments.
Consequently, the CB(HCC') sequence was reduced equidistantly towards its C-terminus and the following constructs were obtained: CB(HCC'1) (E242_D333; #17), CB(HCC'2) (V255_D333; #18), CB(HCC'3) (Q268_D333; #19), CB(HCC'4) (R281_D333; #20). With lower fragment size, the constructs were located more and more homogeneously within the cytoplasm and the nucleus, the typical granules no longer appeared and fewer signals were enriched in the midbody. Finally, the only 52 residues long CB(HCC'4) fragment appeared almost as diffuse as EGFP.
The exposed and highly mobile residue E273, which is located within the restricted region, raised the question whether it is possibly involved in the nuclear localization of CB. We proved this suspicion by specific point mutations: In CB([E273L]SC) (#21), the acidic and polar glutamic acid is exchanged against the neutral and non-polar leucine; in CB([E273del]SC) (#22), the glutamic acid is eliminated. Finally, in CB([Q268_G277del]SC) (#23), the surrounding residues forming a loop are deleted completely. Surprisingly, neither the point mutation nor the point deletion nor the excision of the five residues to each side of E273 produced a significant change in the localization. In contrast, the elimination of the adjacent C-terminal region in CB([L80_L283]SC) (#24) resulted in a considerable decrease of nuclear signal.
These mutational studies suggest that a region within the C-terminal sequence N228_D333 has a great impact on the nuclear localization of CB(SC). The results favour the assumption that an epitope is generated by a spatial arrangement of the respective, yet not defined residues.
Mechanisms of nuclear import
Besides an active import of the CB(SC) polypeptide into the nucleus, a passive transport by diffusion processes combined with retention of the molecule inside the nucleus and caused by affinity to nuclear components, has to be considered. In general, proteins larger than 60 kDa cannot pass the nuclear envelope by mere diffusion. To exclude the possibility of passive diffusion, we used a double-tagged construct (FP-CB(SC)-FP; #5) such increasing the size to 84 kDa, which is well above the exclusion limit.
In spite of their increased size, the double-tagged constructs revealed a nuclear localization comparable to that of their single-tagged analogues (Fig 4C, D). Because of the close neighbourhood of both fluorochromes, fluorescence resonance energy transfer (FRET) is possible when appropriate fluorochromes are used. Indeed, using the combination ECFP-EYFP a FRET effect was observed. As the donor molecule is never completely extinguished, the transfer appears to be incomplete, which is not unusual under the given experimental conditions [43]. FRET would become unlikely if the heavy and the light chain of CB(SC) were segregated by processing or general proteolysis. Therefore, the FRET experiment is a further proof that the single chain form keeps intact and that the entire fusion protein is transported into the nucleus.
We conclude that the underlying mechanism depends on an active transport. Further, these results clearly point out a directed nuclear transport and retention mechanism linked to the structure of the CB(SC) polypeptide. In advanced expression states, punctuate signals were observed, which were carried from the tips of the pseudopodia towards the Golgi area in a directed way and which penetrated the nuclear membrane without a noticeable delay [see Additional file 1].
Induction of cell death
LCLC-103H cells permanently expressing CB(FLM) chimeras were easy to clone. In contrast, cells transfected with CB(SC)-FP proved to be short-lived and all attempts to establish cell clones with a typical CB(SC)-FP expression failed. We therefore assume that this construct has the ability to provoke cell death. We proved whether it localizes in a time-dependent manner and monitored the process of cell death in time-lapse experiments (Fig 9A). Cell death followed a strict scheme which can be described in morphological terms over time starting as early as 12 h post transfection and finishing about 6 h later: (i) nuclear accumulation throughout rising expression level, (ii) formation of cytoplasmic and nuclear granules (around the Golgi and within the nucleus mostly), (iii) directed transport of granules to the Golgi and the nucleus, (iv) disintegration of these organelles followed by (v) rounding up and detachment of the cells from the support, (vi) stationary motility of the cell and cellular collapse [see Additional file 2]. Interestingly, cells which were transfected with FP-CB(HCC')-FP also died ~3–4 h after apparent expression. This took place in a different manner exhibiting membrane blebbing and cell swelling followed by bursting of the cell.
Figure 9 CB induced cell death. A. LCLC-103H cells were transiently transfected with CB(SC)-EYFP and the temporal expression course was continuously (Δt = 2 min) monitored beginning with early signs of fluorescence ~12 h post transfection. The process always followed the same scheme: Initially the yet weak fluorescence is distributed equally in the cytoplasm and in the nucleus. Then, fluorescent granules are formed (partly in the pseudopodia) and transported towards the nucleus passing the Golgi region. Granulation goes ahead with rapidly increasing fluorescence intensity in the nucleus. Finally, the Golgi apparatus disappears and the cell collapses within a few minutes while bubbling. Selected images taken by indicated points of a time-lapse experiment document these events. B. Mortality rates of LCLC-103 H cells transiently expressing CB(SC)-EGFP and control constructs (EGFP, CB(FLM)-EGFP) were determined by propidium iodide staining followed by FACS analysis. To meet falsifications of results that are caused by variable transfection rates (22–63%) the absolute mortality values were normalized to the respective transfection efficiency. Dead cells in consequence of the toxicity of the transfection agent and false positives attributable to autofluorescence were taken into account by subtraction of the respective control values: The number of dead cells (propidium iodide channel) amounted to ~13% in case of mock-transfected control cells; 2% of total were false positive "transfectants" (GFP channel). In comparison to CB(FLM)-EGFP, the toxicity of the CB(SC)-construct was almost twice as high (~78%).
We quantified cell death by propidium iodide staining and fax analysis 16 h post transfection and determined mortality rates for populations transfected with EGFP, CB(FLM)-EGFP, and CB(SC)-EGFP (Fig 9B). The mortality rate of mock-transfected cell populations was below 20%. Cells expressing the EGFP control only revealed mortality rates of ~18%. On the contrary, ~43% of the CB(FLM)-EGFP expressing cells and ~78% of the cells transfected with CB(SC)-EGFP died at that time. The latter population did not recover in the further time course and only a few cells with a very low expression level or aberrant localization survived.
Indications for apoptosis were proved. In some cases, nuclear fragmentation and membrane blebbing were observed. A well-defined evidence of apoptosis in respect to disintegration of the plasma membrane (Annexin V/propidium iodide assay) or DNA fragmentation (DNA ladder) could not be produced. Nevertheless, one should not exclude apoptosis in favour of necrosis from further considerations.
Discussion
Over the last years, the perception about the functions of CB has changed considerably. Recently found evidences assign to this 'lysosomal peptidase' key positions in cardinal processes also outside the lysosomes, like in apoptosis or cancer. Technical advances in microscopy and the development of stable chemical and genetic markers for organelles and molecules now facilitate powerful and direct in vivo approaches. They permit not only the localization of certain proteins but also the investigation of their intracellular transport, their interaction with other proteins, and their enzymatic activities, as well as the study of the cellular response in respect to overexpression or silencing of specific proteins.
In vivo, both normal tissues and especially tumours contain a population of truncated CB, which can be traced to alternative splicing (Fig 1A). Since expression and transport of CB are frequently altered in transformed and malignant cells as well as in cells undergoing apoptotic processes, we gave our attention to the investigation of such CB aberrations. For this purpose, we have labelled several recombinant CB forms by fluorescent proteins and subjected them to living cell imaging by advanced digital microscopy techniques.
Truncated cathepsin B forms
The naturally truncated Δ51CB lacks the complete signal sequence as well as parts of the N-terminal proregion. This product is barred from entering the ER, and thus from further processing and transport by the mannose-6-℗ pathway. However, the residual propeptide contains a MTS which becomes efficient and directs the predominant amount of the respective product to mitochondria [25]. Our own observations confirm this finding. Δ51CB is expressed both in vitro and in vivo as entire 35 kDa product [24]. This corresponds to our findings of further truncated artificial CB sequences irrespective of their size or tagging with markers: all constructs remained intact and they were not cleaved posttranslationally. Earlier assumptions [24] about the possible CB-specific enzymatic activity of Δ51CB were recently questioned [25,26].
Obviously, the propeptide is indispensable for proper in vivo-folding of the mature enzyme with the typical CB activity. Interestingly, a splicing variant of cathepsin L devoid of the signal peptide also appears associated with the nucleus and exhibits a specific cleavage activity [28]. Therefore, one should take into consideration that the truncated form(s) of CB might have cleaving characteristics, which do not become evident in the standard assays. In this report, we prove that neither the completeness of the sequence nor the CB specific enzymatic activity is relevant to the observed nuclear accumulation and induction of cell death.
Nuclear localization
Unlike CB(FLM), which is targeted to the lysosomes via ER and Golgi and partly secreted into the extracellular medium, the cytosol-expressed Δ51CB is mainly addressed to the mitochondria [25]. Our own experiments with the same construct confirm these findings. However, deduced from our measurements, a non-negligible fraction of the expression product can also be found in the nucleoplasm. Inspection of the published data [25] does not contradict our findings. Nuclear fluorescence cannot arise from unspecific decay or cleavage products inasmuch as double-tagged constructs reveal similar results as the single-tagged ones indicating the integrity of the constructs. Further, we propose a targeting signal downstream of the MTS which alternatively may direct CB and derivatives thereof into the nucleus. Obviously, a hierarchy of signals encoded within the CB polypeptide determines its intracellular distribution pattern. The signal peptide and the glycosylation sites are decisive for lysosomal targeting of the FLM-product. The signal peptide and the propeptide containing the MTS are removed during the maturation process. Thus, the nuclear targeting signal might become active after release of the enzyme from lysosomes into the cytosol. In case of the truncated Δ51CB, the MTS is predominant, whereas for the artificially truncated CB forms the nuclear targeting signal is characteristic. In the past, CB was also found in cell nuclei of tumour cells and normal tissues [31,44,45], but until now there are almost no indications to a potential transport mechanism or a specific function. Especially in apoptotic processes, CB and other CB-like peptidases were detected also in the cell nucleus [8,32]. However, these studies still miss a thorough scrutiny for the nuclear localization.
Here, artificially truncated CB-GFP chimaeras were used, from which the Δ72CB-construct came closest to the splicing variant Δ51CB in respect to its size. However, it was devoid of the functional sequence present in Δ51CB that encodes the N-terminal MTS. Not only Δ72CB but also the slightly shorter CB(SC) and other considerably shorter CB fragments were first expressed cytoplasmically as expected. In the sequel, they were enriched within granular structures, which were not consistent with lysosomes or mitochondria as might be supposed. Furthermore, these polypeptides were clearly proved in the nucleoplasm of several cell types. Frequently, nucleoli showed discrete regions of labelling. In contrast to the nucleus, which can be entered by both active and passive transport, the nucleoli are addressed exclusively by interaction with nucleolar building blocks [46]. Immunocytochemistry of CB(SC), which was tagged by a myc-epitope, confirms the results of the GFP-tagging. Though, a slightly higher reticular signal distribution was observed. In addition, this proves that in spite of its size, the fluorescent protein does not sufficiently affect the affinity of CB polypeptides to the respective localization sites.
Based on these results, the capacity of cathepsins particularly in the context of nuclear localization has to be reconsidered [47].
Mutational analysis
There are no clues to an already known NLS in CB according to literature and to our own computational analysis. The findings suggest that the complex differential distribution of artificially truncated CB might depend on distinct targeting signals. To identify the region(s) of a potential nuclear localization signal sequence or a signal patch, respectively, a number of mutated GFP-tagged constructs were produced. Despite the elimination of extensive sequence regions, partly including potential stabilizing elements such as disulfide bridges – e.g. in CB([C211_I243del]SC) –, the specific localization persisted to a high degree. The participation of the CB light chain in this sorting procedure was excluded. The heavy chain determines the nuclear localization only; the region with the highest impact on the specific localization could be narrowed down to its C-terminal subunit. Although constructs smaller than CB(C'1) did not reveal unequivocal results, the smallest of them, CB(C'4), was not targeted specifically. The assumption that the nuclear affinity essentially depends on the prominent acidic and polar surface residue E273, which is found within the relevant region, could not be proved by several specific mutations. Excision of the differential part of CB(C'3) and CB(C'4) did also not affect the localization. The deletion did not imply adjacent residues around the active site H278 in CB(C'4), which also might be important. Hence, the results do not support the existence of a linear signal sequence. Rather a composed signal patch is likely, which evolves from the three-dimensional conformation of the polypeptide. In contrast to linear signals, such signal patches are difficult to identify exactly.
The appearance of CB(SC) and other artificially truncated constructs in the midbody is striking (Fig 4C–G). According to a recent study [48], midbodies have a complex composition. However, only a few peptidases and no cysteine peptidases at all were found therein. Nevertheless, the association of CB constructs with the midbody supports their nuclear occurrence.
Transport mechanisms and interaction with the nuclear matrix
The exclusion limit for free diffusing molecules through the nuclear pores is at ~60 kDa. By applying constructs well above the exclusion limit (~84 kDa) we ruled out passive transport across the nuclear pore complex with high probability. The integrity of the products was proved by immunoblotting and by FRET analysis. In the time-lapse experiments, we noticed a directional transport of granules from across the cells to the Golgi and into the nucleus without any delay at the nuclear envelope. These observations ask for a specific transport system to which the expression product might be hooked into.
Are the imported artificial CB variants possibly retained inside the nucleus because of a specific affinity to nuclear components? To answer this question, a comparative TPM-photobleaching approach was applied. The mobility of CB(SC)-EGFP and ECFP-CB(SC)-EYFP was analysed in living cells by continuous photobleaching and FRAP. We chose the freely diffusing EGFP [41] and the tightly chromatin-bound H2A-EGFP [42] as limiting controls and TIF1A-EGFP as further control with partial mobile and immobile fractions. In both technical variants of the approach, the EGFP measurements obeyed curve shapes characteristic of free diffusion (almost exclusively mobile fraction); in contrast, those of H2A-EGFP were typical for predominantly immobile molecules. Hence, both controls reacted as to be expected and in analogy to former studies [49,50]. The courses of the CB(SC) graphs reflect an intermediate status indicating low immobile and high mobile fractions evolving from limited diffusion inside the nucleus. From this we assume that the artificially truncated CB is able to associate with nuclear matrix components. This nuclear affinity might be transferred to the naturally truncated Δ51CB form. Chromatin as a conceivable partner of interaction could be excluded from considerations by an OPM double labelling experiment of cells using GFP-tagged histone H2A and CB(SC) inasmuch as no colocalization could be observed. The relatively high amount of mobile CB(SC) might depend on scarcity of interaction partners: we have to consider that the CB products are overexpressed, other than their possible counterparts.
Cell viability
It was reported that the naturally truncated Δ51CB was directed to the mitochondria and that the cells died after fragmentation of the nucleus [25]; our observations confirm these findings. We suppose that nuclear targeting of Δ51CB might be overwhelmed by the present MTS.
Removal of this sequence in the artificially truncated Δ72CB and in further modified constructs results in their nuclear targeting and accumulation. Both overexpressed natural and artificial constructs lead to the same consequence, namely nuclear fragmentation and cell death. Neither we nor others [25] could prove that the induced cell death arises from apoptosis.
It was described that cell death can be preceded by a release of mature and active CB from the lysosomes and by the appearance of CB in the nucleus [7]. Our studies of the artificially truncated constructs support these observations. Any truncated forms of CB proved to have no regular CB enzymatic activity. A proper refolding of Δ51CB to an enzymatically active form was demonstrated under in vitro conditions [24]. However, one has also to take into consideration a different cleavage activity or functionality for the truncated variant(s) of CB. Such was recently found in case of truncated cathepsin L [28] and probably also cathepsin H [27].
The significance of results, which are obtained by overexpression, is often a contentious issue. Two arguments support the validity of our results: (i) Generally, expression levels of transfected cell populations diverge largely among individual cells. In case of CB(SC), the response to expression is severe and comprises also cells with obviously negligible expression. (ii) Time lapse video sequences demonstrate a granulation and a directed transport of CB(SC) to the nuclear region followed by fusion with the nucleus (see supplemental material).
Conclusion
Naturally appearing variations of CB and other related enzymes exhibit changed physiological characteristics and function as metabolic regulators in different states of diseases. We examined the nature of truncated CB by mutational analysis of extrinsic CB forms combined with advanced fluorescence microscopy. According to our results, artificially truncated CB forms lacking the MTS accumulated within the cell nucleus by an active transport mechanism and revealed binding affinity to nuclear matrix compounds. The region responsible for nuclear targeting resides in the C-terminal part of the protein. A hierarchy of signals is discussed. Expression of artificially truncated CB affected the cell viability to a large extent. Emerging from this, one has to raise the question whether the traditional understanding of distinct CB populations in terms of "normal" and "aberrant" might be misleading and thus has got to be reconsidered.
Methods
Cell lines
Investigations were performed on LCLC-103H cells derived from a human large cell lung carcinoma (ATCC# CCL5, DSMZ# 384). In addition to the original classification of these cells, it was now found that they are point-mutated in the p53 gene leading to an inactive p53. In control experiments, additional cell lines were used: HeLa (ATCC# CCL2), HEK-293 (DSMZ# 305), Wi-38 (ATCC# CCL75), primary human microvascular endothelial cells (HDMEC), Hep-G2 (DSMZ# 180), COS-7 (ATCC# CRL1651), and MDCK (ATCC# CCL34). The cell lines were cultured according to the recommendations of the suppliers.
Transfection procedure
For microscopical studies, the cells were seeded on 4.2 cm-diameter coverslips or on Lab-Tek® Chambered Cover Glasses (Nunc, Wiesbaden, Germany) at a density of 104cm-2 and transfected 12–15 h later at ~70% confluence with ~150 ng of the appropriate plasmid DNA by FuGENE6™ (Roche Molecular Biochemicals) according to the supplier's instructions. In double transfection experiments, equal masses of DNA were applied.
Fluorescence microscopy
Prior to observation, the coverslips with the transfected cells were mounted in perfusion chamber holders (PeCon, Erbach, Germany). The samples were observed 24–72 h post transfection at 34–36°C and 5% CO2.
Digital wide field microscopy (WFM)
Images were obtained by an Axiovert S100TV (Zeiss, Jena, Germany). It was equipped with long distance condenser and objectives (Neofluar® 10×/0.30 Ph, LD Apochromat® 20×/0.40 Ph, LD Apochromat® 40×/0.60 Ph, Fluar® 40×/1.30 Oil, Plan-Neofluar® 63×/1.25 Oil Ph, C-Apochromat® 40×/1.20 W, Plan-Apochromat® 63×/1.4 Oil), a CCD camera (Orca C4742-95, Hamamatsu Photonics, Hamamatsu, Japan), shutters, macro (Ludl Electronic Products Ltd., Hawthorne, NY, USA) and piezo (Pifoc720, PI, Karlsruhe, Germany) focus drives, and an incubator to guarantee proper growth conditions in long-term experiments. The automated filter wheels (Ludl) contained filters for ECFP (Ex 436/10, DiM 455, Em 480/40), EGFP (Ex 488/20, DiM 505, Em 535/40), EYFP (Ex 515/10, DiM 530, Em 560/40) (Chroma, Brattleboro, VT, USA), and Cresyl Violet (Ex 560/40; DiM 590; Em 600 LP) (Omega Optical Inc., Brattleboro, VT, USA) fluorescence detection. Image acquisition was controlled by the OpenLab software (Improvision, Coventry, UK). Optical slices were partly subjected to built-in deconvolution algorithms and processed to 3D-restoration by the Amira™ software (TGS Europe, Düsseldorf, Germany).
TPM
Single scans and serial images for time-lapses were acquired by an Eclipse TE300 microscope (Nikon, Düsseldorf, Germany) using a Plan-Apochromat® 60×/1.2 W objective (Nikon). A pulsed (13.2 ns) mode-locked Mira900-F Ti:sapphire laser (Coherent, Santa Clara, CA, USA) was pumped by a Verdi™ argon laser (Coherent). TPM was performed at 860 nm; fluorescence was detected consecutively at 470/30 nm (ECFP), at 535/30 nm (EYFP), or at 510/20 nm (EGFP), respectively. Microscopical set-up and image processing were described in detail previously [51].
Bleaching experiments
Nucleoplasmic diffusion of the fluorescent protein chimeras was analysed by photobleaching of living cells as described in [49] and shown exemplarily for the H2A-EGFP construct (Fig 5A). Besides minor modifications to the original work, which had to be introduced for technical reasons and which are described below in detail, TPM was used instead of OPM.
Fluorescence recovery after photobleaching (FRAP)
A nuclear region of 21.8 × 21.8 μm2 was scanned to describe the situation before bleaching (total scanning time: 2.033s; period duration: 5 μs). The fluorescence was monitored at the respective emission wavelengths (see above) with 5 mW incident laser power. No cell damage was observed with the chosen illumination parameters. Within this area, a smaller region of interest (ROI) of 2 × 2 μm2 was bleached completely by 10 continuous scans (total bleaching time: 20.33 s). The selection was reset manually to the region at the beginning and a series of 50 consecutive single scans was recorded; the delay between bleaching and start of recording was ~2 s. About 8% of the initial EGFP fluorescence and up to 30% of the ECFP or EYFP intensity were depleted by the additional post-bleaching irradiation. Mean grey values of the images were evaluated by the Scion Image software (Scion Corporation; Las Vegas, NV, USA). The relative fluorescence intensity values, Irel, were normalized according equation (1)
where T0 is the total intensity before bleaching, Tt the total intensity at various time points t, I0 the intensity of the prebleached ROI, and It the intensity of the ROI at the corresponding time points. The share of the mobile fraction, Fm, was calculated by equation (2)
where I∞ is the fluorescence in the ROI reaching a plateau after complete recovery, Ii the intensity during prebleach and Ib the intrinsic background fluorescence, which was negligibly low in our case. The effective diffusion coefficient, Deff, is conversely proportional to the characteristic diffusion time, τD, which can be determined from the time value at I∞/2. For a more complex view of the calculations refer to [52,53].
Continuous photobleaching
The same basic parameters and evaluation tools as described for FRAP were used. A series of 30 consecutive images was grabbed within a square ROI of 2 μm border width. Overview scans before and after the series were taken for comparison. The grey values of the ROI were normalized to their respective initial value and fitted by a regression function which was composed of two partial exponential terms by the SigmaPlot software (Systat Software, Inc., Point Richmond, CA, USA).
Determination of the mortality rates
Mortality of transiently transfected LCLC-103H cell populations was quantified after a propidium iodide (Sigma-Aldrich, Taufkirchen, Germany) staining by a FACSCalibur™ flow cytometer (Becton-Dickinson, Heidelberg, Germany). The cell concentration was 2–8 × 105 cellsml-1 in a measure volume of 50 μl. Efficiency of transfection was determined by EGFP fluorescence in the FITC channel; dead cells stained by propidium iodide were monitored in the propidium iodide channel (excitation at 488nm each). The data were evaluated by the EPICS® profile analyser software (Coulter Corp., Hialeah, FL, USA).
Apoptosis tests
LCLC-103H cells transiently expressing CB(SC)-EGFP were checked for plasma membrane integrity and DNA fragmentation as indicators for apoptosis.
Plasma membrane perforation: At the cellular level, early stage apoptosis was examined by the AnnexinV-assay as described by the supplier (Roche Molecular Biochemicals, Mannheim, Germany). Cells were stained by AnnexinV and propidium iodide and subjected to fluorescence microscopy and FACS analysis.
DNA fragmentation: Genomic DNA was isolated and cleared of RNA using the DNA ladder kit (Roche Diagnostics, Mannheim, Germany) according to the instructions. Apoptosis was triggered by 5 μm etoposide (Sigma-Aldrich) or MG115 (Sigma-Aldrich) in control cells. The samples were then subjected to agarose gelelectrophoresis.
Cytochemistry
Besides GFP, synthetic living cell markers and immunolabels were used.
Living cell staining
Nuclear DNA was stained with 2 μgml-1 Hoechst33342 (Sigma-Aldrich), lysosomes were marked with 50 nM LysoTracker Red (Molecular Probes, Leiden, The Netherlands).
Immunolabelling
Methanol fixed cells immobilized onto poly-lysine coated cover slips were treated with antibodies (Abs) against CB (primary Ab: α-hCB-shIgG, BioAss, Herrsching, Germany; secondary Ab: α-sh-dIgG(HC+LC)-Cy3, Dianova, Hamburg, Germany), GFP (primary Ab: α-GFP-rIgG, Clontech, Heidelberg, Germany; secondary Ab: α-r-gIgG(HC+LC)-Cy3, Dianova), and the myc-epitope (primary Ab: α-myc-mIgG; secondary Ab: α-m-dIgG-Texas Red). Finally, the cover slips were mounted onto slides with Mowiol (Calbiochem, Schwalbach, Germany).
Determination of the enzymatic activity
CB-specific activity was determined invivo by digital widefield microscopy (WFM). The fluorogenic substrate (Z-Arg-Arg)2-cresyl violet was added into the culture medium at 25 μm and the following enzymatic reaction was monitored by the release of the fluorescent dye cresyl violet in 1 min intervals over a period of 10 min. For quantitative purposes, we used invitro activity assays, which were performed as described [54].
SDS-PAGE and Western Blot analysis – Proteins were separated on 12% polyacrylamide SDS gels followed by an electro transfer to a nitrocellulose membrane and detected by appropriate Abs using alkaline phosphatase by standard procedures.
Cloning and mutational analysis
The sequence encoding the single chain form of human CB, CB(SC), was amplified from a known cDNA (IMAGE clone ID 380482; Deutsches Ressourcenzentrum für Genomforschung, Berlin, Germany) by PCR. It was subcloned into pcDNA3 (Invitrogen; NV Leek, NL) and tagged by appropriate fluorescent proteins (Clontech) analogously to the CB(FLM) sequence as described previously [55]. The CB(SC) sequence served as template for further mutation variants. Additional ATG start codons ensured proper translation. The artificially truncated sequence Δ72CB appeared as a by-product in the cloning process of CB(FLM) caused by an internal KpnI restriction site. Specific mutations (insertion, deletion, conversion) were introduced into the CB sequence by site-specific mutagenesis using standard PCR procedures. All constructs were verified by DNA sequencing.
Synthesized forward (fd) and reverse (rv) primers (Pharmacia; DKFZ) were used for PCR amplification of modified CB constructs. They contained restriction sites for cloning into pcDNA3 (underlined in the following primer sequences): SC-fd-KpnI: GGGGTACCATGCTGCCTGCAAGCTTCGATG, myc-SC-fd-KpnI: GGGGTACCATGGAGCAGAAGCTGATCTCCGAGGAGGACCTGCTGCCTGCAAGCTTCGATGCACGG, SC-X-rv-NotI: GGCGGCCGCTTAATCGGTGCGTGGAATTCCAGC, [ΔC'pro]SC-rv-SalI: GGGTCGACGATCTTTTCCCAGTACTGATCG, LC-rv-SalI: GGGTCGACATTGGTGTGGATGCAGATGCGG, HC-fd-KpnI: GGGGTACCATGGTCAGCGTGGAGCTGTCGG, HCN'-rv-SalI: GGGTCGACATTGTATCCGTAGTGCTTGTCC, HCC'-fd-KpnI: GGGGTACCATGAATTCCTACAGCGTCTCCA, HCC'1-fd-KpnI: GGGGTACCATGGAGATCTACAAAAACGGCC, HCC'2-fd-KpnI: GGGGTACCATGGTGTATTCGGACTTCCTGC, HCC'3-fd-KpnI: GGGGTACCATGCAACACGTCACCGGAGAGA, HCC'4-fd-KpnI: GGGGTACCATGCGCATCCTGGGCTGGGGAG, [E273L]SC-fd: CCAACACGTCACCGGACTGATGATGGGTGGCCATG, [E273L]SC-rv: CATGGCCACCCATCATCAGTCCGGTGACGTGTTGG, [E273del]SC-fd: ACCAACACGTCACCGGAATGATGGGTGGCCATGCC, [E273del]SC-rv: GGCATGGCCACCCATCATTCCGGTGACGTGTTGGT, [Q268_G277del]SC-fd: TACAAGTCAGGAGTGTACCATGCCATCCGCATCCTG, [Q268_G277del]SC-rv: CAGGATGCGGATGGCATGGTACACTCCTGACTTGTA, [L80_L283]SC-rv-SalI: GGGTCGACCAGGATGCGGATGGCATGGCCA.
The spatial orientation of the mutated regions was visualized by the molecular modelling software PyMOL™ (v0.97, © DeLano Scientific LLC, San Carlo, California, USA). The reconstruction was based on crystallographic data of a mature CB protein (PDB-Id: 1huc).
List of abbreviations
4MβNA, 4-methoxy-β-naphthylamide; CB, cathepsinB; DSMZ, Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH; FLM, full length message; ECFP/EGFP/EYFP, enhanced cyan/green/yellow fluorescent protein; fd, forward; FP, fluorescent protein; FRAP, fluorescence recovery after photobleaching; FRET, fluorescence resonance energy transfer; HC, heavy chain; LC, light chain; MTS, mitochondrial targeting signal; NLS, nuclear localization signal; OPM, one-photon laser scanning microscopy; ℗, phosphate; ROI, region of interest; rv, reverse; SC, single chain; TIF1A, transcription initiation factor 1A; TPM, two-photon laser scanning microscopy; WFM, digital widefield microscopy; Z-Arg-Arg, benzyloxycarbonyl-Arginine-Arginyl
Authors' contributions
FB carried out the molecular, biochemical, and microscopical studies, contributed to the cell culture, and drafted the manuscript. CD participated in the molecular cloning and the biochemical characterization, and contributed to the preparation of cells. ES carried out the cytochemical and enzymatic characterization of the cells, contributed to the microscopical studies, supervised the project, and drafted the manuscript. All authors have read and approved the final manuscript.
Supplementary Material
Additional File 1
Transient expression of CB(SC)-EYFP in LCLC-103H cells (phase contrast, fluorescence channel, and inverse grey value representation of the fluorescence signal). The sequence describes the formation of granules, their fusion, and the transport to the nucleus accompanied by breakup of the Golgi. WFM, LD Apochromat® 40×/0.60 Ph. Δt = 5 min, total time = 195 min. Video format: MPEG1. The abbreviation "MM" (mature message) is used equivalently to the term "SC" (single chain).
Click here for file
Additional File 2
Transient expression of ECFP-CB(SC)-EYFP in LCLC-103H cells (fluorescence channels, phase contrast, and superimposition of the last fluorescence frames). The double-tagged construct is expressed in the cytoplasm and accumulated within the nucleus. No significant differences among the two fluorescence channels can be observed suggesting that the polypeptide remains intact and is imported into the cell nucleus as a whole. Only the expressing cells undergo cell death and reveal membrane blebbing, which is typical for apoptotic processes. WFM; obj. Neofluar® 10×/0.30 Ph. Δt = 3 min, total time = 250 min. Video format: MPEG1. The abbreviation "MM" (mature message) is used equivalently to the term "SC" (single chain).
Click here for file
Acknowledgements
We would like to thank the following persons for their support: Bernd Werle (Thoraxklinik, Heidelberg, Germany; enzymatic activity measurements), Karsten Richter (DKFZ; OPM), Helmut Acker and Torsten Porwol (MPI für Molekulare Physiologie, Dortmund, Germany; TPM), and Andreas Hunziker (DKFZ; sequencing). We are also grateful to Birgit Krahling (DKFZ) for providing the primary endothelial cells, Ron van Noorden (Amsterdam; Netherlands) for the (Z-Arg)2-cresyl violet substrate, Jörg Uhlmann (DKFZ) for the α-myc-Abs, Gergana Dobreva (DKFZ) for TIF1A-EGFP, Tobias A. Knoch (DKFZ) for H2A-EGFP, and particularly to Kathrin Müntener (University of Zürich, Switzerland) for the generous provision of the Δ51CB-EGFP construct. We appreciate Christina Barther (Heidelberg), Christoph Spiess (Stanford University, USA), and Malte Wachsmuth (KIP, University of Heidelberg, Germany) for critical reading of the manuscript.
==== Refs
Salvesen GS A lysosomal protease enters the death scene J Clin Invest 2001 107 21 22 11134173
Reinheckel T Deussing J Roth W Peters C Towards specific functions of lysosomal cysteine peptidases: phenotypes of mice deficient for cathepsin B or cathepsin L Biol Chem 2001 382 735 741 11517926
Baici A Lang A Horler D Kissling R Merlin C Cathepsin B in osteoarthritis: cytochemical and histochemical analysis of human femoral head cartilage Ann Rheum Dis 1995 54 289 297 7763107
Cavallo-Medved D Sloane BF Cell-surface cathepsin B: understanding its functional significance Curr Top Dev Biol 2003 54 313 341 12696754
Podgorski I Sloane BF Cathepsin B and its role(s) in cancer progression Biochem Soc Symp 2003 70 263 276 14587299
Roshy S Sloane BF Moin K Pericellular cathepsin B and malignant progression Cancer Metastasis Rev 2003 22 271 286 12785001
Roberts LR Kurosawa H Bronk SF Fesmier PJ Agellon LB Leung WY Mao F Gores GJ Cathepsin B contributes to bile salt-induced apoptosis of rat hepatocytes Gastroenterology 1997 113 1714 1726 9352877
Vancompernolle K Van Herreweghe F Pynaert G Van de Craen M De Vos K Totty N Sterling A Fiers W Vandenabeele P Grooten J Atractyloside-induced release of cathepsin B, a protease with caspase-processing activity FEBS Lett 1998 438 150 158 9827536
Guicciardi ME Deussing J Miyoshi H Bronk SF Svingen PA Peters C Kaufmann SH Gores GJ Cathepsin B contributes to TNF-alpha-mediated hepatocyte apoptosis by promoting mitochondrial release of cytochrome c J Clin Invest 2000 106 1127 1137 11067865
Foghsgaard L Wissing D Mauch D Lademann U Bastholm L Boes M Elling F Leist M Jäättela M Cathepsin B acts as a dominant execution protease in tumor cell apoptosis induced by tumor necrosis factor J Cell Biol 2001 153 999 1010 11381085
Foghsgaard L Lademann U Wissing D Poulsen B Jäättela M Cathepsin B mediates tumor necrosis factor-induced arachidonic acid release in tumor cells J Biol Chem 2002 277 39499 39506 12185082
Werneburg NW Guicciardi ME Bronk SF Gores GJ Tumor necrosis factor-α-associated lysosomal permeabilization is cathepsin B dependent Am J Physiol Gastrointest Liver Physiol 2002 283 G947 G956 12223355
Yuan XM Li W Dalen H Lotem J Kama R Sachs L Brunk UT Lysosomal destabilization in p53-induced apoptosis Proc Natl Acad Sci USA 2002 99 6286 6291 11959917
Canbay A Guicciardi ME Higuchi H Feldstein A Bronk SF Rydzewski R Taniai M Gores GJ Cathepsin B inactivation attenuates hepatic injury and fibrosis during cholestasis J Clin Invest 2003 112 152 159 12865404
Hentze H Lin XY Choi MS Porter AG Critical role for cathepsin B in mediating caspase-1-dependent interleukin-18 maturation and caspase-1-independent necrosis triggered by the microbial toxin nigericin Cell Death Differ 2003 10 956 968 12934070
Liu N Raja SM Zazzeroni F Metkar SS Shah R Zhang M Wang Y Bromme D Russin WA Lee JC Peter ME Froelich CJ Franzoso G Ashton-Rickardt PG NF-kappaB protects from the lysosomal pathway of cell death EMBO J 2003 22 5313 5322 14517268
Madge LA Li JH Choi J Pober JS Inhibition of phosphatidylinositol 3-kinase sensitizes vascular endothelial cells to cytokine-initiated cathepsin-dependent apoptosis J Biol Chem 2003 278 21295 21306 12663669
Michallet MC Saltel F Preville X Flacher M Revillard JP Genestier L Cathepsin-B-dependent apoptosis triggered by antithymocyte globulins: a novel mechanism of T-cell depletion Blood 2003 102 3719 3726 12893746
Gong Q Chan SJ Bajkowski AS Steiner DF Frankfater A Characterization of the cathepsin B gene and multiple mRNAs in human tissues: evidence for alternative splicing of cathepsin B pre-mRNA DNA Cell Biol 1993 12 299 309 8494608
Berquin IM Cao L Fong D Sloane BF Identification of two new exons and multiple transcription start points in the 5'-untranslated region of the human cathepsin-B-encoding gene Gene 1995 159 143 149 7622042
Yan S Berquin IM Troen BR Sloane BF Transcription of human cathepsin B is mediated by Sp1 and Ets family factors in glioma DNA Cell Biol 2000 19 79 91 10701774
Yan S Sloane BF Molecular regulation of human cathepsin B: implication in pathologies Biol Chem 2003 384 845 854 12887051
Berardi S Lang A Kostoulas G Horler D Vilei EM Baici A Alternative messenger RNA splicing and enzyme forms of cathepsin B in human osteoarthritic cartilage and cultured chondrocytes Arthritis Rheum 2001 44 1819 1831 11508434
Mehtani S Gong Q Panella J Subbiah S Peffley DM Frankfater A In vivo expression of an alternatively spliced human tumor message that encodes a truncated form of cathepsin B. Subcellular distribution of the truncated enzyme in COS cells J Biol Chem 1998 273 13236 13244 9582368
Müntener K Zwicky R Csucs G Rohrer J Baici A Exon skipping of cathepsin B: mitochondrial targeting of a lysosomal peptidase provokes cell death J Biol Chem 2004 279 41012 41017 Epub 2004 Jul 15 15262981
Muentener K Willimann A Zwicky R Svoboda B Mach L Baici A Folding competence of N-terminally truncated forms of human procathepsin B J Biol Chem 2005 Jan 18
Waghray A Keppler D Sloane BF Schuger L Chen YQ Analysis of a truncated form of cathepsin H in human prostate tumor cells J Biol Chem 2002 277 11533 11538 Epub 2002 Jan 16. 11796715
Goulet B Baruch A Moon NS Poirier M Sansregret LL Erickson A Bogyo M Nepveu A A cathepsin L isoform that is devoid of a signal peptide localizes to the nucleus in S phase and processes the CDP/Cux transcription factor Mol Cell 2004 14 207 219 15099520
Strohmaier AR Porwol T Acker H Spiess E Tomography of cells by confocal laser scanning microscopy and computer-assisted three-dimensional image reconstruction: localization of cathepsin B in tumor cells penetrating collagen gels in vitro J Histochem Cytochem 1997 45 975 983 9212823
Pietras RJ Roberts JA Cathepsin B like enzymes: Subcellular distribution and properties in neoplastic and control cells from human ectocervix J Biol Chem 1981 256 8536 8542 7263668
Spiess E Brüning A Gack S Ulbricht B Spring H Trefz G Ebert W Cathepsin B activity in human lung tumor cell lines: ultrastructural localization, pH sensitivity, and inhibitor status at the cellular level J Histochem Cytochem 1994 42 917 929 8014475
Pizzorno MC Nuclear cathepsin B-like protease cleaves transcription factor YY1 in differentiated cells Biochim Biophys Acta 2001 1536 31 42 11335102
Riccio M Di Giaimo R Pianetti S Palmieri PP Melli M Santi S Nuclear localization of cystatin B, the cathepsin inhibitor implicated in myoclonus epilepsy (EPM1) Exp Cell Res 2001 262 84 94 11139332
Faubion WA Guicciardi ME Miyoshi H Bronk SF Roberts PJ Svingen PA Kaufmann SH Gores GJ Toxic bile salts induce rodent hepatocyte apoptosis via direct activation of Fas J Clin Invest 1999 103 137 145 9884343
Roberts LR Adjei PN Gores GJ Cathepsins as effector proteases in hepatocyte apoptosis Cell Biochem Biophys 1999 30 71 88 10099823
Müntener K Zwicky R Csucs G Baici A The alternative use of exons 2 and 3 in cathepsin B mRNA controls enzyme trafficking and triggers nuclear fragmentation in human cells Histochem Cell Biol 2003 119 93 101 12610728
Kessel D Luo Y Mathieu P Reiners JJ Jr Determinants of the apoptotic response to lysosomal photodamage Photochem Photobiol 2000 71 196 200 10687394
Feldstein AE Werneburg NW Canbay A Guicciardi ME Bronk SF Rydzewski R Burgart LJ Gores GJ Free fatty acids promote hepatic lipotoxicity by stimulating TNF-alpha expression via a lysosomal pathway Hepatology 2004 40 185 194 15239102
Guicciardi ME Leist M Gores GJ Lysosomes in cell death Oncogene 2004 23 2881 2890 15077151
Erdal H Berndtsson M Castro J Brunk U Shoshan MC Linder S Induction of lysosomal membrane permeabilization by compounds that activate p53-independent apoptosis Proc Natl Acad Sci USA 2005 102 192 197 Epub 2004 Dec 23 15618392
von Arnim AG Deng XW Stacey MG Cloning vectors for the expression of green fluorescent protein fusion proteins in transgenic plants Gene 1998 221 35 43 9852947
Mosammaparast N Ewart CS Pemberton LF A role for nucleosome assembly protein 1 in the nuclear transport of histones H2A and H2B EMBO J 2002 21 6527 6538 12456659
Danell AS Parks JH Fraying and electron autodetachment dynamics of trapped gas phase oligonucleotides J Am Soc Mass Spectrom 2003 14 1330 1339 14652182
Sinha AA Gleason DF Limas C Reddy PK Wick MR Hagen KA Wilson MJ Localization of cathepsin B in normal and hyperplastic human prostate by immunoperoxidase and protein A-gold techniques Anat Rec 1989 223 266 275 2646985
Pietras RJ Szego CM Roberts JA Seeler BJ Lysosomal cathepsin B-like activity: mobilization in prereplicative and neoplastic epithelial cells J Histochem Cytochem 1981 29 440 450 6972405
Carmo-Fonseca M Mendes-Soares L Campos I To be or not to be in the nucleolus Nat Cell Biol 2000 2 E107 E112 10854340
Chapman HA Cathepsins as transcriptional activators? Dev Cell 2004 6 610 611 15130484
Skop AR Liu H Yates J 3rdMeyer BJ Heald R Dissection of the mammalian midbody proteome reveals conserved cytokinesis mechanisms Science 2004 305 61 66 15166316
Phair RD Misteli T High mobility of proteins in the mammalian cell nucleus Nature 2000 404 604 609 10766243
Wachsmuth M Weidemann T Muller G Hoffmann-Rohrer UW Knoch TA Waldeck W Langowski J Analyzing intracellular binding and diffusion with continuous fluorescence photobleaching Biophys J 2003 84 3353 3363 12719264
Bestvater F Spiess E Stobrawa G Hacker M Feurer T Porwol T Berchner-Pfannschmidt U Wotzlaw C Acker H Two-photon fluorescence absorption and emission spectra of dyes relevant for cell imaging J Microsc 2002 208 108 115 12423261
Axelrod D Koppel DE Schlessinger J Elson E Webb WW Mobility measurement by analysis of fluorescence photobleaching recovery kinetics Biophys J 1976 16 1055 1069 786399
Sprague BL Pego RL Stavreva DA McNally JG Analysis of binding reactions by fluorescence recovery after photobleaching Biophys J 2004 86 3473 3495 15189848
Werle B Jülke B Lah T Spiess E Ebert W Cathepsin B fraction active at physiological pH of 7.5 is of prognostic significance in squamous cell carcinoma of human lung Br J Cancer 1997 75 1137 1143 9099961
Bestvater F Knoch TA Langowski J Spiess E Construct conversions caused by simultaneous cotransfection: "GFP-Walking" Biotechniques 2002 32 830 836 11962605
Berquin IM Sloane BF Cathepsin B expression in human tumors Adv Exp Med Biol 1996 389 281 94 8861022
| 15807897 | PMC1087480 | CC BY | 2021-01-04 16:39:10 | no | BMC Cell Biol. 2005 Apr 4; 6:16 | utf-8 | BMC Cell Biol | 2,005 | 10.1186/1471-2121-6-16 | oa_comm |
==== Front
BMC Cell BiolBMC Cell Biology1471-2121BioMed Central London 1471-2121-6-181582630710.1186/1471-2121-6-18Methodology ArticleA strategy to study tyrosinase transgenes in mouse melanocytes Lavado Alfonso [email protected] Ander [email protected] Manuel [email protected] Lluís [email protected] Department of Molecular and Cellular Biology Centro Nacional de Biotecnología (CNB-CSIC) Campus de Cantoblanco, C/ Darwin, 3 28049 Madrid, Spain2 Spanish National Cancer Centre (CNIO) C/ Melchor Fernández Almagro 3 28029 Madrid, Spain3 St Jude Children's Research Hospital 332 N. Laudardale Memphis TN 38105, USA2005 12 4 2005 6 18 18 1 11 2004 12 4 2005 Copyright © 2005 Lavado et al; licensee BioMed Central Ltd.2005Lavado 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 number of transgenic mice carrying different deletions in the Locus Control Region (LCR) of the mouse tyrosinase (Tyr) gene have been developed and analysed in our laboratory. We require melanocytes from these mice, to further study, at the cellular level, the effect of these deletions on the expression of the Tyr transgene, without potential interference with or from the endogenous Tyr alleles. It has been previously reported that it is possible to obtain and immortalise melanocyte cell cultures from postnatal mouse skin.
Results
Here, we describe the efforts towards obtaining melanocyte cultures from our Tyr transgenic mice. We have bred our Tyr transgenic mice into Tyr c-32DSD mutant background, lacking the endogenous Tyr locus. In these conditions, we failed to obtain immortalised melanocytes. We decided to include the inactivation of the Ink4a-Arf locus to promote melanocyte immortalisation. For this purpose, we report the segregation of the Ink4a-Arf null allele from the brown (Tyrp1b) mutation in mice. Finally, we found that Ink4a-Arf +/- and Ink4a-Arf -/- melanocytes had undistinguishable tyrosine hydroxylase activities, although the latter showed reduced cellular pigmentation content.
Conclusion
The simultaneous presence of precise genomic deletions that include the tyrosinase locus, such as the Tyr c-32DSD allele, the Tyr transgene itself and the inactivated Ink4a-Arf locus in Tyrp1B genetic background appear as the crucial combination to perform forthcoming experiments. We cannot exclude that Ink4a-Arf mutations could affect the melanin biosynthetic pathway. Therefore, subsequent experiments with melanocytes will have to be performed in a normalized genetic background regarding the Ink4a-Arf locus.
==== Body
Background
Eukaryotic genes are organised on chromosomes in units known as expression domains, that are believed to include all regulatory elements required for correct gene expression [1]. We use the mouse tyrosinase locus (Tyr) as an experimental model to study mammalian expression domains [2,3]. The mouse Tyr gene is located in chromosome 7 [4], encodes the rate-limiting enzyme in melanin biosynthesis and is tightly regulated during development, being exclusively expressed in neural crest-derived melanocytes and optic cup-derived retinal pigment epithelium (RPE) cells [5,6].
Classically, the approach used to functionally identify regulatory elements has been testing a series of DNA constructs containing different amounts of regulatory sequences in transgenic animals. Minigene Tyr constructs were able to rescue the albino phenotype of recipient animals, but displayed variability in pigmentation [7,8]. In contrast, the generation of transgenic mice with a 250 kb yeast artificial chromosome (YAC) covering the entire mouse Tyr locus completely rescued the albino phenotype, resulting in mice that were indistinguishable from agouti wild-type pigmented mice [9,10]. These results pointed to the existence of important regulatory elements, absent in previous standard constructs, such as the locus control region (LCR), identified 15 kb upstream of the mouse Tyr promoter [11,12]. The LCR is necessary to establish the proper expression pattern of the mouse tyrosinase gene. The absence of the LCR resulted in weaker pigmentation, variegated expression in the melanocytes and RPE cells and delayed retinal pigmentation in transgenic mice [13]. Moreover, two binding boxes for nuclear factors within the LCR core, known as boxes A and B, were identified by in vitro analysis [14] and, recently, have been incorporated into a new boundary activity associated within the LCR region [15]. We have generated transgenic mice with new YAC Tyr transgenes carrying a range of specific mutations within the LCR region [10].
To address a more detailed study, both at the functional and structural level (using biochemical and cellular approaches), of the role of LCR-variants in these different transgenes, a number of problems had to be solved, including: (1), the dispersed nature of melanocytes which prevented us from direct analysis of relevant tissues, such as skin, where many other unrelated cell types are found; (2), the tyrosinase albino allele (Tyr c), present in all recipient mouse strains used for the generation of transgenic mice, carrying a reported point-mutation within the coding region that results in a non-functional protein, without the transcriptional status of the locus being affected [16-18], and, [3], although it has been demonstrated that it is possible to obtain mouse melanocyte immortal cell lines from postnatal skins [19-22], often it becomes difficult to overcome the senescence period that all primary cell cultures undergo.
In this study we describe our efforts and the strategy to obtain melanocyte cell cultures from YAC Tyr transgenic mice in a genetic background lacking the endogenous mouse tyrosinase gene, and the effect of the inactivation of the Ink4a-Arf locus [23] on proliferation, senescence and tyrosinase activity of established melanocyte cell lines.
Results and discussion
Transfer of YAC Tyr transgenes from albino outbred NMRI mice (Tyr c / Tyr c) to a Tyr c-32DSD / Tyr c-32DSD background
Previous Tyr transgenic mice have been generated in albino outbred NMRI mice [9,10,12]. All observed pigmentation is due to the expression of the Tyr transgene but the presence of mutant albino Tyr protein and mRNA from the host Tyr c allele [16-18] could interfere in subsequent cellular, biochemical and genomic analyses. The albino 32DSD mutant mouse (Tyr c-32DSD) carries a deletion of the entire Tyr locus, encompassing ~200 kb of mouse chromosome 7 [24]. Therefore, a breeding program was established between Tyr transgenic and albino 32DSD mice in order to obtain Tyr transgenes in animals lacking the endogenous Tyr alleles. Genotype analysis of the resulting mice was carried out by Southern blot. An EcoR I restriction fragment length polymorphism (RFLP) in exon 2 detected by the RFLP probe allows the identification of the transgenic Tyr allele (17 kb), derived from the mouse C3H strain, and the endogenous counterpart albino Tyr c NMRI allele (12 kb) (9) (Fig. 1).
Figure 1 Transfer of YAC Tyr transgenes from a Tyr c / Tyr c background to a Tyr c-32DSD / Tyr c-32DSD background. Southern blot analysis of representative mice obtained during the breeding program, using the RFLP [9] and p19Arf E1 probes. Lanes: (M) 1 kb-ladder (Invitrogen), (1) YAC Tyr YRT2/∅ [9]; Tyr c-32DSD / Tyr c, (2) YRT2/∅ ; Tyr c-32DSD / Tyr c-32DSD, (3) ∅/∅ ; Tyr c-32DSD / Tyr c, and (4) ∅/∅ ; Tyr c-32DSD / Tyr c-32DSD. Note: ∅ indicates the absence of the transgene. Tg: YAC Tyr transgene derived signal. Tyr c: endogenous albino tyrosinase locus. p19: The p19Arf E1 probe (kindly provided by M. Malumbres), detecting exon 1 of the single-copy p19arf gene, was employed as an internal control for DNA loading and comparisons.
Melanocyte primary cultures of YAC Tyr/∅ ; Tyr c-32DSD/ Tyr c-32DSD
To gain further insight on a series of YAC Tyr transgenic mice carrying a range of deletions around the LCR [10,12], we decided to prepare cell lines that could be representative of these animals. Chromatin analyses cannot be done in tissue samples obtained directly from transgenic animals, due to the low number of cells expressing the Tyr gene (RPE cells and melanocytes) and the complexity of the tissues or organs containing these cells (eye and skin, respectively). In addition, the presence of the mutated, but transcriptionally active [16-18], albino Tyr locus in all transgenic mice generated to date could interfere with the interpretation and the acquisition of experimental data. To avoid this problem we mobilised the YAC Tyr transgenes to a genetic background lacking the endogenous mouse gene, as shown in Fig. 1, and then we tried to establish melanocyte cultures from these mice.
Mouse melanocyte immortal cell lines can be derived from postnatal skins [19-22]. Mouse crosses were established with parental lines to obtain pigmented pups with the desired genotype (YAC Tyr / ∅ ; Tyr c-32DSD / Tyr c-32DSD). Genotype analysis was not necessary to distinguish between transgenic and non transgenic pups, because melanin, clearly visible in the eye and in the skin at this stage, could only derive from the YAC Tyr transgene expression. Melanocyte primary cultures from dorsal skin of heterozygous YAC Tyr transgenes in homozygous Tyr c-32DSD background were prepared as described [19]. Individual melanocytes and small melanocytes colonies appeared at culture day 10. The number of cells increased slowly until day 25–35, when melanocytes showed signs of senescence. Most cells died by day 85–90 and no immortal cell lines could be established (Fig. 2A–C). Similar results were obtained from all YAC Tyr derivative transgenes bred to homozygous Tyr c-32DSD mice.
Figure 2 The absence of at least one Ink4a-Arf allele overcomes the senescence of primary melanocytes in culture. Melanocytes from heterozygous YAC Tyr transgenic in homozygous Tyr c-32DSD background became large, flat, vacuolated and highly pigmented (A, B). No surviving melanocytes were detected in the culture dishes after the senescence step (C). Melanocytes from Ink4a-Arf +/- (G, H, I) and Ink4a-Arf -/- (J, K, L) mutant mice in a Tyrp1B background are black, small, bipolar, pale and without significant sings of senescence, as compared to wild type Ink4a-Arf +/+ melanocytes (D, E, F). A: culture at day 10; B: culture at day 18; C: culture at day 48; D, G, J: cultures at day 43; E, H, K: cultures at day 57; and, F, I, L: cultures at day 82. Scale ebars in C, D, G, J = 200 μm and in A, B, E, F, H, I, K, L = 150 μm.
Mouse melanocyte primary cultures and their corresponding immortalised cell lines have been established from a number of mutant mice [25-29], although this type of cell lines can be sometimes difficult to achieve, as it may be inferred from these listed publications, in which reported cell lines have been generated by the same laboratory. A number of parameters can influence the success in obtaining immortalised melanocytes from mice. First, skins from mouse pups are used as starting material, carrying bacteria and other microorganisms that can contaminate the cultures. Second, and most important, melanocytes, as any somatic cell line in culture, undergo a senescence step previous to their immortalisation. Due to the low number or surviving cells after this senescence step, cells need to be cultured continuously during a minimum of 3–6 months to obtain an immortal cell line [19,20]. In most of the cultures we did not observe melanocytes after the senescence step (Fig. 2A–C). These results were obtained with all different primary cultures, regardless of their genotype, indicating a problem at the immortalisation step.
Segregation of the Ink4a-Arf locus from the Tyrp1 b locus
It has been reported that melanocyte immortal cell lines (i.e. melan-a and melan-c) lack the p16 protein, most likely due to the lost of the Ink4a-Arf locus during the culture process [30]. Melanocytes from Ink4a-Arf (-/-) null mice proliferate exponentially without showing any signs of senescence, thus it has been proposed that the generation of melanocyte immortal cell lines in an Ink4a-Arf null background would be much easier [30]. Comparable results had been obtained before with fibroblast cultures from Ink4a-Arf homozygous mutant mice [31]. The absence of p16 leads to the inhibition in the inactivation of CDK4 and CDK6. These kinases inactivate the retinoblastoma pathway, promoting the proliferation of the cells [32,33]. Therefore, we decided to mobilise the Ink4a-Arf null allele into our YAC Tyr transgenic mice.
First, we had to remove the brown mutation co-segregating with the Ink4a-Arf null allele. The murine brown locus corresponds to the gene encoding the Tyrosinase-related protein 1 (Tyrp1), an enzyme which is also implicated in the melanin biosynthetic pathway [34]. It was originally noted that the coat colour of Ink4a-Arf null mice was paler than their wild-type littermates [23]. Indeed, biochemical evidence was presented that was highly suggestive of defective Tyrp1 activity in Ink4a-Arf null melanocytes [35]. Homozygous Tyrp1 b mutant mice display less tyrosinase activity than wild type, because mutated forms of the Tyrp1 protein affect tyrosinase processing [36]. Notably, the Tyrp1 locus, is close to the Ink4a-Arf locus, in mouse chromosome 4, at a distance of 8.7 Mb (mouse ENSMBL, build 32). Most inbred laboratory strains with defective Tyrp1 activity carry the recessive brown allele, Tyrp1 b that contains a single amino acid change at a critical residue in the Tyrp1 protein [37]. To test directly whether the Ink4a-Arf null allele was linked to the Tyrp1 b allele, two single nucleotide polymorphisms (SNPs) that characterize the Tyrp1 b allele and that can be diagnosed by subsequent restriction enzyme digestion [37] were used. In particular, a Tyrp1 b-linked SNP at exon 4 eliminates a Taq I restriction site and, similarly, another Tyrp1 b – linked SNP at exon 5 eliminates an Hga I restriction site. Using these two markers, we directly proved that the original Ink4a-Arf null mice were indeed homozygous for the Tyrp1 b allele, thus explaining their paler coat colour (Fig. 3). The Tyrp1 b recessive allele linked to the mutated Ink4a-Arf allele was probably derived from the ES cell line originally used for targeting the Ink4a-Arf locus, namely WW6 ES cells, which had a complex genetic background [38].
Figure 3 Generation of an inbred Ink4a/Arf -/-; Tyrp1 B / Tyrp1 Bmouse strain. Illustrative examples of the genotyping of the Tyrp1 locus by PCR and digestion with Taq I (A) or Hga I (B). Gel lanes correspond to: (1) Ink4a/Arf -/- ; Tyrp1 b / Tyrp1 b, (2) Ink4a/Arf +/- ; Tyrp1 B / Tyrp1 b, (3) Ink4a/Arf -/- ;Tyrp1 B / Tyrp1 b and (4) Ink4a/Arf -/- ; Tyrp1 B / Tyrp1 B.
Ink4a/Arf-null strain in a pure C57BL/6J genetic background and unlinked from the mutant Tyrp1 b allele were obtained to avoid the confounding presence of the Tyrp1 b allele. Ink4a-Arf +/- mice were backcrossed seven times with wild-type C57BL/6J mice, eventually yielding Ink4a-Arf +/- ; Tyrp1 B/ Tyrp1 b mice. These mice were intercrossed to produce a number of Ink4a-Arf -/- mice that were in their majority Ink4a-Arf -/- ; Tyrp1 b/ Tyrp1 b and, accordingly, had a brown coat. Exceptionally, one mouse (from a total population of 27 Ink4a-Arf -/- mice) was identified as an Ink4a-Arf -/- but had a black coat. This mouse turned out to represent a recombinant with an Ink4a-Arf -/- ; Tyrp1 B/ Tyrp1 b genotype. From this animal, and after the appropriate crosses, a strain of mice in C57BL/6J background that were Ink4a-Arf -/- ; Tyrp1 B / Tyrp1 B was obtained (Fig. 3), and used for subsequent experiments.
Melanocyte primary cultures from Tyrp1 B/ Tyrp1 B Ink4a-Arf mutant mice
To evaluate the effect of the inactivation of the Ink4a-Arf locus in Tyrp1 B / Tyrp1 B genetic background on the immortalisation of mouse melanocytes, melanocyte primary cultures from dorsal skin of Ink4a-Arf +/+, Ink4a-Arf +/-, Ink4a-Arf -/- pups in a Tyrp1 Bbackground were prepared (Fig. 2D–L), using the same described procedures [19]. Individual melanocytes and colonies appeared at day 10. The number of cells increased slowly in Ink4a-Arf +/+ (Fig. 4A) and, around day 35, melanocytes increased their size and lost their bipolar shape. Few or none Ink4a-Arf +/+ melanocytes survived the senescence step, and most of these melanocytes died by day 85–90 (Fig. 2D–F). In contrast, Ink4a-Arf +/- and Ink4a-Arf -/- kept their bipolar shape and small size, with a limited number of cells showing any sign of senescence (Fig. 2G–L). Proliferation of Ink4a-Arf -/- melanocytes was higher than in Ink4a-Arf +/- melanocytes, and both were much higher than in Ink4a-ARF +/+ melanocytes (Fig. 4A). By day 45 of culture, Ink4a-Arf -/- melanocytes entered the exponential phase of growth. In contrast, Ink4a-Arf +/- melanocytes did not show evidences of exponential growth until day 65 (Fig. 4A). This delay could be explained by LOH (loss of heterozygosity), spontaneously occurring in Ink4a-Arf +/- melanocytes and affecting the remaining wild-type allele of the Ink4a-Arf locus, a common event that is known to take place both in vitro [39] and in vivo [40]. Similar results have been reported [30] using independent mouse colonies.
Figure 4 The effect of Ink4a-Arf locus on cell growth and Tyr enzymatic activity. (A) Cell growth of Ink4a-Arf -/- (open circles), Ink4a-Arf +/- (closed circles) and Ink4a-Arf +/+ (crosses) melanocytes. Each point represents one subculture step and represents the mean count of cells from three replicates. (B) Tyrosine hydroxylase activity in albino melan-c (white bars), [Ink4a-Arf +/-; Tyrp1 B / Tyrp1 B ] (dark-grey bars), [Ink4a-Arf -/- ; Tyrp1 B / Tyrp1 B ] (light-grey bars) and melan-a (black bars) cells. (C) Melanin content in the same cellular types, as in (B). Bars represent the mean value (+/- SD) from three replicates. Tyrosinase activity and melanin content were measured in melanocyte primary cultures at passage 8 (~day 70 of culture). Statistically significant differences (t-Student test) are indicated as follows: * p < 0,1; ** p < 0,01; *** p < 0,001.
Tyrosinase activity and melanin content in the melanocyte cultures from Ink4a-Arf mutant mice
A valid approach to analyse the role of the different regulatory regions of the mouse Tyr gene in the transcription of the locus and, eventually, in the amount of mature protein being made, is the measurement of the enzymatic activity of the derived Tyr protein. We measured the levels of Tyr enzymatic activity and melanin content using reported procedures [41] in melanocyte cultures, to study the influence of the presence or absence of the p16 protein in the expression of the Tyr gene. Tyrosine hydroxylase activity values from Ink4a-Arf -/- and Ink4a-Arf +/- cell extracts were undistinguishable and significantly lower to values obtained in melan-a cells (Fig. 4B). However, the quantity of melanin in the proliferative Ink4a-Arf -/- melanocytes cell cultures was significantly lower than in Ink4a-Arf +/- or melan-a cells (Fig. 4C).
Differences in melanin content could be explained by different individual cell culture response to TPA or/and CT that are present in cell culture medium. However, the observed differences in cellular pigmentation were maintained after removing CT from cell culture medium. In addition, TPA is always required as an additive to maintain cellular proliferation. Increasing the amount of TPA does not result in a parallel increase in pigmentation, opposite to what is observed with CT.
Differences in melanin content could also be explained by the control of the retinoblastoma (RB) protein by p16, and the observation that RB protein interacts with the transcription factor microphthalmia (Mift) [42], that controls the expression of the Tyr gene. Differences between Ink4a-Arf mutant cells and melan-a could be due to the presence of a number of additional alterations in this latter immortal cell line, such as the loss of expression of p16Ink4a [30]. The recent generation of mice with increased gene dosage of Ink4a-Arf will be instrumental to further investigate the influence of this locus on Tyr activity [43].
Conclusion
With all these results we can conclude that the simultaneous presence of: [1] at least one mutant allele of the Ink4a-Arf locus; [2], the Tyr c-32DSD mutant albino allele in homozygosis and; [3], the presence of the relevant Tyr transgene in heterozygosis, are required for the establishment and the study of immortal mouse melanocyte cultures from transgenic mice carrying Tyr constructs. Finally, we cannot exclude that Ink4a-Arf mutations could affect the melanin biosynthetic pathway. Therefore, experiments with mouse melanocytes will have to be performed in a normalized genetic background regarding the Ink4a-Arf locus.
Methods
Animals
Four types of mice were used in this study: YRT2 YAC tyrosinase heterozygous transgenic mice in an albino outbred NMRI background (YRT2/∅ ; Tyr c / Tyr c) (line #1999) [9], 32DSD radiation induced albino mutant mice (Tyr c-32DSD/ Tyr c-32DSD) [24], homozygous Ink4a-Arf mutant mice (Tyr + / Tyr + ; Tyrp1 b / Tyrp1 b ; Ink4a-Arf -/-) in C57BL/6J genetic background [23] and wild-type pigmented C57BL/6J mice. All experiments complied with local and European legislation concerning vivisection and the experimentation and use of animals for research purposes.
Southern blot analysis
The discrimination of the endogenous tyrosinase gene from the YAC-tyrosinase transgenes was performed as previously described, using the RFLP probe, containing exon 2 of the mouse tyrosinase gene [9,12]. To obtain an endogenous internal control, membranes were co-hybridised with a single-copy mouse gene, the p19Arf E1 probe, a EcoR I DNA fragment, 230 bp in length, containing exon 1 of the p19Arf gene (pRSp19arfE1 plasmid, generous gift from M. Malumbres). In brief, genomic DNA was isolated from mice tail tips and prepared for southern blot as described [12]. 15–20 μg of genomic DNA were digested with EcoR I (Roche, Basel, Switzerland), fractionated by horizontal gel electrophoresis in 0,8% agarose and transferred to a Hybond-N nylon membrane (Amersham, Buckinghamshire, UK) by capillary blotting. RFLP and p19Arf E1 DNA probes were labelled with [α32P] dCTP using the High Prime labelling kit (Roche). Membranes were hybridised in Southern hybridisation solution (0.25 M Na2HPO4 pH = 7.2, 7% SDS, 1% BSA) overnight at 65°C, washed at 65°C in 20 mM Na2HPO4 pH = 7.2, 1% SDS, 1 mM EDTA pH = 8 and resulting blots exposed for 1–3 days and scanned with a PhosphorImager (Molecular Dynamics, Sunnyvale, CA, USA). Quantification of the hybridisation signals was performed using the ImageQuant v1.2 software (Molecular Dynamics).
Genotyping of the Ink4a-Arf and Tyrp1 loci
The Ink4a-Arf and Tyrp1 loci were genotyped by PCR as follows. The Ink4a-Arf wild-type allele was specifically detected using primers that amplify Ink4a-Arf exon 2: mp16F, 5'-ATGATGATGGGCAACGTTC-3' and mp16R, 5'-CAAATATCGCACGATGTC-3'. The Ink4a-Arf null allele, which has a neo-cassette substituting exons 2 and 3 [23], was genotyped using primers that hybridise, respectively, with the neo-cassette (oligo Neo) and with Ink4a-Arf flanking genomic sequences (oligo R1): Neo, 5'-CTATCAGGACATAGCGTTGG-3' and R1, 5'-AGTGAGAGTTTGGGGACA GAG-3'. To genotype the Tyrp1 locus, two different PCR reactions were performed to amplify, respectively, exons 4 and 5 of the Tyrp1 gene. Amplification of exon 4 was performed using primers: Tyrp1-4F, 5'-CTGCGATGTCTGCACTGATGACTT-3' and Tyrp1-4R, 5'-AGGGTATCGTACTCTTCCAAGGAT-3'. Amplification of exon 5 was performed using primers: Tyrp1-5F, 5'-ACAGCACTGAGGGTGGACCAATC-3' and Tyrp1-5R, 5'-AGGGTATCGTACTCTTCCAAGGAT-3'. After amplification, the product corresponding to exon 4 was digested with Taq I, and the product corresponding to exon 5 was digested with Hga I. Digestion mixtures were separated in standard agarose gels and visualized with ethidium bromide. The Tyrp1 bmutant allele does not contain neither of the previous restriction sites, Taq I and Hga I, whereas both enzymes digest the Tyrp1 B wild-type allele. PCR reactions had 3 mM MgCl2, 1% DMSO, 2.5 mM dNTPs (Epicentre, Madison, WI, USA), 20 pmol of each primer, and 0.25 μl of Taq-Gold (Applied Biosystems, Foster City, CA, USA). In addition, each reaction had approximately 100 ng of genomic DNA extracted from the tail tips. Annealing temperature was 60°C and PCR reactions were carried on for 30 amplification cycles.
Melanocyte cultures
Melanocyte cultures were prepared, essentially, as previously described [19]. Briefly, dorsal skin biopsies were obtained from pups of all investigated genotypes between +19.5 and +22.5 d.p.c. stages (postnatal P2–P3). Dorsal skin was split in 5 μg/ml trypsin (Sigma, St. Louis, MO, USA) in PBS and the epidermal layer then minced with a pair of surgical blades in 250 μg/ml trypsin and 200 μg/ml EDTA in PBS. Cells were cultured on a feeder layer of mitomycin-treated murine XB2 keratinocytes [44]. The cells were grown in RPMI-1640 medium containing 2 mM glutamine, 10% fetal calf serum, 100000 U/l penicillin, 100 mg/l streptomycin sulphate (all from Invitrogen, Carlsbad, CA, USA), 200 nM tetradecanoyl phorbol acetate (TPA) (Sigma) and 200 pM cholera toxin (CT) (Sigma), at 37°C, 95% humidity and 10% CO2 pressure. Explant cultures from different donor mice were kept separate. Passages were made when cultures became subconfluent, and melanocytes were counted at each passage. Feeder cells were added when necessary. Control melan-a and melan-c cells, derived from inbred C57BL/6J and outbred albino LAC-MF1 mice, respectively, were cultured as previously described [19,20].
Quantification of melanin and tyrosinase enzymatic activities
Melanin contents in whole cell extracts were measured by spectrophotometer essentially as described [45]. In brief, 6 × 106 cells from ~day 70 of culture were collected and homogenised in 300 μl of PBS, 100 μl of homogenate incubated for 14–16 hours at room temperature with 900 μl of 2 M NaOH, 20% DMSO and absorbance measured at 470 nm.
Tyrosinase enzymatic activities were recorded following described assays [41,46,47]. In brief, for tyrosine hydroxylase activity, 6 × 106 cells from ~day 70 of culture were collected and cell extracts were prepared in 10 mM Sodium Phosphate buffer pH = 6.8 to which Tween-20 (Igepal) was added (1% final concentration) prior the assay. Reaction volume included 10 μl of L-DOPA 250 mM, 10 μl of L-[3,5-3H]-Tyrosine mix (450 μl of L-Tyrosine 262 μM in 10 mM Sodium Phosphate buffer pH = 6.8 and 50 μl L-[3,5-3H]-Tyrosine [1 mCi/ml, 46 Ci/mmol, Amersham]), 20 μl of Sodium Phosphate buffer pH = 6.8 and 10 μl of cell extracts, was incubated for 1 hour at 37°C, and stopped by adding 450 μl trichloracetic acid (TCA) 1% (Merck, Darmstadt, Germany). A small amount of absorbing substrate (active carbon [Merck] and Celite 545 [Fluka, St. Gallen, Switzerland], 1:1) was added, mixed for 30 min at room temperature and centrifuged. Radioactivity from clear supernatants (100 μl) was measured in a β-scintillation counter (Beckmann, Fullerton, CA, USA).
Authors' contributions
AL carried out the molecular genetic, cellular, biochemical and statistical studies and drafted the manuscript, MS conceived of the segregation of ink4a-Arf from mutant brown locus, AM carried out the genotype analysis of ink4a-Arf mice. LM conceived of the general study, and participated in its design and coordination, in collaboration with MS, and helped to draft the manuscript. MS also helped to draft the manuscript. All authors read and approved the final manuscript.
Acknowledgements
This work was supported by funds from the Spanish Ministry of Science and Technology Bio2000-1653 and Bio2003-08196 to LM and SAF2002-03402 to MS. All experiments complied with local and European legislation concerning vivisection and the experimentation and use of animals for research purposes. The authors are grateful to D. Bennett and E. Sviderskaya for melan-a and melan-c cells, for their continuous support and for generous teaching of melanocyte culture methods, to E. Rinchik, S. Shinpocks and P. Hunsicker (ORNL) for kindly providing albino 32DSD cryopreserved mouse embryos, to J. Fernandez and S. Montalbán for rescueing the 32DSD mouse stock at CNB, to J.C. García-Borrón for useful comments and to P. Cozar and M. Cantero for technical assistance with mouse colonies. Correspondence and request for materials should be addressed to Lluís Montoliu [email protected].
==== Refs
Bonifer C Developmental regulation of eukaryotic gene loci: which cis-regulatory information is required? Trends Genet 2000 16 310 315 10858661 10.1016/S0168-9525(00)02029-1
Giraldo P Montoliu L Size matters: use of YACs, BACs and PACs in transgenic animals Transgenic Res 2001 10 83 103 11305364 10.1023/A:1008918913249
Montoliu L Gene transfer strategies in animal transgenesis Cloning Stem Cells 2002 4 39 46 12006155 10.1089/153623002753632039
Regales L Giraldo P Garcia-Diaz A Lavado A Montoliu L Identification and functional validation of a 5' upstream regulatory sequence in the human tyrosinase gene homologous to the locus control region of the mouse tyrosinase gene Pigment Cell Res 2003 16 685 692 14629727 10.1046/j.1600-0749.2003.00100.x
Beermann F Schmid E Schutz G Expression of the mouse tyrosinase gene during embryonic development: recapitulation of the temporal regulation in transgenic mice Proc Natl Acad Sci U S A 1992 89 2809 2813 1557387
Gimenez E Lavado A Giraldo P Montoliu L Tyrosinase expression is not detected in mouse brain outside the retinal pigment epithelium cells European J of Neurosc 2003 18 2673 2676 10.1046/j.1460-9568.2003.02992.x
Beermann F Ruppert S Hummler E Bosch FX Muller G Ruther U Schutz G Rescue of the albino phenotype by introduction of a functional tyrosinase gene into mice EMBO J 1990 9 2819 2826 2118105
Schedl A Beermann F Thies E Montoliu L Kelsey G Schutz G Transgenic mice generated by pronuclear injection of a yeast artificial chromosome Nucleic Acids Res 1992 20 3073 3077 1620604
Schedl A Montoliu L Kelsey G Schutz G A yeast artificial chromosome covering the tyrosinase gene confers copy number-dependent expression in transgenic mice Nature 1993 362 258 261 8459851 10.1038/362258a0
Giraldo P Montoliu L Artificial chromosome transgenesis in pigmentary research Pigment Cell Res 2002 15 258 264 12100491 10.1034/j.1600-0749.2002.02030.x
Porter S Larue L Mintz B Mosaicism of tyrosinase-locus transcription and chromatin structure in dark vs. light melanocyte clones of homozygous chinchilla-mottled mice Dev Genet 1991 12 393 402 1822431 10.1002/dvg.1020120604
Montoliu L Umland T Schutz G A locus control region at -12 kb of the tyrosinase gene EMBO J 1996 15 6026 6034 8947025
Gimenez E Giraldo P Jeffery G Montoliu L Variegated expression and delayed retinal pigmentation during development in transgenic mice with a deletion in the locus control region of the tyrosinase gene Genesis 2001 30 21 25 11353514 10.1002/gene.1028
Ganss R Montoliu L Monaghan AP Schutz G A cell-specific enhancer far upstream of the mouse tyrosinase gene confers high level and copy number-related expression in transgenic mice EMBO J 1994 13 3083 3093 8039502
Giraldo P Martinez A Regales L Lavado A Garcia-Diaz A Alonso A Busturia A Montoliu L Functional dissection of the mouse tyrosinase locus control region identifies a new putative boundary activity Nucleic Acids Res 2003 31 6290 6305 14576318 10.1093/nar/gkg793
Jackson IJ Bennett DC Identification of the albino mutation of mouse tyrosinase by analysis of an in vitro revertant Proc Natl Acad Sci U S A 1990 87 7010 7014 2119500
Shibahara S Okinaga S Tomita Y Takeda A Yamamoto H Sato M Takeuchi T A point mutation in the tyrosinase gene of BALB/c albino mouse causing the cysteine-serine substitution at position 85 Eur J Biochem 1990 189 455 461 2110899 10.1111/j.1432-1033.1990.tb15510.x
Yokohama T Silversides DW Waymire KG Kwon BS Takeuchi T Overbeek PA Conserved cysteine to serine mutation in tyrosinase is responsible for the classical albino mutation in laboratory mice Nucleic Acids Res 1990 18 7293 7298 2124349
Bennett DC Cooper PJ Dexter TJ Devlin LM Heasman J Nester B Cloned mouse melanocyte lines carrying the germline mutations albino and brown: complementation in culture Development 1985 105 379 385 2806130
Bennett DC Cooper PJ Hart IR A line of non-tumorigenic mouse melanocytes, syngeneic with the B16 melanoma and requiring a tumour promoter for growth Int J Cancer 1987 39 414 418 3102392
Tamura A Halaban R Moellmann G Cowan JM Lerner MR Lerner AB Normal murine melanocytes in culture In Vitro Cell Dev Biol 1987 23 519 522 3610949
Halaban R Moellmann G Tamura A Kwon BS Kuklinska E Pomerantz SH Lerner AB Tyrosinases of murine melanocytes with mutations at the albino locus Proc Natl Acad Sci U S A 1988 85 7241 7245 3140237
Serrano M Lee H Chin L Cordon-Cardo C Beach D DePinho RA Role of the INK4a locus in tumor suppression and cell mortality Cell 1996 85 27 37 8620534 10.1016/S0092-8674(00)81079-X
Rinchik EM Stoye JP Frankel WN Coffin J Kwon BS Russell LB Molecular analysis of viable spontaneous and radiation-induced albino (c)-locus mutations in the mouse Mutat Res 1993 286 199 207 7681531
Spanakis E Lamina P Bennett DC Effects of the developmental colour mutations silver and recessive spotting on proliferation of diploid and immortal mouse melanocytes in culture Development 1992 114 675 680 1618134
Sviderskaya EV Bennett DC Ho L Bailin T Lee ST Spritz RA Complementation of hypopigmentation in p-mutant (pink-eyed dilution) mouse melanocytes by normal human P cDNA, and defective complementation by OCA2 mutant sequences J Invest Dermatol 1997 108 30 34 8980282 10.1111/1523-1747.ep12285621
Sviderskaya EV Novak EK Swank RT Bennett DC The murine misty mutation: phenotypic effects on melanocytes, platelets and brown fat Genetics 1998 148 381 390 9475748
Suzuki T Li W Zhang O Novak EK Sviderskaya EV Wilson A Bennett DC Roe BA Swank RT Spritz RA The gene mutated in cocoa mice, carrying a deffect of organelle biogenesis, is a homologue of the human Hermansky-Pudlak syndrome-3 gene Genomics 2001 78 30 37 11707070 10.1006/geno.2001.6644
Suzuki T Li W Zhang O Karim A Novak EK Sviderskaya EV Hill SP Bennett DC Levin AV Nieuwenhuis HK Fong CT Castellan C Miterski B Swank RT Spritz RA Hermasky-Pudlak syndrome is caused by mutations in HPS4, the human homolog of the mouse light-ear gene Nat Genet 2002 30 321 324 11836498
Sviderskaya EV Hill SP Evans-Whipp TJ Chin L Orlow SJ Easty DJ Cheong SC Beach D DePinho RA Bennett DC p16(Ink4a) in melanocyte senescence and differentiation J Natl Cancer Inst 2002 94 446 454 11904317
Serrano M Lin AW McCurrach ME Beach D Lowe SW Oncogenic ras provokes premature cell senescence associated with accumulation of p53 and p16INK4a Cell 1997 88 593 602 9054499 10.1016/S0092-8674(00)81902-9
Serrano M Hannon GJ Beach D A new regulatory motif in cell-cycle control causing specific inhibition of cyclin D/CDK4 Nature 1993 366 704 707 8259215 10.1038/366704a0
Sharpless NE DePinho RA The INK4A/ARF locus and its two gene products Curr Opin Genet Dev 1999 9 22 30 10072356 10.1016/S0959-437X(99)80004-5
del Marmol V Beermann F Tyrosinase and related proteins in mammalian pigmentation FEBS Lett 1996 381 165 168 8601447 10.1016/0014-5793(96)00109-3
Sviderskaya EV Gray-Schopfer VC Hill SP Smit NP Evans-Whipp TJ Bond J Hill L Bataille V Peters G Kipling D Wynford-Thomas D Bennett DC p16/cyclin-dependent kinase inhibitor 2A deficiency in human melanocyte senescence, apoptosis, and immortalization: possible implications for melanoma progression J Natl Cancer Inst 2003 95 723 732 12759390
Toyofuku K Wada I Valencia JC Kushimoto T Ferrans VJ Hearing VJ Oculocutaneous albinism types 1 and 3 are ER retention diseases: mutation of tyrosinase or Tyrp1 can affect the processing of both mutant and wild-type proteins FASEB J 2001 15 2149 2161 11641241 10.1096/fj.01-0216com
Zdarsky E Favor J Jackson IJ The molecular basis of brown, an old mouse mutation, and of an induced revertant to wild type Genetics 1990 126 443 449 2245916
Ioffe E Liu Y Bhaumik M Poirier F Factor SM Stanley P WW6: an embryonic stem cell line with an inert genetic marker that can be traced in chimeras Proc Natl Acad Sci U S A 1995 92 7357 7361 7638196
Obata M Lee GH Kanda H Kitagawa T Ogawa K Loss of heterozygosity at loci on chromosome 4, a common genetic event during the spontaneous immortalization of mouse embryonic fibroblasts Mol Carcinog 1997 19 17 24 9180924 10.1002/(SICI)1098-2744(199705)19:1<17::AID-MC3>3.0.CO;2-K
Orlow I Lacombe L Hannon GJ Serrano M Pellicer I Dalbagni G Reuter VE Zhang ZF Beach D Cordon-Cardo C Deletion of the p16 and p15 genes in human bladder tumors J Natl Cancer Inst 1995 87 1524 1529 7563186
Gimenez E Lavado A Giraldo P Cozar P Jeffery G Montoliu L A transgenic mouse model with inducible Tyrosinase gene expression using the tetracycline (Tet-on) system allows regulated rescue of abnormal chiasmatic projections found in albinism Pigment Cell Res 2004 17 363 370 15250938 10.1111/j.1600-0749.2004.00158.x
Yavuzer U Keenan E Lowings P Vachtenheim J Currie G Goding CR The Microphthalmia gene product interacts with the retinoblastoma protein in vitro and is a target for deregulation of melanocyte-specific transcription Oncogene 1995 10 123 134 7824265
Matheu A Pantoja C Efeyan A Criado LM Martín-Caballero J Flores JM Klatt P Serrano M Increased gene dosage of Ink4a/Arf results in cancer resistance and normal aging Genes Dev 2004 18 2736 2746 15520276 10.1101/gad.310304
Rheinwald JG Green H Serial cultivation of strains of human epidermal keratinocytes: the formation of keratinising colonies from single cells Cell 1975 6 331 334 1052771 10.1016/0092-8674(75)90183-X
Donatien PD Orlow SJ Interaction of melanosomal proteins with melanin Eur J Biochem 1995 232 159 164 7556145 10.1111/j.1432-1033.1995.tb20794.x
Pomerantz SH Tyrosine hydroxylation catalysed by mammalian tyrosinase: an improved method of assay Biochem Biophys Res Comun 1964 16 188 194 10.1016/0006-291X(64)90359-6
Solano F García-Borron JC Norlund JJ, Boissy RE, Hearing VJ, King RA, Ortonne JP Advances in enzymatic analysis of melanogenesis The Pigmentary System: Physiology and Pathophysiology 1998 New York: Oxford University Press 461 471
| 15826307 | PMC1087481 | CC BY | 2021-01-04 16:39:10 | no | BMC Cell Biol. 2005 Apr 12; 6:18 | utf-8 | BMC Cell Biol | 2,005 | 10.1186/1471-2121-6-18 | oa_comm |
==== Front
BMC Evol BiolBMC Evolutionary Biology1471-2148BioMed Central London 1471-2148-5-271581997910.1186/1471-2148-5-27SoftwareDeduction of probable events of lateral gene transfer through comparison of phylogenetic trees by recursive consolidation and rearrangement MacLeod Dave [email protected] Robert L [email protected] Ford [email protected] Eric [email protected] GenomeAtlantic, 1721 Lower Water Street, Suite 401, Halifax, NS, B3J 1S5, Canada2 Department of Biochemistry & Molecular Biology, Dalhousie University, 5850 College St., Halifax, NS, B3H 1X5, Canada2005 8 4 2005 5 27 27 26 11 2004 8 4 2005 Copyright © 2005 MacLeod et al; licensee BioMed Central Ltd.2005MacLeod 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
When organismal phylogenies based on sequences of single marker genes are poorly resolved, a logical approach is to add more markers, on the assumption that weak but congruent phylogenetic signal will be reinforced in such multigene trees. Such approaches are valid only when the several markers indeed have identical phylogenies, an issue which many multigene methods (such as the use of concatenated gene sequences or the assembly of supertrees) do not directly address. Indeed, even when the true history is a mixture of vertical descent for some genes and lateral gene transfer (LGT) for others, such methods produce unique topologies.
Results
We have developed software that aims to extract evidence for vertical and lateral inheritance from a set of gene trees compared against an arbitrary reference tree. This evidence is then displayed as a synthesis showing support over the tree for vertical inheritance, overlaid with explicit lateral gene transfer (LGT) events inferred to have occurred over the history of the tree. Like splits-tree methods, one can thus identify nodes at which conflict occurs. Additionally one can make reasonable inferences about vertical and lateral signal, assigning putative donors and recipients.
Conclusion
A tool such as ours can serve to explore the reticulated dimensionality of molecular evolution, by dissecting vertical and lateral inheritance at high resolution. By this, we mean that individual nodes can be examined not only for congruence, but also for coherence in light of LGT. We assert that our tools will facilitate the comparison of phylogenetic trees, and the interpretation of conflicting data.
==== Body
Background
Phylogeny in a context of LGT
The Tree of Life has long been a central metaphor of evolutionary theory, and is a useful framework upon which to build a natural classification. However, understanding the phylogeny of species from gene trees is a very challenging task, both technically and conceptually.
First, not all genes are useful phylogenetic markers. Paralogs resulting from gene duplications occurring prior to the divergence of the organismal lineages being studied can easily confuse an analysis, as would happen, for instance, when unknowingly constructing a tree for mammals based on hemoglobin sequences from some species and myoglobins from others. For different reasons (see [1] for a detailed review), orthologs themselves must be scrutinized carefully. Notably, the phylogenetic signal in genes can be reduced by their limited size, radiative (star-like) evolution they may have undergone or the phenomenon of mutational saturation. Much of the time, the quality of phylogenetic signal is too low to resolve all of the problems under consideration in a single analysis. Multiple gene trees are often then exploited because interestingly, in each gene tree, some nodes may in fact receive significant support. Increasing the number of markers might then resolve the full tree, piecemeal. However, before accepting these individual relationships as solid, the phylogenetic quality and congruence between markers must also be tested.
Artifacts of tree reconstruction or the use of an inappropriate evolutionary model can lead to robust but artifactual groupings in gene trees. Recombination events producing mosaic genes can also blur phylogenetic trees and may produce controversial groupings. As a consequence, software has been developed to test that the signal carried by a single gene is not self-contradictory [2,3]. But even non-mosaic orthologs for different genes can be in conflict, when one of them has been laterally transferred.
This last source of conflicting signal is a major concern. All living systems from viruses [4] to eukaryotes [5] can participate in the transfer of genetic material. Lateral transfer occurs within domains of life, but also across domains, for different markers. There is now broad general agreement that lateral gene transfer (LGT) is a major force in the evolution of prokaryotes [5-7]. Additional evidence suggests that gene transfer might also be an important evolutionary mechanism in protist evolution [8]. For instance, Andersson et al. [9] recently reported that alanyl-tRNA synthetase had been transferred from Nanoarchaeota to Diplomonads and Parabasalids. The same authors [10] showed that LGT has affected both eukaryotes and prokaryotes with respect to glutamate dehydrogenase.
One may choose to ignore conflicting signal as if it were noise, even if legitimate evolutionary events underlie it. If such a choice is made, incompatibilities between different trees can be resolved by supertree [11-15] or a posteriori consensus approaches [16]. Supertree methods assemble an input set of separate phylogenetic trees with shared taxa into a larger tree [13,17] (or several trees). By fitting variously supported clades together, they allow large phylogenies based on different characters to be constructed rapidly and have been applied to a broad range of species [18]. Consensus approaches, such as obtained by concatenating sequences [19] or by averaging over a large number of genes [20] produce resolved phylogenies by overwhelming noise with signal that is presumed to be systematically congruent and historically true, though weak.
However, if this "noise" is in fact bona fide phylogenetic signal, then the tree cannot display a historical picture of evolution: internal nodes do not accurately represent the evolutionary process. It might be useful in taxonomy, indeed represent the best possible compromise for taxonomic purposes, but such a graph could do little to satisfy our understanding of molecular evolution. The averaging employed by these methods irons over what are perhaps the most interesting wrinkles.
However, it should still be possible to make progress towards the deeper goal of understanding genome and organism history, if we are willing to abandon the formalism of the tree, itself just a convenient metaphor for genealogical relationships. Horizontal inheritance, as complicated as it is to deal with within the constraints of the Tree paradigm, is simply a facet within a more realistic model of genealogy yet to develop fully. SplitsTree [2] and its kind are excellent at representing this higher dimensionality, though the webs that it draws are still just graphs of relationships. Little about evolutionary history can be gleaned from these graphs, though they do suggest compelling hypotheses.
Here, we are interested in retracing evolutionary history as much as possible, attending to both the vertical and horizontal axes of inheritance. We do not think that LGT should be dismissed as noise and discarded in the construction of a phylogeny, but neither do we wish to abandon the concept of vertical inheritance, which by the very nature of cell division, must be a factor in genealogy, overwhelmingly so in the short term. Instead of forcing each gene to fit a given model, therefore, our aim is to extract as much of a gene's phylogenetic information as we can, vertical and horizontal inheritance modes included, and cross-reference the information with that obtained from other genes. We call the resulting graphs syntheses, because they have both tree-like and web-like parts [21]. In doing so, we reckon as the supertree users, consensus makers or concatenation advocates do that recognizing the vertical backbone of a synthesis graph is an important part of the task, but nevertheless, just a part of the ultimate goal of phylogenetics.
Using phylogenetic tree comparison to suggest LGT
This exercise requires an accurate identification of LGT. Many methods researching similarity/dissimilarity among genes exist to assess the occurrence of LGT [5,8,22], with varying reliability. Koski and Golding [23] showed that best BLAST hits do not necessarily represent the closest sequence relatives, casting doubt on the reliability of BLAST based approaches (at least when done naively). While Daubin et al. [24] showed that, in the bacterial realm at least, results from genome content and compositional approaches to detect LGT cannot always be taken as evidence for genetic exchange. In fact, it seems that phylogenetic analysis, even if it is more time consuming and might fail to detect transfers between more closely related organisms [5], is still one of the most reliable ways to investigate LGT [25]. Indeed, observing a close and robust phylogenetic relationship between gene sequences of distantly related organisms to the exclusion of gene sequences of more closely related organisms likely indicates LGT [26]. Alternative explanations, still possible a priori, would be artifacts of tree reconstruction, or complicated cases of reciprocal gene loss.
Highlighting conflicting signal between presumably correct gene trees is necessary in identifying LGT, although tree comparison is not trivial. Broad comparisons can be conducted between pairs of trees to determine if they are significantly different [15,27]. For instance, the incompatibility of different topologies can be assessed by an AU test for a given gene, using maximum likelihood estimates [28]. However, such global comparisons only indicate that there is a disagreement between the history of genes, not what this disagreement is. An alternative pruning of selected taxa and subsequent statistical tests of incongruency [29-31] can be attempted to identify the species misplaced in one of the two trees, considering one tree as a reference. Species to be pruned are generally chosen after an observation of bootstrap support values [32], as those robustly located in different parts of the trees under comparison. If the incongruence persists after these taxa have been removed, those species are then generally not considered responsible for the incompatibility between the trees. Conversely, if the pruning of species with significantly different locations in the trees eliminates incongruence, then these species could be considered as recipients of LGT, if no artifact is otherwise suspected of causing their odd placement in one of the topologies. In practice, this pruning approach is generally time-consuming and inefficient, due to the high number of possible pruning combinations, exacerbated by a possible misinterpretation of results due to taxon-sampling effects.
Recently, Addario-Berry et al. [33] implemented a promising comparison between a reference tree and a gene tree, overcoming some of these difficulties. Interestingly, their program posits lateral transfer schemes and scenarios to rationalize the differences between trees. These evolutionary scenarios, based on LGT, provide an explicit and accurate description of possible transfer events. In addition to the identification of a recipient species, they suggest a donor species, in the reference tree or outside of it. To do so, they estimate the minimum number of transfers necessary to explain disagreements between the pair of trees. This is done using the following criterion: since A and B are siblings in the gene tree, either the ancestral gene AB must have been present in the last common ancestor of A and B in the species tree or a lateral transfer event has occurred from the A lineage to the B lineage (or vice versa). At present, this method compares only one gene tree against the reference tree, and only if they include exactly the same taxa. In addition, it works only for rooted directed trees, which unfortunately limits its application. Notably, the requirement of strictly bifurcating trees forbids the collapse of unsupported nodes, thus greatly exaggerating the difference between trees.
In this paper, we revisit Addario-Berry et al.'s search for evolutionary scenarios to describe LGT but also insist on the detection of common vertical features among trees. We propose a pair of programs for dealing with multiple phylogenetic tree comparison, against a given reference tree. These trees may (or may not) differ in their species content, topological relationships, label positions, bootstrap support, and branch lengths.
Implementation
Horizstory
Comparing trees for two different genes amounts to exchanging branches within the first genes' tree (editing it) to match another's, and assessing the relative likelihood of this scenario. The topological comparison of phylogenetic trees is inherently difficult, as the number of possible edit paths (where one rearranges the branches of one tree to match another's) increases rapidly with the number of differences between trees. Where edits are individually independent, they may furthermore occur in any order within an edit path (evolutionary scenario), thus exploding their number factorially. Without resorting to heuristics and its inherent approximations, and without constraining the types of possible edits (apart from those that are biologically impossible: LGT with one's ancestor), we are therefore limited to comparing trees that are relatively similar. However, if one abandons the constraint of assuming that trees are fully resolved [33], which is most often not the case, then many apparent differences disappear and the problem becomes more tractable. One should be concerned with explaining only robust differences between trees, which can be determined by bootstrap support, for example, in ML phylogenies.
It is also important to compare the right trees, which amounts to a wise choice of reference tree against which test trees (typically gene-based trees) are compared. A good reference tree may be one that minimizes the overall number of differences it finds when matched with a variety of test trees; thus references based on concatenated sequences, supertrees, or genomic phylogenies should be most suitable for this purpose. Still, one might be uncertain about the exact choice of reference tree (or its optimal rooting), where alternatives are equally attractive. One could then use each candidate reference tree in turn as a reference against the set of test trees, and measure the extent of LGT predicted with each choice. This may suggest that one reference tree is better than another in this context, a result interesting on its own with respect to the issue of organismal phylogeny and the "true" tree.
Our approach to the comparison of phylogenetic trees, with the purpose of detecting lateral gene transfer (LGT) as well as determining the degree of concordance of vertical inheritance relationships, makes use of a recursive procedure of consolidation and rearrangement. Consolidation involves the simplification of identical topological features (vertical inheritance relationships) by collapsing such features [for example, a triplet-taxon relationship (("A","B"),"C") in common to both trees is collapsed to the single-taxon "((A,B),C)"]. This is followed in a second step by the collapse of compatible topological features. A compatible feature is, for example, a relationship of (("A","B"),"C") in the reference tree and an unresolved relationship of ("A","B","C") in the candidate tree, leading to " [A,B],C]". Compatible features do not necessarily support vertical relationships, but neither do they provide evidence for lateral gene transfer. Once the pair of trees to be compared is thus simplified, each of the candidate tree's leaves is moved to every alternative node in its tree in turn, with each move being tested by consolidation. Where simplification is possible (where the topologies can further converge), the move is productive and launches another pathway of rearrangement, where further rearrangements and simplifications are tried until the pair of trees can converge to identity, or until the pathway is abandoned. A pathway is abandoned, if, for example, it would require more steps than another pathway already explored.
We suppose that a rearrangement that can bring a pair of trees closer to one another topologically, is equivalent to "undoing" an event of lateral gene transfer. The position from which a taxon had to be moved in order to make the trees more similar is taken to be the LGT donor, whereas the taxon being moved is then the recipient. Given that rearrangement reconstitutes reference topologies (vertical inheritance relationships), presumed LGT targets are disqualified from suggesting such relationships by being pruned from the trees prior to the next recursive rearrangement.
Last but not least, rearrangement pathways must go to completion in order to be reported, resulting in full convergence of the pair of trees, but some edit distances are shorter than others, and suggest a smaller number of LGT events. For reasons of evolutionary parsimony as well as computational economy, this conservative route should be preferred.
Our scenarios can include different kinds of LGT events. Events of LGT may be nested, or otherwise intertwined, needing reversal by multiple rearrangements. The clade founded by an LGT donor may have subsequently had its species membership obfuscated by later exchanges of genetic material, yielding an unnatural assemblage of nomenclatural tags (species labels) in a presumed lineage. We distinguish such intermediary groupings in our output using an asterisk, indicating an ambiguity in deducing the root taxon, that being the actual organism that served as LGT donor. We also found it necessary to indicate (using a prime mark) when an LGT event cannot be attributed directly to a clade found in the reference tree, but rather to a phantom sister of that clade. This signals a 'basal transfer', which is observed when a taxon migrates out of its own clade to sit just outside of that clade in the candidate tree. Since LGT cannot occur with one's ancestor, the best explanation in a context of LGT is that the taxon received genetic material from a sister clade which happens not to be represented among the taxa in the dataset under investigation. Other nomenclatural marks in our program's output, mentioned above, include the parenthesis indicating an identical topological feature, and the square bracket indicating a compatible topological feature. The program's output thus consists of a list of the rearrangement pathways and the consequently deduced vertical and lateral (LGT) features. The frequencies of these features are also summarized in the output for each the pair of trees, and a global summary is presented for all pairs of trees analyzed.
For various reasons, a user might have a collection of trees to compare that include different taxa, or a user might more generally wish to exclude specific taxa from certain individual phylogenetic analyses. Where trees have different complements of taxa, pruning of taxa outside of the intersection of the two sets is done automatically. Where a pair of trees includes common taxa that the user wishes to exclude from analysis explicitly, he or she need only supply files listing which taxa should be excluded, either globally (applied to the reference tree and all of the candidate trees), or locally (applied to the reference tree and a candidate tree on a one-by-one basis). Pruning might be prescribed if, for example, one suspects causes other than LGT to be responsible for some of the differences between a pair of trees, such as long branch attraction. Absence of such files indicates that no pruning is desired.
We have successfully tested Horizstory with many different sets of real data, including hundreds of trees with 13 species and several dozen trees with over 27 species. Simple and nested LGTs simulated by manual modifications of a reference tree were also correctly reported in the scenarios.
Lumbermill
Lumbermill is a phylogeny editor, written in Java, for drawing trees and syntheses, using Horizstory output as input. It realizes our notion of synthesis (see Fig. 2) by first representing a vertical phylogram as its backbone. Line thickness is drawn in proportion to the percentage of genes supporting a given grouping; common patterns of vertical inheritance are thus easy to identify. (The assumption here is that over the short term, vertical inheritance is the dominant pattern). Overlaid onto this backbone are links that represent presumed LGTs, completing the synthesis.
Figure 2 Synthesis diagram. The vertical-inheritance backbone representing the input reference tree is shown in dark blue, with the line thickness of an internal branch corresponding to the frequency of its support across the whole dataset. Putative LGT events are in orange, connecting donors (circles) with recipients (arrowheads); where there are multiple possible donor candidates, these converge onto a double arrowhead (see text). Where the apparent donor of a gene falls outside of the taxa included in the analysis, one is created as a basal group taxon, indicated in light blue. In order to avoid graphical congestion, branches in the tree may be artificially extended, as dotted segments. Colours are editable, and links are interactive. Clicking on node 3A, for example, displays the following message: 3A: set006r, branch length:0.03755, thickness:0.25, files supported: mvin.tre:2/2, where support for the node is 2/2 for mvin.tre (both edit paths support the node), and 0 (unsupported) for hypoprot.tre, biob.tre and n6methylase.tre. The segment's thickness, therefore, is simply (2/2)/4.
The images trees or syntheses are highly editable, allowing the user to change fonts and line colours to provide a customized view of the data. The order of a node's descendants can be swapped. The backbone can be rerooted on another node or even unrooted. To help describe events in the synthesis, nodes are labeled numerically by distance from the root then alphabetically from top to bottom. In addition to this, each link is interactive, and when clicked on, displays information such as the proportion of genes inherited along this link, and their names.
Often, multiple equally parsimonious LGT scenarios are proposed by Horizstory in order to explain differences between a given gene tree and the reference topology. Lumbermill allows the user to restrict the display of individual LGT events to those suggested by a specified minimum fraction of the scenarios, such as 1.0 (strict consensus) or values greater than 0.5 (majority rule). One can also elect to omit the display of specific LGTs for some genes in Lumbermill, such as when the conflicting signal involving this gene is found to be due to some cause other than lateral gene transfer.
Putative LGT events are drawn as arrows originating from a donor (indicated by a circle) and terminating at a recipient. Since the exact time at which a transfer occurred cannot be determined, the relative order of multiple arrows on any given segment is irrelevant, as is the position of an arrow on the segment. In order to avoid clutter in busy regions of the tree, we chose to extend segments to provide more room. Such artificially extended segments are drawn as dotted lines so as not to confuse them with actual branch lengths (solid lines).
When genes have apparently been inherited from a taxon missing from the reference tree, we insert a basal group in the tree where appropriate. This donor, a contemporary clade absent from the current dataset and that may or may not have since gone extinct, collects all of the LGT events originating from outside of a represented clade. It is meant as a convenient catchall, and where multiple LGTs appear to originate from such a given basal group, it is understood that different donors may actually have contributed genes independently.
Our method allows for nested LGT scenarios, where a compound donor in the evolutionary scenario for a gene is not represented as an organismal clade in the reference tree. In this case, one cannot therefore point to an actual donor for the gene at this intermediary step in the scenario, since its identity is ambiguous, appearing to parent species from different clades. Lumbermill deals with such organismal assemblages by indicating several candidate donors all leading to the same target. It is assumed that the LGT event in question involved a single donor parenting one or more species in this assemblage, or when a basal group is also indicated, a single donor related to the parent of one or more of the species in the assemblage. These LGT involving such intermediate assemblages are represented using a double-headed arrow.
Results
We provide an example for four genes of the gamma-proteobacterial core compared with the gamma-proteobacterial species tree [34], deduced from a simultaneous study of 205 genes by Lerat et al. (Our previous critique of this tree [21] was focused on its statistical support, not its biological reasonableness). In order to illustrate the workings of Horizstory on a simple example, Figure 1 displays the analysis of a tree for the virulence factor MviN against this proposed gamma-proteobacterial species tree. The two steps required to transform the MviN tree into one compatible with the reference tree are in fact equivalent in the two scenarios shown (differing only in hypothetical intermediates), and can thus be represented more simply in the user-interface layer of our software, Lumbermill, described above.
Figure 1 Step-by-step example of the consolidation and rearrangement method of Horizstory. Where indicated as resolved, nodes were supported by a bootstrap value better than 80%. Two minimal scenarios were found in the analysis of the MviN tree shown in this example, both suggesting an equivalent pair of lateral gene transfer events (as illustrated in Lumbermill, Figure 2). Other (longer or unproductive) scenarios are not shown. A productive move is one where the gene tree gains further similarity to the reference tree, permitting further consolidation (see text). Ba, Buchnera aphidicola; Ec, Escherichia coli; Hi, Haemophilus influenzae; Pa, Pseudomonas aeruginosa; Pm, Pasteurella multocida; Se, Salmonella enterica; Vc, Vibrio cholerae; Wg, Wigglesworthia glossinidia; Xa, Xanthomonas axonopodis; Xc, Xanthomonas campestris; Xf, Xylella fastidiosa; YpC, Yersinia pestis CO92; YpK, Yersinia pestis KIM.
Among their 205 markers, Lerat et al. [34] identified two proteins, the virulence factor MviN and the biotin synthase BioB, for which the position of Pseudomonas aeruginosa conflicted with the species tree. Our example (Fig. 2) includes these two genes, with two more randomly chosen from among the remaining 203 of Lerat et al.'s set: a putative ORF and N6-adenine-specific methylase. It shows that some but not all recent divergences are reasonably supported by the set of four genes that contributed to this synthesis (e.g., the monophyly of the endosymbionts Buchnera aphidicola and Wigglesworthia glossinidia has a thick link), whereas deeper nodes are generally not so well supported, suggesting that this reference tree is not robust. This is consistent with the fact that mutational saturation of molecular sequences contributes to the erasure of phylogenetic information as time progresses [35], and several reports indicate that even with large amounts of data, the resolution of deep phylogenies will continue to be elusive [36,37].
Lumbermill displays both the support for a clade as well as the putative shuffling of genes with other clades, and can therefore begin to address this issue. For instance, Figs. 1 and 2 show that the ancestor of enterobacteria (represented by Y. pestis, E. coli, S. enterica, B. aphidicola and W. glossinidia) acquired its copy of MviN from a sister group of the gamma-proteobacteria by LGT, and subsequently the non-endosymbiotic enterobacteria donated this gene to P. aeruginosa. Fig. 2 also shows that a copy of BioB was laterally transferred from a sister group of the gamma-proteobacteria to the ancestor of P. multocida and H. influenzae.
Many estimates, described in the help file of Lumbermill (available online), are also implemented in Lumbermill to help investigate the distribution of the phylogenetic signal in the synthesis, to get information about the processes of LGT and a rough description of genetic mosaicism by node, as presented in Fig. 3. This last feature of Lumbermill addresses concerns raised by important publications [38-40], describing organisms as chimerae. Although a tree-like framework can still be appropriate to relate species for which no mosaicism is detected, a web [2] or a synthesis such as ours is required where mosaicism is evident. Fig. 3 displays the relative proportion of vertical inheritance (in blue) and of lateral transfer (in orange), phylogenetically assessed for each node of the synthesis. The information is displayed in two columns: the first represents lateral transfer occurring since the previous node only, while the second shows a cumulative estimate of all LGTs since the origin. Such summaries can address the issue relating phylogenetic distance with propensity for exchange, known to exist within "species" [41,42] and postulated for larger clades [43,44].
Figure 3 Estimate of genomic mosaicism. This application screenshot displays a table indicating the degree of vertical and of horizontal inheritance inferred for each node in the synthesis (see Fig. 2). The middle two columns of the table summarize the distribution of vertical (blue) and of lateral (orange) support for the segment immediately preceding each node, and the vertical and lateral support for the whole string of segments leading to a particular node, respectively. For example, B. aphidicola shows a 14% degree of accumulated mosaicism since the tree's root (from the small sample of genes used in this analysis), but no recent mosaicism (in the segment preceding its divergence from W. glossinidia, node 5A). Numerical values for the proportions are given by clicking on a bar. Node 1B is left entirely blank for want of phylogenetic evidence.
Conclusion
Horizstory and Lumbermill constitute a pair of phylogenetic tools to compare multiple trees, and to display this comparison in a useful and flexible format. The software is freely available to the community and downloadable from , and it will continue to be built upon in order to add further summary analyses. Already, we believe that they provide important information to the evolutionary biologist such as the frequency and direction of putative LGT events. We also gain information on the number of lateral exchanges undergone by each gene, the uniqueness of the LGT event (a unique event generates a thin horizontal link) or the fact that some species exchanged multiple genes on a regular basis (thick horizontal lines). Finally, we can assess the robustness of a tree's backbone. One could use these analyses in order to explore the evolution of phylogenetic signal and the relative power of phylogenetic methods to solve a given taxonomic issue. By proposing specific LGT events, our system may furthermore bracket the temporal coexistence of lineages, given that LGT can only occur between contemporary species. Otherwise, one cannot know much about the temporal coincidence of anything but extant species, where fossil evidence is lacking. This may be combined with ecological and biogeochemical hypotheses, and thereby help to understand the propagation of biological innovation throughout evolutionary time.
Availability and requirements
*Project name: Horizstory and Lumbermill
*Project home page: (all programs and test files alternatively available from EB upon request)
*Operating system(s): Platform independent (Horizstory requires a command line interface)
*Programming languages: Horizstory – C++, Lumbermill – Java
*Other requirements: C++ Compiler (Horizstory), Java 1.4.0 or higher (Lumbermill)
*License: Horizstory – GNU GPL, Lumbermill – none
*Any restrictions to use by non-academics: Horizstory – none, Lumbermill – none
note: Horizstory was designed and tested using gcc 3.3, available at
Application details
Whereas Lumbermill has an intuitive graphical user interface, Horizstory is a command-line C++ program whose options we describe below. Its signature is:
HorizStory -f listOfTreeFiles -r indexOfReferenceTree -m minimumBootstrap [-t fractionThreshold] [-bias twoLetterCode] [-s timeLimitPerTreeInSeconds] [-v]
where square brackets indicate optional parameters; the default fraction threshold is 0.0, the default bias is none, the default time limit is 86400 s (1 day), and the default verbose mode (the -v option) is false.
-f: Names a file that first specifies the number of entries (e.g., 7), then specifies that number of file names, representing a reference tree and trees to be tested against it, in any order. Non-reference trees must include bootstrap values for each node (or they are considered to be equal to 100%), but the reference tree need not include such information. Our method requires trees to be rooted, but can deal with multifurcated trees.
-r: The index, within the listOfTreeFiles, of the reference tree's file. The reference tree is assumed to be topologically resolved and rooted; where the user wishes an unresolved feature (a multifurcation) within the reference tree, a branch length of 0.0 can be specified. Bootstrap values may or may not be present in the tree; if present they are ignored unless 0, which again serves to collapse a node.
-m: Branch lengths of 0.0 in a candidate tree are automatically unresolved; otherwise bootstrap values are compared with this integer parameter in order to decide whether or not the node should be collapsed. For example, a weak feature (A,B) below threshold in ((A,B),C) collapses to (A,B,C).
-t: When trees are quite different, many pathways of rearrangement might be found that can transform a candidate tree into a topology compatible with the reference tree. In some of these pathways, exotic LGT events might be proposed that are not proposed by the majority of other pathways. This parameter, therefore, allows one to limit the summary output of the program to LGT that are suggested by at least a certain fraction of the pathways, e.g. 0.1.
-bias: One may wish to minimize or to maximize proposed events of LGT involving nested groups or basal groups. The analysis is first done to minimize the length of a rearrangement scenario, but then the output can be filtered by one of the following options if desired.
n: consider paths with the minimum number of nested groups, from among the set of shortest edit paths.
np: consider paths with the minimum number of nested groups, and of those, with the minimum number of phantom sisters.
nP: consider paths with the minimum number of nested groups, and of those, with the maximum number of phantom sisters.
N: consider paths with the maximum number of nested groups.
...and so on (Np, NP, p, pn, pN, P, Pn and PN).
-s: The user may specify a time limit for the analysis of each tree, in seconds. Tests on pairs of randomly generated trees indicate that processing time is exponential, O(cn), with the number of differences in the trees after initial consolidation, n, with c = 6.1, and on our test system (a 2.0-GHz PowerPC G5), with O being approximately 1.5 μs. Where a pair of trees must be compared, but the required time is prohibitive, the user may opt to increase the minimum bootstrap (-m) requirement, in order to compare only the most resolved portions of the trees.
-v: When specified, this parameter enables verbose output to the terminal window. It does not affect the output printed to file.
"Pruning files" are given the identical name of the referenceTree, but appended with ".pru", for global prunings, or are given the identical name of a candidate tree (appended with ".pru") for local prunings. The file specifies the number of taxa to be pruned, followed by that number of taxon names.
Authors' contributions
EB originated the study and provided critical feedback to DM, who implemented Lumbermill, and to RLC, who implemented Horizstory. WFD contributed to the writing of the manuscript.
Acknowledgements
This work was supported by GenomeAtlantic, and by a grant from CIHR (MOP4467).
==== Refs
Gribaldo S Philippe H Ancient phylogenetic relationships Theor Popul Biol 2002 61 391 408 12167360 10.1006/tpbi.2002.1593
Huson DH SplitsTree: analyzing and visualizing evolutionary data Bioinformatics 1998 14 68 73 9520503 10.1093/bioinformatics/14.1.68
Makarenkov V T-REX: reconstructing and visualizing phylogenetic trees and reticulation networks Bioinformatics 2001 17 664 668 11448889 10.1093/bioinformatics/17.7.664
Filee J Forterre P Laurent J The role played by viruses in the evolution of their hosts: a view based on informational protein phylogenies Res Microbiol 2003 154 237 243 12798227 10.1016/S0923-2508(03)00066-4
Lawrence JG Ochman H Reconciling the many faces of lateral gene transfer Trends Microbiol 2002 10 1 4 11755071 10.1016/S0966-842X(01)02282-X
Doolittle WF Phylogenetic classification and the universal tree Science 1999 284 2124 2129 10381871 10.1126/science.284.5423.2124
Nelson KE Clayton RA Gill SR Gwinn ML Dodson RJ Haft DH Hickey EK Peterson JD Nelson WC Ketchum KA McDonald L Utterback TR Malek JA Linher KD Garrett MM Stewart AM Cotton MD Pratt MS Phillips CA Richardson D Heidelberg J Sutton GG Fleischmann RD Eisen JA White O Salzberg SL Smith HO Venter JC Fraser CM Evidence for lateral gene transfer between Archaea and Bacteria from genome sequence of Thermotoga maritima Nature 1999 399 323 329 10360571 10.1038/20601
Huang J Mullapudi N Sicheritz-Ponten T Kissinger JC A first glimpse into the pattern and scale of gene transfer in Apicomplexa Int J Parasitol 2004 34 265 274 15003488 10.1016/j.ijpara.2003.11.025
Andersson JO Sarchfield SW Roger AJ Gene transfers from Nanoarchaeota to an ancestor of Diplomonads and Parabasalids Mol Biol Evol
Andersson JO Roger AJ Evolution of glutamate dehydrogenase genes: evidence for lateral gene transfer within and between prokaryotes and eukaryotes BMC Evol Biol 2003 3 14 12820901 10.1186/1471-2148-3-14
Daubin V Gouy M Perrière G Bacterial molecular phylogeny using supertree approach Genome Inform Ser Workshop Genome Inform 2001 12 155 164 11791234
Eulenstein O Chen D Burleigh JG Fernandez-Baca D Sanderson MJ Performance of flip supertree construction with a heuristic algorithm Syst Biol 2004 53 299 308 15205054 10.1080/10635150490423719
Semple C Daniel P Hordijk W Page RD Steel M Supertree algorithms for ancestral divergence dates and nested taxa Bioinformatics 2004 20 2355 2360 15073021 10.1093/bioinformatics/bth246
Creevey CJ McInerney JO Clann: investigating phylogenetic information through supertree analyses Bioinformatics 2005 21 390 2 15374874 10.1093/bioinformatics/bti020
Thorley JL Page RD RadCon: phylogenetic tree comparison and consensus Bioinformatics 2000 16 486 7 10871273 10.1093/bioinformatics/16.5.486
Lapointe FJ Cucumel G The average consensus procedure: combination of weighted trees containing identical or overlapping sets of taxa Syst Biol 1997 46 306 312
Steel M Dress AWM Bocker S Simple but fundamental limitations on supertree and consensus tree methods Syst Biol 2000 49 363 368 12118411 10.1080/10635159950173870
Thomas GH Wills MA Szekely T A supertree approach to shorebird phylogeny BMC Evol Biol 2004 4 28 15329156 10.1186/1471-2148-4-28
Baldauf SL Roger AJ Wenk-Siefert I Doolittle WF A kingdom-level phylogeny of eukaryotes based on combined protein data Science 2000 290 972 977 11062127 10.1126/science.290.5493.972
Gophna U Doolittle WF Charlebois RL Weighted genome trees: refinements and applications J Bacteriol
Bapteste E Boucher Y Leigh J Doolittle WF Phylogenetic reconstruction and lateral gene transfer Trends Microbiol 2004 12 406 411 15337161 10.1016/j.tim.2004.07.002
Zhaxybayeva O Gogarten JP An improved probability mapping approach to assess genome mosaicism BMC Genomics 2003 4 37 12974984 10.1186/1471-2164-4-37
Koski LB Golding GB The closest BLAST hit is often not the nearest neighbor J Mol Evol 2001 52 540 542 11443357
Daubin V Lerat E Perrière G The source of laterally transferred genes in bacterial genomes Genome Biol 2003 4 R57 12952536 10.1186/gb-2003-4-9-r57
Zhaxybayeva O Lapierre P Gogarten JP Genome mosaicism and organismal lineages Trends Genet 2004 20 254 260 15109780 10.1016/j.tig.2004.03.009
Levasseur C Lapointe FJ War and peace in phylogenetics: a rejoinder on total evidence and consensus Syst Biol 2001 50 881 891 12116638 10.1080/106351501753462858
Page RD GeneTree: comparing gene and species phylogenies using reconciled trees Bioinformatics 1998 14 819 20 9918954 10.1093/bioinformatics/14.9.819
Shimodaira H An approximately unbiased test of phylogenetic tree selection Syst Biol 2002 51 492 508 12079646 10.1080/10635150290069913
Lecointre G Rachdi L Darlu P Denamur E Escherichia coli molecular phylogeny using the incongruence length difference test Mol Biol Evol 1998 15 1685 1695 9866203
Walsh DA Bapteste E Kamekura M Doolittle WF Evolution of the RNA polymerase B' subunit Gene (rpoB') in Halobacteriales: a complementary molecular marker to the SSU rRNA gene Mol Biol Evol 2004 21 2340 2351 15356285 10.1093/molbev/msh248
Zelwer M Daubin V Detecting phylogenetic incongruence using BIONJ: an improvement of the ILD test Mol Phylogenet Evol 2004 33 687 93 15522796
Felsenstein J Confidence limits on phylogenies: an approach using the bootstrap Evolution 1985 40 783 791
Addario-Berry L Hallett M Lagergren J Towards identifying lateral gene transfer events Pac Symp Biocomput 2003 8 279 290 12603035
Lerat E Daubin V Moran NA From Gene Trees to Organismal Phylogeny in Prokaryotes:The Case of the gamma-Proteobacteria PLoS Biol 2003 1 E19 12975657 10.1371/journal.pbio.0000019
Moreira D Philippe H Molecular phylogeny: pitfalls and progress Int Microbiol 2000 3 9 16 10963328
Teichmann SA Mitchison G Is there a phylogenetic signal in prokaryote proteins? J Mol Evol 1999 49 98 107 10368438
Philippe H Snell EA Bapteste E Lopez P Holland PW Casane D Phylogenomics of eukaryotes: impact of missing data on large alignments Mol Biol Evol 2004 21 1740 1752 15175415 10.1093/molbev/msh182
Ribeiro S Golding GB The mosaic nature of the eukaryotic nucleus Mol Biol Evol 1998 15 779 788 9656480
González V Bustos P Ramírez-Romero MA Medrano-Soto A Salgado H Hernández-González I Hernández-Celis JC Quintero V Moreno-Hagelsieb G Girard L Rodríguez O Flores M Cevallos MA Collado-Vides J Romero D Dávila G The mosaic structure of the symbiotic plasmid of Rhizobium etli CFN42 and its relation to other symbiotic genome compartments Genome Biol 2003 4 R36 12801410 10.1186/gb-2003-4-6-r36
Rivera MC Lake JA The ring of life provides evidence for a genome fusion origin of eukaryotes Nature 2004 431 152 155 15356622 10.1038/nature02848
Lawrence JG Catalyzing bacterial speciation: correlating lateral transfer with genetic headroom Syst Biol 2001 50 479 496 12116648
Lawrence JG Gene transfer in Bacteria: speciation without species? Theoret Pop Biol 2002 61 449 460 12167364 10.1006/tpbi.2002.1587
Gogarten JP Doolittle WF Lawrence JG Prokaryotic evolution in light of gene transfer Mol Biol Evol 2002 19 2226 2238 12446813
Koonin EV Makarova KS Aravind L Horizontal gene transfer in prokaryotes: quantification and classification Annu Rev Microbiol 2001 55 709 742 11544372 10.1146/annurev.micro.55.1.709
| 15819979 | PMC1087482 | CC BY | 2021-01-04 16:37:17 | no | BMC Evol Biol. 2005 Apr 8; 5:27 | utf-8 | BMC Evol Biol | 2,005 | 10.1186/1471-2148-5-27 | oa_comm |
==== Front
BMC Fam PractBMC Family Practice1471-2296BioMed Central London 1471-2296-6-141583109910.1186/1471-2296-6-14Research ArticleIdentifying future models for delivering genetic services: a nominal group study in primary care Elwyn Glyn [email protected] Adrian [email protected] Rachel [email protected] Peter [email protected] Jonathon [email protected] Centre for Health Sciences Research, Cardiff University, 56 Park Place, Cardiff, CF10 3AT, Wales2 Institute of Medical Genetics, Cardiff University, Heath Park, Cardiff, CF14 4XN, Wales3 Primary Care Group, University of Wales Swansea, SA2 8PP, Wales2005 14 4 2005 6 14 14 11 11 2004 14 4 2005 Copyright © 2005 Elwyn et al; licensee BioMed Central Ltd.2005Elwyn 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
To enable primary care medical practitioners to generate a range of possible service delivery models for genetic counselling services and critically assess their suitability.
Methods
Modified nominal group technique using in primary care professional development workshops.
Results
37 general practitioners in Wales, United Kingdom too part in the nominal group process. The practitioners who attended did not believe current systems were sufficient to meet anticipated demand for genetic services. A wide range of different service models was proposed, although no single option emerged as a clear preference. No argument was put forward for genetic assessment and counselling being central to family practice, neither was there a voice for the view that the family doctor should become skilled at advising patients about predictive genetic testing and be able to counsel patients about the wider implications of genetic testing for patients and their family members, even for areas such as common cancers. Nevertheless, all the preferred models put a high priority on providing the service in the community, and often co-located in primary care, by clinicians who had developed expertise.
Conclusion
There is a need for a wider debate about how healthcare systems address individual concerns about genetic concerns and risk, especially given the increasing commercial marketing of genetic tests.
==== Body
Background
'Imagine what it would be like if doctors could look at your medical future' says an advert in the Harvard Business Review [1]. If companies are already marketing the possibility of a genetic manipulated future, is it not time doctors considered how to manage what will become a growing area of work? Genetics will alter the face of medicine, as the inherited components of common diseases and cancers are uncovered and patients' awareness of susceptibility increases. This trend is likely to lead to more individuals approaching primary care for guidance, for genetic risk assessment or counselling and perhaps testing [2,3]. Many practitioners are confident that generalists can absorb this demand by increasing their knowledge and making use of new software [4,5]. Others are sceptical and fear primary care clinicians will not want the additional task of providing genetic advice [6], adamant that this new role – both in time and effort – goes beyond what can be realistically provided. Patients are not concerned with such debates: they look for convenient access to effective services [7]. Should primary care respond to these demands and if so, how? If not, are we guilty of allowing technological developments in predictive testing and commercial pressures overtake the ability of services to react appropriately, leaving individuals with genetic concerns unsupported? [8,9]
Genetics clinics are experiencing workload increases [10]. Referrals to cancer genetic services are a particular area of demand as predictive genetic testing for breast, ovarian and colorectal cancer becomes more widely known [10]. To meet this demand, genetic services are experimenting with new methods of service delivery. A number of models are reported and include initiatives such as initial risk assessments by mailed questionnaire or telephone interview or the use of letters to transmit risk information to referred patients without brining them to clinic [11,12]. Some services are examining the use of non-medical genetic counsellors to assess family pedigrees and are using nurses to undertake specialised genetic counselling. Some services have explored the use of video-conferencing in order to increase access to populations that live in rural areas [13]. Nevertheless, the basic service delivery model is still one predicated on specialist care: general practitioners refer patients to the secondary (or tertiary) care sector. As yet, there has been no guideline that indicates that some genetic issues should be managed in primary are, although guidelines are emerging that ask generalists to stratify risk by categorising patients according to risk and to only refer those above a suggested threshold. However, as demand for genetic services continues to grow, the viability of this service delivery model needs to be examined.
As elsewhere in the UK, genetic services in Wales are provided at a regional level, using a distributed clinic approach in three centres (North, West and the South East). Referrals are accepted from a range of sources, and for cancer genetic referrals a triage system involving postal questionnaire and telephone assessment is used to stratify patients into low, medium and high-risk groups. Given increasing demand however, there is speculation that genetics could be integrated into community settings, perhaps in family practice, close to kinship relationships and in a context of continuous care. However, there are also concerns about patients living 'at risk' with no obvious means of support [10]. Genetic risk assessment is a task requiring a detailed history, and where there are uncertainties regarding neoplasm in other family members, validation of disease by pathological confirmation is necessary. To determine individual risk, family members have to be contacted and blood samples obtained for DNA analysis from living affected relatives. This process is lengthy, complex and depends on effective communication between health care professionals, their patients, family members and between health professionals themselves. Is it feasible that primary care services can be re-designed to enable genetic risk assessment and counselling for those identified as moderate to high risk to take place and then referred onwards to the appropriate specialist service?
This study aimed to enable general practitioners to generate possible delivery models for genetic services and then critically assess these different possibilities over the next five to ten years. To facilitate the emergence of a wide range of options and possibilities a modified nominal group technique was employed.
Methods
General practitioners were invited to discuss developments in genetics, emphasising cancer genetics, and to consider how services could be re-designed to meet anticipated growth in patient demand for genetic advice, counselling and further management. To obtain a varied sample, meetings were arranged in three different locations (Swansea, Newport and Cardiff). Two mailings were circulated to all practitioners in the relevant catchment areas using details available to the postgraduate offices. Approximately 150 practitioners were mailed each area.
The meetings were described as having two broad aims: firstly, to inform general practitioners about developments in genetics and how these might translate into patient concerns. Secondly, to involve practitioners in considering how genetic services could best meet this anticipated increased demand. The meetings took place between March and September 2003, at broadly the same time as the implementation of the new contract for general practice in the UK [14].
The structure of the meetings was standardised. A specialist in medical genetics (JG) provided an overview of recent advances in genetics and advised participants of the developments in predictive testing and pharmacogenetics. The predicted increasing demand for genetic advice was discussed. The presentation described the process of assessing family history, verifying verbal reports of diseases in other family members, assessing and communicating genetic risk to patients.
After this presentation, a modified nominal group technique was used [15]. Participants were asked to write, without conferring, a list of possible service delivery models that could be designed to provide genetic services to patients. Participants were asked to think creatively about news kinds of services and, for the purposes of the exercise, to put aside concerns about funding or the future of general practice. A facilitator then asked each participant to describe the models they proposed, one at a time, and they were outlined on flip charts. At this stage, the facilitator (GE) clarified and categorised the models, to remove duplications and arrived at an agreed list of service delivery designs. After this list had been agreed, participants were asked to conduct brief (10–20 minute) small group discussions in order to discuss the advantages and disadvantages of each model. Each practitioner then independently ranked the agreed models and the results were shared and discussed. In order to compare the results of the meetings, rankings given to different service delivery models were given inverse scores, and totals calculated to determine overall rankings.
Results
Three events were held at which a nominal group process was undertaken immediately after a short standardised presentation about recent developments in genetics and cancer genetics. A total of 37 practitioners participated: they represent practitioners who were mostly in the mid-careers and were from a range of practice sizes. Compared to other similarly advertised events, attendance was approximately 50% lower: details of the attendees are provided in Table 1.
Table 1 Details of general practitioner participants at three nominal group events
Meeting Area Number at meeting Gender, (Mean number of years in practice) Mean Practice Size (Whole Time Equivalent Doctors)
South West Wales 13 9 male, 4 female (17). 5
South East Wales 9 5 male, 4 female (13). 2
Cardiff 15 6 Male, 6, female, 3 not specified, (14). 4.6
Participants proposed lists of service models ranging from patient self-referral to telephone-based services such as NHS Direct, to maintaining existing arrangements where general practitioners continue to act as gatekeepers to other services. After clarifying the nature of each proposed model, the list was summarised and distinctive service delivery models were given a short descriptive names. Small group discussions were conducted to examine their pros and cons. At this point, each individual practitioner was asked to rank his or her preferred service models. The hypothetical service delivery models, their ranking at each meeting and overall ranking are provided in Table 2.
Table 2 Genetic service delivery models proposed and ranked
Service Models Meeting 1 Meeting 2 Meeting 3
Rank Score Rank Score Rank Score Score Total
Community based service provided by genetic counsellors, not managed by general practice, but could be located in practices or local community centres to provide local patient assessment and advice. 1 8 1 8 3 6 22
Enhanced primary care: a service located within primary care, with specialists in genetic risk assessment, with support made possible by information technology and software applications. 2 7 3 6 1 8 21
Special 'genetic' clinics: this model was suggested so that the privacy and discretion analogous to 'genitourinary clinics' was built in, and where self-referral is possible and anonymity and confidentiality respected. 3 6 4 5 2 7 18
Traditional gatekeeper model: where general practitioners undertake an initial assessment, using standardised referral guidelines, and refer patients who are not categorised as 'low' risk. 4 5 2 7 4 5 17
Direct access telephone service: a 'genetics direct' model where patients have their genetic pedigrees assessed by counsellors with assess to pedigree software tools. 5 4 5 4 5 4 12
Drop in service for genetic assessment: e.g. similar to the Citizen Advice Bureau model. 6 3 - - - - 3
Private service: patients with concerns are directed to commercial providers either in the UK or elsewhere. 7 2 - - - - 2
Pharmacy led service: patients with concerns are directed to pharmacists, who could also undertake pharmacogenetic profile testing and offer lifestyle advice. 8 1 - - - - 1
Table 2 illustrates the range of possible models generated during the nominal groups and shows that general practitioners are willing to consider a range of service delivery methods. Compared to the other two meetings, practitioners at the first meeting (south west Wales) generated many innovative approaches, including a telephone-based service integrated with an on-line family pedigree software and a suggestion that community pharmacists might wish to develop a role as genetic counsellors. Another suggestion was a 'drop in' centre as an attempt to 'de-medicalise' the assessment of an individual's genetic make-up. Some practitioners were aware of commercial companies selling genetic tests on the Internet and proposed that any initiatives linked to these should be privately funded and not covered as part of the state funded health care system.
The results of the ranking exercise showed all participants were agreed on one point: that the current gatekeeper model of primary care was not going to be able to deliver the genetic assessment, counselling and possibly testing that patients would require. However, participants felt genetic services should remain close to primary care, either in community settings or co-located and delivered by practitioners with specialist skills in this area, such as a general medical or nurse practitioner with a special interest. These models, including the current delivery model, were ranked higher than innovations based on telephone-based assessment alone or those located in different contractor professions. No model was a clear favourite among all three groups.
Discussion
Principal findings
Those general practitioners who accepted the invitation to hear about genetic developments iand take part in the nominal group process did not believe current systems were sufficient to meet anticipated patient demand for genetic services. Within the group process practitioners proposed a wide range of different service options, although no single option emerged as a clear preference for those participating. Surprisingly no argument was put forward for genetic assessment and counselling being central to family practice, neither was there a voice for the view that the family doctor should become skilled at advising patients about predictive genetic testing and be able to counsel patients about the wider implications of genetic testing for patients and their family members.
The views emerging from the groups reflected those of practitioners who are in routine NHS practice and therefore may not be considering the wider or future impact of not undertaking genetic assessment in primary care. Nevertheless, all the preferred models put a high priority on providing the service in the community, and often co-located with general practice. Although not directly addressed in the suggested service models, it was clear that the general practitioners had an open mind about which professional group would be best placed to undertake genetic counselling. Their understanding of the current secondary care model was that nurses with special training undertake the assessments.
Strengths and weaknesses
The use of the nominal group process is strength of this study. It is a recognised means of allowing participants to give free rein to ideas, without constraint. The only limitation on the process is the knowledge and experience of those taking part in the group. It was unfortunate that the sample size was small. On the basis of prior attendance at postgraduate events similar to this one, 60 participants had been expected. The small sample size may be explained by the subject being given a low priority: this area of practice is not yet seen as having urgency in the mind of service based general practitioners. Participants had noted a slight increase demand for genetic advice, particularly among women concerned about breast or ovarian cancer, however increased requests for other types of predictive genetic testing had been experienced. Some will regard the non-specialist perspective of this work as a weakness but the study was purposefully designed to obtain the preferences of service-based general practitioners as a 'bottom-up' exercise to identify the delivery models felt to be appropriate and applicable in the evolving primary care context.
Results in context
Placed in the context of publications that describe the expected impact of predictive genetic testing [16,17] there are few studies in which the effectiveness of different service delivery models has been examined. Holloway reported a study demonstrating the economy of using postal questionnaires compared to specialist nurse interviews as a means of assessing familial breast cancer risk [12]. Campbell reported a cluster randomised trial of a GP based genetic clinic, versus the normal practice of referral to a regional service and showed a larger increase in referral rates when clinics were based in primary care and that patients from the GP clinic had an inappropriate level interest and expectation of the appropriateness of genetic testing [10]. Elwyn reported reactions to a nurse-led triage system [18]. These studies did not consider other possible service delivery models. The work reported in this study provides a wider canvas of possible models and novel approaches. Genetic counselling and testing services need to combine accurate, comprehensive genetic risk assessment with methods that can be both local and sensitive to family contexts. It may be possible for general practice in the UK to develop expertise in genetic assessment, using the possibility of contracting for enhanced services.
Conclusion
As a result of developments in genetics, there are increasing demands being made on primary care and genetic services to address patient concerns and manage requests for genetic advice, risk assessment and testing. Innovations in the ways services could be delivered have started to appear but there is a lack of discussion and planning about how the NHS intends to deal with the impact of the new genetics. General practitioners agree that the current referral processes and structures are unlikely to meet anticipated needs, but they do not have agreement about how services could be re-designed to meet anticipated demand. The practitioners suggested a range of potential service delivery models: the common thread among those which were ranked highest was that the service should be located close to communicates and work in close liaison, or embedded in, primary care provision. There is a need for a wider debate about how healthcare systems address how individual concerns about genetic risk are counselled and managed, especially given the likely commercial marketing of genetic tests. This study demonstrates that there is no obvious preferred solution to the problem of designing a service or system to provide genetic advice and assessment to an increasing number of patients and an implicit sense of scepticism about the likely impact of the new genetics, echoing other commentators [19]. It was also unclear how clinicians were expected to integrate genetic information about individuals and their relatives across an integrated electronic patient record shared between healthcare organisations – an area where concerns about data security and confidentiality abound. General practitioners are willing to suggest a range of models but there is no clear preference. From the groups, it was presumed that as a routine service, primary care practitioners would not undertake detailed genetic assessments but were open to the concept of a specialist in this area operating in a primary care arena. The practitioners did not differentiate between genetic concerns that were likely to be more frequent and therefore might be considered to become part of the primary care service, such as the stratification of women with a family history of breast or ovarian cancer into risk categories and then referring those who were calculated to be above population risk. Whether the aggregated views of this selected sample would find resonance among other stakeholders needs to be explored.
Authors' contributions
GE had the original idea for the study and all authors developed the proposal. PD and GE led the nominal group sessions. GE, AE and PD contributed to the analysis and interpretation of the data. GE wrote the first draft of the paper and all authors contributed to subsequent drafts. GE acts as guarantor.
Competing interests
The author(s) declare that they have no competing interests.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This study was funded by the Tenovus Foundation. The views expressed in the paper represent those of the authors and not necessarily those of the funders. Wendy Jones provided support and group facilitation at the events, Jenny Kowalczuk provided assistance during the drafting of the article.
==== Refs
Harvard Business Review 2004 82
Robins R Metcalfe S Integrating genetics as practices of primary care Soc Sci Med 2004 59 223 233 15110415 10.1016/j.socscimed.2003.10.025
Rich EC Burke W Heaton CJ Haga S Pinsky L Short MP Acheson L Reconsidering the family history in primary care J Gen Intern Med 2004 19 273 280 15009784 10.1111/j.1525-1497.2004.30401.x
Watson EK Shickle D Qureshi N Emery J Austoker J The 'new genetics' and primary care: GPs' views on their role and their educational needs Family Practice 1999 16 420 425 10493715 10.1093/fampra/16.4.420
Emery J Walton R Murphy M Austoker J Yudkin P Chapman C Coulson A D G Fox J Computer support for interpreting family histories of breast and ovarian cancer in primary care: comparative study with simulated cases BMJ 2000 321 28 32 10875832 10.1136/bmj.321.7252.28
Kumar S Resisting revolution: generalism and the new genetics Lancet 1999 354 1992 1993 10622314 10.1016/S0140-6736(99)04073-8
Iredale R Brain K Edwards L Gray J France E The information and support needs of women at high risk of familial breast and ovarian cancer: how can cancer genetic services give patients what they want? Familial Cancer 2003 2 119 121 14649669 10.1023/A:1025706907045
Kinmonth AL Reinhard J Bobrow M Pauker S The new genetics: Implications for clinical services in Britain and the United States BMJ 1998 316 767 770 9529418
Emery J Hayflick S The challenge of integrating genetic medicine into primary care BMJ 2001 322 1027 1030 11325768 10.1136/bmj.322.7293.1027
Campbell H Holloway S Cetnarskyj R Anderson E Rush R Fry A Gorman D Steel M Porteous M Referrals of women with a family history of breast cancer from primary care to cancer genetics services in South East Scotland Br J Cancer 2003 89 1650 1656 14583764 10.1038/sj.bjc.6601348
Phelps C Platt K France L Gray J Iredale R Delivering information about cancer genetics via letter to patients at low and moderate risk of familial cancer: a pilot study in Wales Familial Cancer 2004 3 55 59 15131407 10.1023/B:FAME.0000026838.92085.ec
Holloway S Porteous M Cetnarskyj R Anderson E Rush R Fry A Gorman D Steel M Campbell H Patient satisfaction with two different models of cancer genetic services in south-east Scotland Br J Cancer 2004 90 582 589 14760368 10.1038/sj.bjc.6601562
Gray J Brain K Iredale R Alderman J France E Hughes H A pilot study of telegenetics J Telemed Telecare 2000 6 245 247 11027129 10.1258/1357633001935329
General Practitioners Committee Your contract, your future 2002 London, British Medical Association
Jones J Hunter D Mays N and Pope C Using the Delphi and nominal group technique in health services research Qualitative Research in Health Care 1999 London, BMJ Books
Harper PS Hughes HB Raeburn JA Clinical genetics services into the 21st century. Summary of a report of the Clinical Genetics Committee of the Royal College of Physicians J R Coll Physicians Lond 1996 30 296 301 8875372
Knottnerus JA Community genetics and community medicine Family Practice 2004 20 601 606 10.1093/fampra/cmg519
Elwyn G Iredale R Gray J Reactions of general practitioners to a triage controlled referral system for cancer genetics Family Practice 2002 19 65 71 11818352 10.1093/fampra/19.1.65
Holtzman NA Marteau TM Will genetics revolutionize medicine? N Engl J Med 2000 343 141 144 10891526 10.1056/NEJM200007133430213
| 15831099 | PMC1087483 | CC BY | 2021-01-04 16:29:13 | no | BMC Fam Pract. 2005 Apr 14; 6:14 | utf-8 | BMC Fam Pract | 2,005 | 10.1186/1471-2296-6-14 | oa_comm |
==== Front
BMC GenetBMC Genetics1471-2156BioMed Central London 1471-2156-6-201582320210.1186/1471-2156-6-20Research ArticleDifferential selection and mutation between dsDNA and ssDNA phages shape the evolution of their genomic AT percentage Xia Xuhua [email protected] Kwok Yung [email protected] Department of Biology, University of Ottawa, Ottawa, Ontario, K1N 6N5, Canada2 Department of Microbiology, University of Hong Kong, Hong Kong2005 11 4 2005 6 20 20 21 9 2004 11 4 2005 Copyright © 2005 Xia and Yuen; licensee BioMed Central Ltd.2005Xia and Yuen; 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
Bacterial genomes differ dramatically in AT%. We have developed a model to show that the genomic AT% in rapidly replicating bacterial species can be used as an index of the availability of nucleotides A and T for DNA replication in cellular medium. This index is then used to (1) study the evolution and adaptation of the bacteriophage genomic AT% in response to the differential nucleotide availability of the host and (2) test the prediction that double-stranded DNA (dsDNA) phage should exhibit better adaptation than single-stranded DNA (ssDNA) phage because the rate of spontaneous deamination, which leads to C→T or C→U mutations depending on whether C is methylated or not, is about 100-fold greater in ssDNA than in dsDNA.
Results
We retrieved 79 dsDNA phage and 27 ssDNA phage genomes together with their host genomic sequences. The dsDNA phages have their genomic AT% better adapted to the host genomic AT% than ssDNA phage. The poorer adaptation of the ssDNA phage can be partially accounted for by the C→T(U) mutations mediated by the spontaneous deamination. For ssDNA phage, the genomic A% is more strongly correlated with their host genomic AT% than the genomic T%.
Conclusion
A significant fraction of variation in the genomic AT% in the dsDNA phage, and that in the genomic A% and T% of the ssDNA phage, can be explained by the difference in selection and mutation between them.
==== Body
Background
We first present a simple model of DNA replication to show that the genomic AT% of rapidly replicating bacterial species is indicative of the relative availability of nucleotides A and T in the bacterial cell. By using the genomic AT% of bacterial species as an index of AT availability, we study how bacteriophage would evolve in response to the differential AT availability in different bacterial hosts. We expect natural selection to favour the evolution of phage genomic AT% to take advantage of the differential AT availability in different hosts. In particular, given that the rate of spontaneous deamination, which results in C→T mutation (when the C is methylated or C→U mutations (when the C is not methylated), is 100-fold higher in single stranded DNA than in double-stranded DNA [1], we expect the adaptation of phage genomic AT% to their host cellular environment to be more disrupted by the C→T(U) mutations in single-stranded DNA (ssDNA) phage than in double-stranded DNA (dsDNA) phage.
Designate the amount of dATP, dCTP, dGTP and dTTP available for DNA replication as VdA, VdC, VdG and VdT, respectively. Note that these are abstract terms and may not correspond to the cellular concentration of dNTPs or rNTPs. Suppose a single-stranded DNA genome of length L is composed of A, C, G, and T with frequencies NA, NC, NG and NT, respectively (NA + NC + NG + NT = L). The polymerization reaction is characterized as
where Mn• stands for an elongating (or propagating in chemistry terminology) DNA strand with n monomer residues, M is the monomer, and kp is the propagating constant. According to the law of mass action, and assuming that kp is the same for adding any of the four nucleotides to the elongating chain, the elongation rate (r) during DNA replication can be modelled as
Bacterial species often need, and typically are selected, to replicate rapidly. For example, E. coli in unlimited culture conditions can replicate once every 20 minutes. It is therefore reasonable to assume natural selection to operate on increasing r for such organisms. According to equation (2), if VdA is the largest, then r is increased with increasing NA and decreasing NG, NC and NT, with the constraint of NA + NC + NG + NT = L. The maximum r is achieved when NA = L and NC = NG = NT = 0. This means that, in order to maximize r with differential nucleotide availability, the genomic nucleotide usage should evolve to adapt to the availability of nucleotide availability by maximizing the usage of the nucleotide of the highest availability. Similar conclusions have also been derived elsewhere on optimization at the molecular level [2].
One should note that the model above does not consider the effect of differential depletion of the nucleotides. For example, consider that VdA is the largest among the four at the beginning of DNA replication. If a rapidly replicating genome is made entirely of A, then A will be differentially depleted leading to a reduced VdA which consequently may become smaller than VdC, VdG and VdT. This means that the replication of the remaining A-rich part of the genome would be slow, thus compromising the statement above that "The maximum r is achieved when NA = L and NC = NG = NT = 0". However, the qualitative conclusion that, if VdA is larger than VdC, VdG and VdT, then NA should be larger than NG, NC and NT remains correct.
When VdC = VdG = VdA = VdT = V, then equation (2) becomes:
so that r is independent of NA, NC, NG, and NT. This might be interpreted to mean that, with equal availability of the nucleotides for DNA replication, there is no selection on genomic nucleotide usage and genomic nucleotide frequencies can vary freely. However, the replication of a large, rapidly elongating and AT-rich genome may differentially reduce VdC, VdG, VdA, and VdT. For example, rapid replication of a large AT-rich genome will reduce VdA and VdT and increase the time for adding the remaining A and T to the elongation chain. Thus, even with VdC = VdG = VdA = VdT = V at the beginning of the replication, we would still expect the genomic AT% to be near 50% instead of fluctuating to extreme values.
For a double-stranded genome where NA = NT = NAT and NC = NG = NCG, equation (2) becomes
If VdA*VdT >> VdC*VdG, then increasing NAT in the genome will increase r, with the maximum r achieved when NAT = L and NCG = 0, i.e., the genome should evolve towards AT-richness. Again, this assumes no differential depletion of A and T and should be interpreted qualitatively to mean that, with VdA*VdT >> VdC*VdG, we should have NAT > NCG.
If VdA*VdT = VdC*VdG, then r becomes independent of NAT and NGC. However, this again does not necessarily mean that there is no selection to constrain genomic AT% and that genomic AT% can consequently vary freely. As we have argued before, a large, rapidly replicating and AT-rich genome will differentially reduce nucleotides A and T and lead to VdA*VdT << VdC*VdG which is unfavourable for replicating an AT-rich genome. Thus, with VdA*VdT = VdC*VdG, we expect the genomic AT% to be near 50% instead of fluctuating to extreme values.
In summary, we expect an extremely GC-rich bacterial genome to indicate high VdC*VdG, an extremely AT-rich bacterial genome to indicate high VdA*VdT, and a bacterial genome with GC% = 50% to indicate (VdA*VdT) ≈ (VdC*VdG).
Based on the reasoning above, we may infer that different genomic AT% values in different bacterial species indicate different AT availability in the cells of these bacterial species. By using the genomic AT% of bacterial species as an index of AT availability, we now study how bacteriophage genomic GC% evolve in response to different nucleotide availability in different hosts.
Assuming that it is beneficial for the phage to replicate its genome rapidly, we can make two testable predictions. First, a phage genome should evolve to become AT-rich in a host with a high genomic AT% (indicating VdA*VdT >> VdC*VdG in its cell), and GC-rich in a host with a low genomic AT% (indicating VdA*VdT << VdC*VdG in its cell). This will lead to a positive correlation between the phage genomic AT% and the host genomic AT%. Such a correlation has in fact been known for a long time [3]. Second, because the rate of spontaneous deamination, which leads to C→T or C→U mutations depending on whether C is methylated or not, is about 100-fold higher in the ssDNA than in dsDNA [1], we expect such mutations to reduced the effectiveness of natural selection to optimize the genomic AT% of the ssDNA phage in response to their host genomic AT%. In particular, with low host AT availability, natural selection should favour the reduction of the phage genomic AT%, but the C→T(U) mutation mediated by the spontaneous deamination in the ssDNA phage would counteract against natural selection and increase the genomic AT% of the ssDNA phage. In addition, because genomic A% and T% can evolve independently in ssDNA phage, we can specifically predict an increase in the genomic T% in ssDNA phage without an associated increase in the genomic A%. We will test these predictions.
Results
The positive relationship between the phage genomic AT% and their host genomic AT% is shown separately for the dsDNA and ssDNA phages (Fig. 1). Such a positive relationship itself is trivial because the relationship has been known for nearly 40 years [3]. However, the difference between the dsDNA and ssDNA phages is scientifically significant. The regression line for the ssDNA phage has a higher intercept and a lower slope than that for the dsDNA phage (Fig. 1).
Figure 1 Relationship between the phage genomic AT% and the host genomic AT%. Data points for ssDNA and dsDNA phages are plotted separately with their respective linear regression lines.
We can employ the general linear model (GLM) to test the statistical difference of the two regression lines:
where PhageAT and HostAT are AT% of the phage and host genomes, respectively, and PhageType is of two categories (i.e., dsDNA and ssDNA) coded by a binary dummy variable with 0 for ssDNA and 1 for dsDNA. If B2 and B3 are not significantly greater than 0, then there is no significant difference between the two regression lines.
The parameters of the general model in equation (5) can be evaluated by the GLM procedure in SAS [4], which uses the method of least-squares to fit general linear models. The resulting B2 is -10.156 which is statistically significant (t = 2.83, p = 0.0028). Similarly, B3 = 0.135, t = 2.04, p = 0.0221. The other parameters are B0 = 18.503 (t = 6.08, p < 0.0001), and B1 = 0.734 (t = 12.93, p < 0.0001)
Given the evaluated parameters in equation (5), the relationships between PhageAT and HostAT for dsDNA and ssDNA phages are
The increased intercept and decreased slope in the ssDNA phage relative to the dsDNA phage is easy to interpret in light of the finding that the rate of spontaneous deamination, which increases the C→T(U) mutation rate, is about 100-fold higher in ssDNA than in dsDNA [1]. This spontaneous deamination features prominently among all other factors contributing to the degradation of DNA [5]. When host genomic AT% is low (the left extreme of Fig. 1), indicating low availability of nucleotides A and T in the cellular medium according to equations (2) and (4), natural selection should cause the phage genome to reduce its AT%, but the C→T(U) mutation mediated by the high rate of spontaneous deamination in the ssDNA phage goes against natural selection and increases phage genomic AT%. In other words, the C→T(U) mutations reduce the effect of the natural selection to push the phage genomic AT% downwards. This would raise the intercept and decrease the slope of the regression for the ssDNA phage relative to the regression line for the dsDNA phage.
Note that the C→T(U) mutations act in the same direction as the natural selection when the host genomic AT% is high, indicating high availability of nucleotides A and T in the cellular medium according to equations (2) and (4). In this case, natural selection should favour phage genomes to become AT-rich, and the C→T(U) mutation mediated by the high rate of spontaneous deamination in the ssDNA phage also increases phage AT%, i.e., the two act in the same direction. Such an interpretation is consistent with the right side of Fig. 1 in which few points are below the regression line and with little scatter above and below the regression line, especially when the host genomic AT% is extremely high.
To further substantiate this interpretation, we can test whether the increased intercept and decreased slope for the regression line of the ssDNA phage in Fig. 1 is really due to an increase in the genomic T% instead of the genomic AT%. This can be done because A and T do not need to be equal to each other in number for ssDNA. We expect an increased genomic T% but not genomic A% in the ssDNA phage. Such an inference is consistent with plotting the genomic A% and T% separately for the ssDNA phage against the host AT% (Fig. 2).
Figure 2 The genomic A% and T% of the ssDNA phage plotted against their host genomic AT%. The regression lines are separately fitted for the phage genomic A% and T%, respectively
We can test the statistical significance of the difference between the two regression lines in Fig. 2 by using the general linear model similar to the approach in equations (5) and (6). The regression line for the genomic T% has a significantly increased intercept (P = 0.0068, one-tailed test) and decreased slope (P = 0.0323, one-tailed test). Also, the relationship between the phage genomic A% and the host genomic AT% is stronger than that between the phage genomic T% and the host genomic AT%, with the Pearson correlation coefficient being 0.87857 for the former and 0.60249 for the latter.
The results above corroborate our interpretation that C→T(U) mutations contribute significantly to the relationship in nucleotide frequency distribution between the phage genome and the host genome. In particular, the increased intercept and decreased slope for ssDNA phage in Fig. 1 can be largely attributed to the C→T(U) mutations mediated by the spontaneous deamination.
The pattern in Fig. 2, however, can have an alternative explanation. First, it is important to note that the host genomic AT% is only indicative of VdA*VdT. If VdT is similar in all hosts, but VdA differs substantially among hosts, then VdA*VdT will also differ substantially and phage genomic AT% will consequently be selected to adapt to the host environment of different VdA*VdT. However, for ssDNA phages in such a scenario with the hosts differing much in VdA but little in VdT, only the genomic A%, but not the genomic T%, of the ssDNA phages will show a good correlation with the host genomic AT%. This is also consistent with the pattern in Fig. 2.
Discussion
Mutation and selection are the two sculptors of nature, but the effect of mutation on the evolution of genomes and proteins is only recently appreciated [6-9], notably after the pioneering work of Sueoka [10]. The C→T(U) mutations mediated by spontaneous deamination [1,5,11,12], in particular, have been invoked to explain the strand-asymmetry in nucleotide frequency distribution in vertebrate mitochondrial genomes [13-15], in the bacterial genomes [16-18], and in coding sequences [19-22]. In this paper, we have shown how the C→T(U) mutations can operate together with selection to shape the genomic AT% of dsDNA phage and the genomic A% and T% in ssDNA phage.
Previous studies have shown that spontaneous mutation appears to be AT-biased in different genomes and genetic backgrounds [23-26], and the evidence is convincing based on the comparison between functional genes and their pseudogene counterparts [25,26]. However, mutation alone is often insufficient to explain the observed genetic variation.
Two different kinds of AT-richness have been documented for mitochondrial genomes alone demanding two different explanations [2]. The first kind is represented by (1) the insect mitochondrial genomes where most codons end with A and T and (2) the mammalian mitochondrial D-loop which is not transcribed and very AT-rich. Both the D-loop and the third codon position of protein-coding genes evolve rapidly. In the insect mitochondrial genomes, the number of A-ending codons roughly equals the number of T-ending codons. In the D-loop, the number of A and T are distributed roughly equally in the two strands. This first kind of AT-richness was attributed to AT-biased mutation [2]. The second kind of AT-richness is represented by the coding sequences in vertebrate mitochondrial genomes, where most codons in four-fold degenerate codon families end with A but few end with T. This cannot be explained by the mutation hypothesis invoking AT-biased mutation because such mutations would lead to roughly equal number of A-ending and T-ending codons in four-fold degenerate codon families.
The large number of A-ending codons with few T-ending codons in mammalian mitochondrial genomes prompted the proposal of the transcription hypothesis of codon usage [2], based on the observation that cellular concentration of ATP is much higher than that of the other three rNTPs [Table 2.1 in [27-29], pp. 4–5]. For example, in the exponentially proliferating chick embryo fibroblasts in culture, the concentration of rATP, rCTP, rGTP and rUTP, in units of (moles × 10-12 per 106 cells), is 1890, 53, 190, and 130, respectively, in 2-hour culture, and 2390, 73, 220, and 180, respectively, in 12-hour culture. The transcription hypothesis of codon usage states that, with the high availability of rATP and relatively low availability of the other three rNTPs, the transcription efficiency can be increased by maximizing the use of A in the third codon position of protein-coding genes.
The variation of the genomic AT% in the dsDNA phage and the genomic A% and T% in the ssDNA phage in our study cannot be explained by the C→T(U) mutations alone, and we believe that the correlations shown in Figs. 1,2 are mainly the work of natural selection favouring the AT-rich phage in AT-rich hosts and AT-poor phage in AT-poor hosts. The data from ssDNA phage helped us to conclude that it is the C→T(U) mutations, instead of AT-biased mutations, are mainly responsible for the difference between the ssDNA and dsDNA phages we observe in Figs. 1,2. The results in this paper corroborate our previous finding [15] that spontaneous deamination has profound effect on the strand-biased nucleotide and codon frequency distributions and on the codon-anticodon adaptation in another kind of intracellular genomes, i.e., the vertebrate mitochondrial genomes.
Conclusion
The phage genomic AT% has evolved in response to the availability of A and T in their host cell. In particular, the difference in the relationship between the ssDNA phage and dsDNA phage, can be partially explained by the difference in (1) selection operating to maximize the rate of DNA replication and (2) the C→T(U) mutation mediated by the high rate of spontaneous mutations in the ssDNA phage.
Methods
We have downloaded complete bacteriophage genomes for 79 dsDNA phages and 27 ssDNA phages in GenBank format from NCBI [30]. Nucleotide frequencies and AT% for each viral genome was calculated from the GenBank sequence files. The host species of each viral genome is taken from the "specific_host" entry in the FEATURES table of the phage sequence file [see Additional file 1], and the genomic AT% for the host species is calculated as follows.
The genomic AT% for bacterial hosts with a genomic sequence
If the bacterial host is a particular strain of a particular species, and if the genomic sequence of that particular strain is available, then the genomic AT% was calculated from the genomic sequence. If the "specific_host" does not include a specification of the strain, and if several strains of the same bacterial species have complete genomic sequences, then the genomic AT% of the host species is the weighted average of these genomic sequences. This group consists of 82 cases out of the total of 118 phage-host pairs.
The genomic AT% for bacterial hosts without a genomic sequence
Among the total of 118 phage-host pairs, 36 cases have the bacterial host without a completely sequenced genome. The genomic AT% of these bacterial hosts is estimated from sequences retrieved from GenBank as follows. First, we perform an ENTREZ search of the host species name with the limit of sequences set to "Genomic DNA/RNA" and with the exclusion of ESTs, STSs, GSSs, and patented sequences. Second, we deleted all plasmid sequences in the retrieved files. From the remaining sequences, one might then compute AT% from the retrieved sequences as the weighted average:
where n is the number of retrieved sequences for the host species, NA + Ti is the number of A and T for the ith sequence, and Li is the length of the ith sequence.
One problem with this calculation is that some genes from the same bacterial host have been sequenced and deposited multiple times and the resulting PA + T would tend to be over-represented by those genes present in multiple copies. For example, among 292 DNA sequences deposited in GenBank for Acinetobacter sp., 152 are rRNA sequences (mostly 16S rRNA sequences), and all sequences deposited in GenBank for Roseobacter sp. are rRNA sequences. For this reason, we have first identified these genes by BLASTing [31] the sequences against each other with E-value = 0.0001 and calculated AT% by representing each set of multiply sequenced genes by the consensus sequence.
Note that this treatment may still suffer from biases. For example, DNA sequences of extreme GC% values (e.g., extremely GC-poor ones) may be more difficult to obtain than those with middle-range GC% values and are consequently underrepresented in GenBank. For this reason, we have also chosen the longest DNA sequence for each bacterial host as a genomic sample. The two sets of AT% values, with one set calculated as the weighted average and the other from the longest sequence in GenBank, are highly correlated (r = 0.975). The conclusions reached in this paper remains the same regardless of which set of the host AT% values we use. We present here only numerical results from representing each of these 36 bacterial hosts without a genomic sequence by its longest GenBank sequence. We have also performed an analysis by using only completely sequenced genomes. The sequence retrieval and analysis were carried out by using DAMBE [32,33].
In a comparative study, one should not treat each bacteriophage or bacterial host as providing an independent data point because of shared ancestry. For illustration, consider an extreme case with two clades of bacterial hosts, with each clade of species having close phylogenetic relationship and each infected by a clade of closely related phage species. We would essentially have only two data points when studying the relationship in AT% between the bacteriophage and the host, no matter how many species of bacterial hosts or bacteriophage species we have in each clade. Ideally, one should perform a phylogeny-based comparison as was done before [e.g., [34,35]]. Unfortunately, it is difficult to build a phage tree because, while a tree can be reconstructed for the bacterial species by using universally shared genes such as rRNA sequences, there is no such shared sequence among bacteriophage species. However, the phage genomic AT% appears to show little phylogenetic inertia. For example, for pairs of bacteriophage species (say A and B) in our study that share homology in protein-coding genes (indicating phylogenetic affiliation), the similarity in AT% between A and B is, in general, smaller than the similarity in AT% between each of them and their respective hosts (i.e., between A and the host of A and between B and the host of B). For this reason, we have adopted a technically undesirable, but approximately true, assumption of little phylogenetic inertia in bacteriophage AT%, with a caution for the reader that the probabilities associated with significance tests may not be exact. This assumption is somewhat justifiable based on the recent documentation of the lack of phylogenetic inertia in GC% (or AT%) of bacterial genomes [6]. If there is little phylogenetic inertia in bacterial genomes, then there should be even less phylogenetic inertia in the phage genomes because the latter evolve much faster than the former.
Authors' contributions
XX and KYY conceived the project. XX obtained funding, carried out the project and drafted the paper. KYY participated in revising four earlier versions of the manuscript.
Supplementary Material
Additional File 1
List of bacteriophage and their hosts. Also included are the genomic AT% of the host species and the phage nucleotide frequencies, in html format.
Click here for file
Acknowledgements
This study is supported by grants from NSERC (Natural Science and Engineering Research Council of Canada)'s discovery, equipment and strategic grants and from the University of Ottawa to XX. We thank Donal Hickey, Guy Drouin, Gareth Palidwor, and Jason Popescu for comments and discussion, and three anonymous reviewers for clarifying a number of ambiguities in an earlier version of the paper.
==== Refs
Frederico LA Kunkel TA Shaw BR A sensitive genetic assay for the detection of cytosine deamination: determination of rate constants and the activation energy Biochemistry 1990 29 2532 2537 2185829 10.1021/bi00462a015
Xia X Maximizing transcription efficiency causes codon usage bias. Genetics 1996 144 1309 1320 8913770
Gibbs A Primrose S A correlation between the genome compositions of bacteriophages and their hosts. Intervirology 1976 7 351 355 799626
SAS Institute Inc. SAS/STAT User's guide. Version 6, Volume1. 1989 1 4th Cary, NC , SAS Institute Inc. 943
Lindahl T Instability and decay of the primary structure of DNA. Nature 1993 362 709 715 8469282 10.1038/362709a0
Gu X Hewett-Emmett D Li WH Directional mutational pressure affects the amino acid composition and hydrophobicity of proteins in bacteria Genetica 1998 102-103 383 391 9720290 10.1023/A:1017028102013
Hickey DA Singer GA Genomic and proteomic adaptations to growth at high temperature Genome Biol 2004 5 117 15461805 10.1186/gb-2004-5-10-117
Wang HC Singer GA Hickey DA Mutational bias affects protein evolution in flowering plants Mol Biol Evol 2004 21 90 96 14595101 10.1093/molbev/msh003
Lobry JR Life history traits and genome structure: aerobiosis and G+C content in bacteria. Lecture Notes in Computer Science 2004 3039 679 686
Sueoka N Correlation bewteen base composition of deoxyribonucleic acid and amino acid composition of proteins. Proceedings of the National Academy of Sciences, USA 1961 47 1141 1149
Frederico LA Kunkel TA Shaw BR Cytosine deamination in mismatched base pairs Biochemistry 1993 32 6523 6530 8329382 10.1021/bi00077a005
Sancar A Sancar GB DNA repair enzymes Annu Rev Biochem 1988 57 29 67 3052275 10.1146/annurev.bi.57.070188.000333
Reyes A Gissi C Pesole G Saccone C Asymmetrical directional mutation pressure in the mitochondrial genome of mammals Mol Biol Evol 1998 15 957 966 9718723
Tanaka M Ozawa T Strand asymmetry in human mitochondrial DNA mutations Genomics 1994 22 327 335 7806218 10.1006/geno.1994.1391
Xia X Mutation and Selection on the Anticodon of tRNA Genes in Vertebrate Mitochondrial Genomes. Gene 2005 345 13 20 15716092 10.1016/j.gene.2004.11.019
McInerney JO Replicational and transcriptional selection on codon usage in Borrelia burgdorferi Proc Natl Acad Sci U S A 1998 95 10698 10703 9724767 10.1073/pnas.95.18.10698
Lobry JR Asymmetric substitution patterns in the two DNA strands of bacteria Mol Biol Evol 1996 13 660 665 8676740
Lobry JR Sueoka N Asymmetric directional mutation pressures in bacteria Genome Biol 2002 3 research58.1 14 10.1186/gb-2002-3-10-research0058
Beletskii A Bhagwat AS Transcription-induced mutations: increase in C to T mutations in the nontranscribed strand during transcription in Escherichia coli Proceedings of the National Academy of Sciences of the United States of America 1996 93 13919 13924 8943036 10.1073/pnas.93.24.13919
Beletskii A Bhagwat AS Correlation between transcription and C to T mutations in the non-transcribed DNA strand Biological Chemistry 1998 379 549 551 9628351
Beletskii A Grigoriev A Joyce S Bhagwat AS Mutations induced by bacteriophage T7 RNA polymerase and their effects on the composition of the T7 genome Journal of Molecular Biology 2000 300 1057 1065 10903854 10.1006/jmbi.2000.3944
Beletskii A Bhagwat AS Transcription-induced cytosine-to-thymine mutations are not dependent on sequence context of the target cytosine Journal of Bacteriology 2001 183 6491 6493 11591695 10.1128/JB.183.21.6491-6493.2001
Marcelino LA Andre PC Khrapko K Coller HA Griffith J Thilly WG Chemically induced mutations in mitochondrial DNA of human cells: mutational spectrum of N-methyl-N'-nitro-N-nitrosoguanidine Cancer Res 1998 58 2857 2862 9661902
Wang RF Campbell W Cao WW Summage C Steele RS Cerniglia CE Detection of Pasteurella pneumotropica in laboratory mice and rats by polymerase chain reaction Lab Anim Sci 1996 46 81 85 8699827
Gojobori T Li WH Graur D Patterns of nucleotide substitution in pseudogenes and functional genes J Mol Evol 1982 18 360 369 7120431
Li WH Wu CI Luo CC Nonrandomness of point mutation as reflected in nucleotide substitutions in pseudogenes and its evolutionary implications Journal of Molecular Evolution 1984 21 58 71 6442359
Kornberg A Baker TA DNA replication. 1992 New York , Freeman 931
Colby C Edlin G Nucleotide pool levels in growing, inhibited, and transformed chick fibroblast cells. Biochemistry 1970 9 917 4313738 10.1021/bi00806a029
Bridger WA Henderson JF Cell ATP 1983 New York. , Wiley
NCBI NCBI Viral genomes
Altschul SF Gish W Miller W Myers EW Lipman DJ Basic local alignment search tool. Journal of Molecular Biology 1990 215 403 410 2231712 10.1006/jmbi.1990.9999
Xia X Data analysis in molecular biology and evolution. 2001 Boston , Kluwer Academic Publishers 277
Xia X Xie Z DAMBE: Software package for data analysis in molecular biology and evolution J Hered 2001 92 371 373 11535656 10.1093/jhered/92.4.371
Xia X Hafner MS Sudman PD On transition bias in mitochondrial genes of pocket gophers. Journal of Molecular Evolution 1996 43 32 40 8660427
Xia X The rate heterogeneity of nonsynonymous substitutions in mammalian mitochondrial genes. Molecular Biology and Evolution 1998 15 336 344 9501500
| 15823202 | PMC1087484 | CC BY | 2021-01-04 16:38:18 | no | BMC Genet. 2005 Apr 11; 6:20 | utf-8 | BMC Genet | 2,005 | 10.1186/1471-2156-6-20 | oa_comm |
==== Front
BMC Mol BiolBMC Molecular Biology1471-2199BioMed Central London 1471-2199-6-81583110110.1186/1471-2199-6-8Research ArticleUse of adenoviral E1A protein to analyze K18 promoter deregulation in colon carcinoma cells discloses a role for CtBP1 and BRCA1 Delouis Cécile [email protected] Philippe [email protected] Madeleine [email protected] Olivier [email protected] Laboratoire de Génétique Oncologique, UMR 8125 CNRS, Institut Gustave Roussy, 39 rue Camille Desmoulins, 94805 Villejuif, France2 PP: Stowers Institute, 1000 E 50th street, Kansas City, MO 64110, USA; OB: UMR 7147, Institut Curie, 26 rue d'Ulm,75248 Paris cedex 05, France2005 14 4 2005 6 8 8 16 11 2004 14 4 2005 Copyright © 2005 Delouis et al; licensee BioMed Central Ltd.2005Delouis 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 promoter of the keratin 18 (K18) gene is 5- to 10-fold more active in tumorigenic (T-type) cell clones derived from the SW613-S human colon carcinoma cell line than in non-tumorigenic (NT-type) clones. We have reported previously that the mechanism responsible for this differential activity is acting on the minimal K18 promoter (TATA box and initiation site). This mechanism does not require the binding of a factor to a specific site on the DNA but involves the acetylation of a non-histone substrate. To get further insight into this mechanism, we investigated the effect of the adenovirus E1A protein on the activity of the K18 promoter, both in T and NT cells.
Results
Wild type adenovirus E1A protein and C-terminal deletion mutants inhibit the K18 promoter, specifically in T-type cells. The domain responsible for this inhibitory effect is located in the 12–25 region of the viral protein. E1A mutants that have lost this region but retain the PLDLS motif (the C-terminal binding site for CtBP1) stimulate the K18 promoter, specifically in NT cells. The inhibitory or stimulatory effects of the different E1A mutants are not dependent on a particular sequence of the promoter. An E1A N-terminal deletion mutant carrying point mutations in the PLDLS motif cannot stimulate the K18 promoter. CtBP1 interacts with CtIP, which is a known partner of BRCA1, itself a component of the RNA polymerase II holoenzyme. The stimulatory effect of two BRCA1 mutants, specifically in NT cells, implicates a tripartite BRCA1-CtIP-CtBP1 complex in the regulation of the K18 promoter.
Conclusion
Since we have shown previously that the K18 promoter is stimulated by deacetylase inhibitors, specifically in NT cells, we conclude that the activity of the promoter is repressed in NT cells by a mechanism involving the recruitment, by a BRCA1/CtIP complex, of CtBP1 and associated deacetylases to the preinitiation complex. We propose a model depicting the mechanism responsible for the differential activity of the K18 promoter between T and NT cells of the SW613-S cell line.
==== Body
Background
The early region 1A (E1A) of adenoviruses encodes two main proteins (243 and 289 aa-long in human adenovirus 2) which are translated from alternatively spliced mRNAs (12S and 13S, respectively). The two proteins have identical N- and C-terminal regions but the larger one (E1A-13S) has an additional domain (46 aa-long in adenovirus 2) located in the central part of the protein. Four regions that are conserved between several human adenoviruses were named CR1, CR2, CR3 and CR4. In adenovirus 2, CR3 almost coincides with the 46 aa-long additional domain present in the E1A-13S isoform. The E1A-12S and E1A-13S proteins are required to activate the transcription of other viral genes. In addition, these proteins interact with multiple cellular proteins to reprogram the expression of many cellular genes in infected cells (reviewed in [1-3]). This is necessary for the virus to replicate its DNA and complete a productive cycle. The E1A proteins were also found to be oncogenic in primary rodent cells. This is a consequence of their ability to interact with key cellular factors, including regulators of the cell cycle such as the pocket proteins (Rb, p107, p130), the related p300 and CBP coactivators of transcription or the cyclin-dependent kinase inhibitors p21Cip1/Waf1 and p27Kip1. Among the other cellular factors known to interact with E1A are the TRRAP, p400 and CtBP1 proteins. CtBP1 is a transcriptional corepressor. Several transcriptional repressors recruit CtBP1 to promoters through a conserved PXDLS motif. CtBP1 represses transcription either directly or by subsequent recruitment of histone deacetylases (HDACs) [4]. E1A proteins interact with CtBP1 by a PLDLS motif located at the C-terminus of the viral proteins [5]. A PLDLS motif is also present in the binding site of the cellular CtIP protein, another interacting partner of CtBP1 [4,6]. The function of CtIP in the cell is unknown but it was found to interact with the BRCT repeats of the BRCA1 tumor suppressor protein [7,8]. CtIP binds to BRCA1 through a site distinct from the PLDLS motif [9]. Thus, CtIP can act as an adapter protein between BRCA1 and CtBP1.
We are interested in the mechanisms involved in transcriptional deregulation of gene expression in the cells of the SW613-S cell line derived from a human colon carcinoma [10]. Analysis of cellular clones isolated from this cell line indicated that it is heterogeneous and composed of a mixture of two main cell types (named here T and NT). T-type cells have a high level of amplification and expression of the c-MYC gene, whose additional copies are located on extrachromosomal elements (double minute chromosomes). NT-type cells present a low level of amplification of the oncogene and the supernumerary copies are integrated into chromosomal DNA [11-14]. T and NT cells markedly differ by several phenotypic traits such as their tumorigenic potential in nude mice, capability to grow in serum-free medium, sensitivity to the induction of apoptosis and cellular morphology [12,15,16]. Genes overexpressed in T cells, as compared to NT cells, have been identified [12,17-20]. Among them, we chose the keratin 18 (K18) gene to investigate the mechanism responsible for its overexpression in T-type cells. We previously reported that this high level of expression is mainly due to an increase in transcriptional rate [17] and that, in transient expression assays, the K18 promoter is much more active in T cells than in NT cells [21]. The mechanism responsible for this higher activity is acting on the minimal K18 promoter (TATA box and initiation site) and does not involve the binding of a factor to a specific site on the DNA [22]. We also found that an acetylation mechanism acting on a non-histone substrate is driving the high activity of the promoter in T cells [23]. In order to get further insight into this mechanism, we now investigated the effect of the adenovirus E1A protein on the activity of the K18 promoter, both in T and NT cells. This protein has been widely used as a powerful tool to identify and/or study important cellular regulatory proteins, in particular factors involved in acetylation mechanisms (p300/CBP, PCAF, TRRAP) [24]. Using a series of E1A mutant proteins, we uncovered a role for the CtBP1 and BRCA1 proteins in the functioning of the K18 promoter in SW613-S cells.
Results
Wild type E1A and E1A mutants differentially inhibit or stimulate the minimal K18 promoter
We previously found that the adenoviral E1A-12S protein has an inhibitory effect on the K18 promoter, specifically in T-type cells of the SW613-S cell line. In contrast, a mutant form of the protein, lacking the CR1 domain (E1A-12S-ΔCR1), has lost this inhibitory effect but gained a stimulatory activity on the K18 promoter, specifically in NT cells [23]. All the E1A mutants that we planned to use in this work were derived from the E1A-13S isoform. Therefore, the effect of the E1A-13S protein and of the E1A-d30-76 mutant protein (equivalent to E1A-12S-ΔCR1) on the activity of the K18(80) promoter (-80 to +21) was assayed. As shown in figure 1A, the E1A-13S protein has also a strong inhibitory effect on the promoter activity in T cells whereas it has no significant effect in NT cells. In contrast, the E1A-d30-76 mutant behaves like the E1A-12S-ΔCR1 protein and stimulates the activity of the promoter, specifically in NT cells.
The promoter of the K18 gene comprises a TATA box and an initiation site (minimal promoter) flanked by an Sp1 binding site (Fig. 1C). The minimal K18(41) promoter (-41 to +21) has a low but differential activity between T and NT cells. The flanking Sp1 binding site is essential for a high activity of the promoter but not for its differential behavior [21]. The effect of the E1A-13S protein and of the E1A-d30-76 mutant was also assayed on the minimal K18(41) promoter (Fig. 1B). The former inhibits the activity of the minimal promoter, specifically in T cells, and the latter stimulates it only in NT cells, indicating that the Sp1 binding site is not essential for these effects. By mutational analysis of the K18 minimal promoter, we have previously shown that the factor(s) responsible for its differential activity between T and NT cells does not bind to a specific DNA sequence [22]. Using the same mutant constructs of the K18 promoter (Fig. 2E), we investigated whether the inhibitory or stimulatory effects of the E1A-13S and E1A-d30-76 proteins were also sequence-independent. As shown in figure 2A–D, all the promoter constructs are inhibited by E1A-13S, specifically in T cells, and are stimulated by the E1A-d30-76 mutant, only in NT cells. These results suggest that, as for the differential activity of the promoter, the effect of these viral proteins is not mediated by a factor which binds to a specific site on the promoter but rather through alterations of protein-protein interactions within the preinitiation complex.
An N-terminal domain of E1A is involved in the inhibitory effect on the K18 promoter
In order to identify the region(s) of the E1A protein involved in the inhibition of the K18 promoter in T-type cells, we first used a series of N- or C-terminal deletion mutants (Fig. 3A). All the C-terminal deletion mutants have a comparable inhibitory effect on the activity of the K18 promoter in T-type cells (Fig. 3B). Mutant E1A-C29, which comprises only the N-terminal 29 aa, seems less efficient at inhibiting the promoter. Further careful comparative analysis indicated that the inhibitory potency of this mutant is about 60% of that of the E1A-13S protein (data not shown). Some of the mutants also acquired an inhibitory potential in NT cells but this point was not explored further. None of the C-terminal deletion mutants display a stimulatory effect on the promoter in NT cells. The localization of the inhibitory domain in the first 29 aa of E1A was confirmed by the analysis of N-terminal deletion mutants (Fig. 3C). Mutant E1A-N25, deleted of the first 25 amino acids, has lost the inhibitory potential in T cells but has acquired the ability to stimulate the promoter in NT cells. This was also the case for all the other N-terminal deletion mutants. The expression level of the constructs coding for the E1A deletion mutants was checked by Western blotting (Fig. 4A).
The results obtained with the N- and C-terminal deletion mutants, assigning a role to the first 25 aa in the inhibitory effect of E1A on the K18 promoter, were apparently contradictory to those obtained with deletion mutants E1A-12S-ΔCR1 and E1A-d30-76 (see above). To solve this issue, a series of small deletion mutants was constructed, spanning the region of aa 2 to 85 (Fig. 5A). With the exception of mutant E1A-d12-25, which has lost the inhibitory potential in T cells (and acquired a stimulatory effect in NT cells), all the other deletions in this region of the protein have no influence on the inhibitory effect of E1A and are unable to confer a stimulatory potential (Fig. 5B). In particular, mutants E1A-d26-35, -d30-45, -d40-60 and -d60-85, whose deletions span the region deleted in mutant E1A-12S-ΔCR1 and E1A-d30-76, have the same properties as wild type E1A with respect to the K18 promoter. The expression level of the E1A constructs used in the above-described experiments was checked by Western blotting (Fig. 4B). From all these results, we concluded that region 12–25 of E1A is involved in its inhibitory effect on the K18 promoter in T-type cells. The large deletion carried by mutants E1A-d30-76 and E1A-12S-ΔCR1 is adjacent to the 12–25 domain. A possible explanation for the results obtained with these mutants is that such large deletions may perturb the conformational structure of the molecule and functionally inactivates the 12–25 domain. It has been reported [25] that E1A mutants with deletions between residues 30–85 may not enter the nucleus, despite the presence of an unaltered nuclear localization signal at their C-terminus.
A C-terminal domain of E1A is involved in the stimulatory effect on the K18 promoter
As mentioned above, all the N-terminal deletion mutants tested possess an activator effect on the K18 promoter, specifically in NT cells (Fig. 3C). Mutant E1A-N190 comprises only the last 99 aa of E1A-13S. On the other hand, none of the C-terminal mutants tested has acquired a stimulatory activity on the promoter in NT cells (Fig. 3B). From these results, we concluded that the region of E1A responsible for the stimulatory effect is located in the C-terminal part of the protein. The CtBP1 protein is known to interact with the PLDLS motif present in this C-terminal region of E1A [5]. To investigate a role for CtBP1 in the stimulatory effect of this region on the K18 promoter in NT cells, we used a previously characterized mutant of the PLDLS domain [6]. A two aa change (PLDLS→PLASS) destroys the capacity of binding to CtBP1 but does not affect the nuclear localization signal which is located nearby. We constructed the E1A-N190-mut281-282 mutant which is mutated on aa 281 (D→sA) and 282 (L→S). The binding of CtBP1 to E1A-N190 and the loss of interaction with E1A-N190-mut281-282 were confirmed by GST pull down assays and Western blotting (Fig. 6A). In transient expression assays (Fig. 6B), the E1A-N190-mut281-282 mutant cannot stimulate the K18 promoter in the NT cells. This result indicates that the binding to CtBP1 plays a role in the stimulatory effect of the C-terminal region of E1A on the activity of the K18 promoter in NT cells. This was confirmed by the observation that an excess of CtBP1 protein can abolish the stimulatory effect of E1A-N190 on the K18 promoter in NT cells (Fig. 6C), although CtBP1 by itself has no effect on the activity of the promoter (Fig. 6D). The expression level of the E1A-N190 and CtBP1 vectors was controlled by Western blot analysis (Fig. 7). A likely explanation for these results is that, in NT cells, the K18 promoter is repressed by a mechanism involving the binding of CtBP1 to the PLDLS motif of a partner protein. Stimulatory mutants such as E1A-N190 disrupt this interaction and relieve the repressed state of the promoter.
The differential activity of the K18 promoter and the inhibitory or stimulatory effect of various E1A mutant proteins are mediated by the minimal promoter ([22]; this work). No specific sequence of the promoter seems necessary for these effects (see above) which likely result from alterations in protein interactions in the preinitiation complex. It was reported that the CtBP1 protein is a corepressor which is recruited to promoters by DNA-bound transcription factors [4]. Therefore, our observations raise the question of how this protein is recruited to the preinitiation complex on the minimal K18 promoter. CtBP1 interacts with the carboxy-terminal interacting protein (CtIP) [6], which is a known partner of the Rb, LMO4 and BRCA1 proteins. A complex comprising LMO4, BRCA1 and CtIP was demonstrated in vivo [26]. The direct recruitment of the Rb protein to the transcription machinery has not been described to date but the presence of the BRCA1 protein in the RNA polymerase II holoenzyme has been reported [27,28]. This observation raises the possibility of a direct association of the BRCA1 protein with the preinitiation complex. The C-terminal BRCT repeats of BRCA1 are responsible for the interaction of this protein with CtIP [7,8]. The nonsense Y1853→STOP mutation identified in some familial breast cancers results in a 10-aa truncation of the second BRCT repeat and prevents the interaction of the mutant protein with CtIP. The possible effect of the wild type BRCA1 protein, of the BRCA1(Y1853→STOP) mutant and of a C-terminal fragment of the BRCA1 protein comprising only the BRCT repeats (pcDNA3-BRCT construct) on the activity of the K18 promoter was tested (Fig. 8A). If the BRCA1(Y1853→STOP) mutant has retained the capacity to be incorporated into the preinitiation complex, it should behave as a dominant negative mutant and stimulate the activity of the promoter, specifically in NT cells, since it is unable to recruit the CtIP-CtBP1 complex. If the BRCT fragment has lost the capacity to be incorporated into the preinitiation complex, it should also behave as a dominant negative mutant and stimulate the activity of the promoter, specifically in NT cells, since it should sequester the CtIP-CtBP1 complex away from the preinitiation complex. Wild type BRCA1 has no effect on the activity of the promoter in transient expression assays. The BRCA1(Y1853→STOP) mutant and the BRCT fragment do have a stimulatory effect on the activity of the K18 promoter, specifically in NT cells. The expression level of the BRCA1 constructs used in these experiments was assessed by Western blotting (Fig. 8B and 8C). The apparent dominant negative effect of the two BRCA1 mutants indicates that the interaction of BRCA1 with CtIP plays an important role in maintaining the repressed state of the K18 promoter in NT cells. Altogether, our results strongly suggest that the CtIP protein acts as an adapter molecule between BRCA1 and CtBP1, allowing the recruitment of CtBP1 to the preinitiation complex and the repression of the K18 promoter in NT cells.
Discussion
Differential activity of the K18 promoter and opposite effects of E1A mutants
We report here that the E1A protein isoforms (12S and 13S) and derivatives of E1A-13S that retain the N-terminal first 29 aa inhibit the activity of the K18 promoter, specifically in T-type cells of the SW613-S colon carcinoma cell line. In contrast, derivatives of the viral protein which have lost the inhibitory region but retain the C-terminal region of E1A stimulate the activity of the K18 promoter, specifically in NT cells. The two phenomena appear to be exclusive: among more than 30 E1A mutants studied (this work and unpublished results), none was found that could both repress the activity of the promoter in T cells and stimulate it in NT cells. This observation suggests that the two phenomena are related and could result from interference with the same and single mechanism. This is further supported by the observation that the stimulatory effect of mutant E1A-N25 is abolished in the presence of the wild-type E1A protein (data not shown). We have previously found that the differential activity of the K18 promoter between T and NT cells is a property of the minimal promoter (K18(41)) [21]. The mechanism responsible for this difference in activity does not involve the binding of a factor to a specific sequence of the promoter [22] but probably results from alterations of protein-protein interactions within the preinitiation complex. The same conclusion was reached here for the stimulatory or inhibitory effects of the various E1A derivatives on the activity of the K18 promoter. These effects were observed with the minimal promoter and with all the mutated versions of the K18 promoter tested. Altogether, these results strongly suggest that both inhibitory and stimulatory mutants of E1A exert their effect by interfering with the very mechanism responsible for the differential activity of the K18 promoter.
A functional domain of the E1A protein located in the 12–25 region
We have found that a unique region of E1A, located between aa 12 and 25, is involved in its inhibitory effect on the activity of the K18 promoter. We previously reported results which suggested that E1A is acting on the p300 or CBP protein to inhibit the activity of the K18 promoter in T-type cells [23]. Two regions of the E1A protein, the N-terminal 1–28 region and the CR1 region, are known to be involved in the interaction of the viral protein with p300/CBP [29]. Within the CR1 region, the p300/CBP interacting domain is most probably located between aa 66 and 72. In the N-terminal region, two p300/CBP interacting domains have been identified. The first one spans residues 2–10 of E1A and the second domain is located between aa 19–28. It was shown [30,31] that aa 11–22 are not essential for the binding of E1A to p300/CBP. Thus, the 12–25 region involved in the inhibition of the K18 promoter appears to be different from the regions of interaction with the p300/CBP proteins. Furthermore, we have found that mutants E1A-ARG2 (R2G – data not shown), E1A-d2-11 and E1A-d61-85 which can no longer bind to p300/CBP, still efficiently inhibit the activity of the K18 promoter in T-type cells. Altogether, our results lead us to conclude that, contrary to what we suggested previously, inactivation of the p300/CBP proteins is most probably not involved in this inhibitory effect. It was reported that aa 1–21 and 55–60 of E1A are involved in its interaction with the PCAF factor [32,33]. The 1–21 region has not been further investigated so far to determine precisely the aa residues responsible for binding to PCAF. However, since the E1A-d40-60 mutant retains an inhibitory potential on the activity of the K18 promoter, this inhibition probably does not involve the PCAF protein. Within the first N-terminal 80 aa of E1A, the 12–26 region has the highest score for the probability of forming an α-helix and this property is conserved through five adenovirus serotypes [34]. The RAP30 subunit of TFIIF and the TATA-box binding protein (TBP) were shown to interact in vitro with aa 1–29 of E1A [35]. Severino et al [36] reported that aa 1–36 of E1A are responsible for the interaction with the RACK1 protein. We have screened a cDNA library by the yeast two-hybrid system using aa 12–29 of E1A as a bait. Clones corresponding to the glyceraldehyde-3-phosphate dehydrogenase (GAPDH), the UREB1 and the MYCBP proteins have been isolated (OB, ML, unpublished results). Experiments are in progress to determine if one of these factors is involved in the regulation of the K18 promoter in SW613-S cells.
A role for CtBP1 and BRCA1 in the regulation of the K18 promoter
E1A mutants that have lost the 12–25 inhibitory region, acquired a stimulatory potential on the activity of the K18 promoter, specifically in NT cells, provided that they retain the C-terminal region of the viral protein. Within this region, aa 281–282, which are part of the CtBP1 binding site, are essential to the stimulatory effect and an excess of CtBP1 protein abolishes this effect. The CtBP1 protein is a corepressor, which represses promoters by HDAC-dependent or HDAC-independent mechanisms. We have previously shown [23] that the activity of the K18 promoter is stimulated by HDAC inhibitors (sodium butyrate or trichostatin), specifically in NT cells. We also made the observation that the K18 promoter could be stimulated in NT cells by forced recruitment to the promoter of a GAL4 fusion protein containing the histone acetyl-transferase (HAT) domain of CBP or of an acidic activator (GAL4-VP16) known to interact with HAT complexes. Mutations of aa residues involved in the HAT activity of CBP or in the interaction of VP16 with HAT complexes strongly reduced the stimulation. We conclude that the activity of the K18 promoter is repressed in NT cells by a mechanism involving the CtBP1 and HDAC proteins. CtBP1 is most probably acting through an HDAC-dependent mechanism because (i) we have reported [23] that the stimulatory effects by sodium butyrate and by the E1A-ΔCR1 mutant are not additive; (ii) forced recruitment of a GAL4-CtBP1 fusion protein on a K18 promoter engineered with GAL4 binding sites strongly inhibits the activity in both T and NT cells and this inhibition cannot be reversed by sodium butyrate (CD, unpublished results). These last results indicate that CtBP1 can inhibit the K18 promoter by an HDAC-independent mechanism but in this case, inhibition is not cell type specific. Stimulation of the activity of the K18 promoter by some E1A mutants in NT cells is most likely due to a relief of the repressive mechanism operating in these cells. The fact that these mutants have no stimulatory effect in T-type cells suggests that the high activity of the promoter reflects a reversal of the same repression mechanism, possibly through the expression of some constitutively high histone/factor acetyl-transferase (HAT/FAT) activity in these cells (see below). Our results are in agreement with those obtained by others. Grooteclaes et al. [37] reported that the K18 gene is derepressed in embryo fibroblasts from mice with homozygous compound knockout of the CtBP1 and CtBP2 genes. In addition, the K18 gene is repressed in knockout cells infected with a CtBP1 retroviral vector. Schuierer et al. [38] found that the E1A protein activates the expression of the AP2α gene by interfering with CtBP1. Sundqvist et al. [39] have reported that the CtBP1 interacting region of E1A relieves the HDAC-dependent repression of transcription by CtBP1.
We propose that CtBP1 and associated HDAC could be recruited to the preinitiation complex by a BRCA1/CtIP complex. Such a role for BRCA1 was not reported previously although it is known that BRCA1 is associated with the RNA polymerase II holoenzyme. Some examples of transcriptional repression by BRCA1 were reported. It represses c-MYC-mediated stimulation of the human telomerase reverse transcriptase promoter [40]. BRCA1 functions as a corepressor for the transcriptional repressor ZBRK1 [41]. Experiments using DNA microarrays have indicated that the BRCA1, CtBP1, CtBP2 and CtIP genes are expressed at comparable levels in T and NT cells of the SW613-S cell line (C. Lavialle, personal communication). This was also the case for genes coding for HDAC-1 to -4 and HDAC-6 to -10. The HDAC-5 gene was found to be overexpressed 3.4 fold in NT cells, as compared to T cells, and this was confirmed by Western blot analysis. However, in transient expression assays, the activity of the K18 promoter was unaffected by overproducing the HDAC-5 protein, both in T and NT cells (CD, unpublished results), indicating that the low level of accumulation of HDAC-5 in T-type cells is not responsible for the high activity of the promoter in these cells. It could be argue that a gene coding for one of these factors is specifically mutated in T cells. This seems highly unlikely since the K18 promoter is also deregulated in stable transfectants obtained from NT cells using a c-MYC expression vector and which acquired all the phenotypic properties of T-type cells (PP, unpublished results).
A model for the differential activity of the K18 promoter
To summarize the observations we made on the differential regulation of the K18 promoter between T and NT cells of the SW613-S cell line, we propose the model presented in figure 9. Several aspects are still speculative but it is compatible with all the results we have obtained so far. The expression level of the K18 gene in NT cells is comparable to that observed in normal epithelial cells of human colon [42]. Our results indicate that, in NT cells, the K18 promoter is in fact subjected to a repression mechanism involving the CtBP1 protein and HDACs, probably in order to keep in check the expression level of the K18 gene. The data presented suggest that CtBP1 and associated HDACs could be recruited to the preinitiation complex by a BRCA1/CtIP complex (Fig. 9A). In our model, the target of HDACs is a factor (S), whose acetylation level is controlled by a balance between HDAC and HAT/FAT activities. The activity of the K18 promoter would be dependent on the acetylation level of S. We have reported previously that the acetylation state of histones H3 and H4, as well as the structure of the chromatin, are the same in T and NT cells in the region of the K18 promoter [23]. In addition, the mechanism responsible for the differential activity is acting on the minimal promoter and no specific sequence is required. Therefore, we propose that S is a non-histone protein which is a component of the preinitiation complex and is acetylated by a FAT activity.
In T-type cells, the FAT activity recruited to the preinitiation complex would be higher, because the enzyme is overproduced or more active in these cells. The high FAT activity would shift the balance toward an hyperacetylated state of S. We propose that this protein with FAT activity (F) is the target of the 12–25 domain of E1A (Fig. 9B). The viral protein, or its mutant forms retaining a functional 12–25 domain, would inhibit the FAT activity. This would result in the inhibition of the promoter. It is very likely that the factor with FAT activity is not the p300/CBP or PCAF proteins because we identified E1A mutants disabled in their capacity to bind to these factors but that are still able to efficiently inhibit the activity of the promoter in T-type cells. In the context of our model and as a novel candidate FAT activities, it is interesting to note recent reports describing the autoacetylation of the general transcription factor TFIIB [43] and of the RAP30 subunit of TFIIF [44].
E1A mutants with no functional 12–25 domain but retaining the C-terminal PLDLS motif would prevent by competition the recruitment of CtBP1 and HDACs to the preinitiation complex (Fig. 9C). In NT cells, this would shift the balance towards hyperacetylation of S despite the low FAT activity of the F protein, resulting in a stimulation of the promoter. Such mutants are not expected to have an effect in T-type cells since the high FAT activity in these cells already supersede that of HDACs. Finally, the proposed model offers an explanation to the observation that inhibition of the promoter in T cells and stimulation in NT cells are apparently exclusive phenomena. Indeed, according to the model, no E1A mutant is expected to have both capabilities.
Conclusion
The promoter of the keratin 18 gene is deregulated in cells of the human colon carcinoma cell line SW613-S by an unusual mechanism that is acting on the minimal promoter and involves alteration of an acetylation mechanism acting on a non-histone substrate. We report here that the adenoviral E1A protein and some of its mutants specifically interfere with this mechanism through two regions of the protein: the C-terminal CtBP1 binding domain and a domain spanning aa 12–25. Our results lead us to conclude that, in colon epithelial cells, the expression level of the K18 gene is kept in check by a repression mechanism involving the CtBP1, HDAC and BRCA1 proteins. This mechanism is altered in SW613-S colon carcinoma cells that overexpress the K18 gene. Since it is acting at the level of the preinitiation complex, its alteration most probably participates in the deregulated expression of many genes in these tumor cells.
Methods
Cell lines, transfection and luciferase assays
The origin of the SW613-S cell line and cell culture conditions have been described previously [10,16]. Clones SW613-3 (T-type cells) and SW613-B3 (NT-type cells) were used in this study. Transfection and luciferase assays were performed in triplicate for each construct in each experiment, as described previously [21,22] and six micrograms of each plasmid were used, unless otherwise stated. In cotransfection experiments where the reporter plasmid was co-introduced with expression vectors coding for various polypeptides, the reference assay comprised cells transfected with the reporter plasmid and with an amount of empty vector corresponding to the molar equivalent of the expression vector used. In each experiment, the two cell types were also transfected in parallel with the pSVluc construct and the activity of every promoter construct was expressed relative to that of the SV40 early promoter. We have shown previously that this viral promoter is equally active in both cell types [21]. This is also the case for the human cytomegalovirus IE1 gene promoter which drives the expression of the various polypeptides encoded by the vectors used in co-transfection experiments (CD, unpublished observations).
Plasmids constructions
Construction of plasmids pK18(41)luc, pK18(80)luc and pSVluc has already been reported [22]. Plasmids pE1A-13S, pE1A-C249, -C192, -C139, -C109, -C79, -N25, -N76, -N120, -N132 [33] were given to us by Dr R.M. Evans (Salk Institute for Biological sciences, La Jolla, USA). Plasmids pE1A-d2-11 and pE1A-d26-35 [45] were gifts of Dr. M. Fuchs (Dana-Farber Cancer Institute, Boston, USA). Plasmids pcDNA3-CtBP and pGEX-CtBP [6,39] coding for the human CtBP1 protein were given to us by Dr. C. Svensson (Uppsala University, Sweden). Plasmids pCBRCA1C (coding for wild type BRCA1) and pcDNA3-BRCA1(Y1853→STOP) were provided by Dr J. Feunteun (Institute Gustave Roussy, Villejuif, France). A fragment coding aa 1640 to 1864 of BRCA1 was amplified by the polymerase chain reaction (PCR) from plasmid pGEX4T1-BRCT [46], a gift from Dr N. Dalla Venezia (Université Médicale Rockefeller, Lyon, France), and inserted into the vector pcDNA3. A DNA fragment coding for 3 tag HA motifs was then inserted upstream of the BRCT coding region to yield plasmid pcDNA3-BRCT. Plasmid pE1A-d30-76 was obtained by inserting into plasmid pE1A-N76 a fragment amplified by PCR from plasmid pE1A-13S and coding for aa 1 to 29 of E1A. For all E1A constructs, aa numbering refers to the E1A-13S isoform. Plasmids pE1Ad30-45, pE1A-d40-60, pE1A-d61-85 were constructed by inserting into the pCMX-PL1 vector two fragments coding respectively for aa 1–29 and 46–289, 1–39 and 61–289 or 1–60 and 86–289 of E1A-13S. Plasmid pE1A-d12-25 was obtained by inserting into plasmid pE1A-N25 a double-stranded synthetic oligonucleotide coding for aa 1 to 11. Plasmid pE1A-C29 was derived from plasmid pE1A-d30-76 by deleting the sequence corresponding to aa 77 to 289. Plasmids pE1A-N190 and pE1A-N190-mut281-282 were constructed by inserting into the pCMX-PL1 vector a fragment amplified by PCR coding respectively for the wild type sequence of aa 190 to 289 or a sequence with two mis-sense mutations on aa 281 (D→A) and 282 (L→S). The same PCR fragments were also inserted into plasmid pGEX-1LambdaT coding the glutathion-S-transferase (GST) protein to obtain plasmids pGEX-E1A-N190 and pGEX-E1A-N190-mut281-282. All constructs coding for E1A proteins were checked by sequencing.
GST-pull down assays
A 40 ml overnight culture of bacteria transformed with the appropriate plasmid was diluted 10-fold with culture medium and incubated for one hour at 37°C with shaking. Induction was for 3 hours in the presence of isopropylthiogalactoside (1 mM). The bacteria were pelleted by centrifugation (10 min, 4000 rpm), resuspended in 20 ml of phosphate-buffered saline (PBS) containing Complete Inhibitor EDTA free (Roche) and 1 mM phenyl-methylsulfonyl fluoride (PMSF) (Sigma) and disrupted by sonication (twice 15 s, 50% of full power). The suspension was adjusted to 1% of Triton X-100 and incubated for 30 min at 4°C. After centrifugation (30 min, 13 000 rpm, 4°C), the supernatant was collected and stored at -80°C as one ml aliquots. The concentration of the GST-fusion protein was determined by purifying it from a one ml aliquot using glutathione beads. Quantification was carried out by migration on a polyacrylamide gel, using a range of known quantities of bovine serum albumin as a standard. For the pull-down assays, 10 μg of GST fusion proteins and 50 μl of a 50% slurry of glutathione Sepharose 4B beads (Pharmacia Biotech) were mixed and diluted to 1 ml with PBS. Incubation was for 2 hours at 4°C with gentle shaking. Beads were washed twice with PBS and once with the protein extraction buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% NP40 and 1 mM EDTA). The beads were incubated with cellular protein extracts for 2 hours at 4°C, washed with protein extraction buffer and boiled in loading buffer.
Western blotting
Samples of cellular extracts (300 μg of proteins) or from the pull-down assays were analyzed by migration on 15% SDS-polyacrylamide gels (6% for the BRCA1 and BRCA1(Y1853→STOP) proteins). Western blotting onto nitrocellulose membranes was performed using a semi-dry transfer procedure [47] or a liquid transfer method (400 mA overnight in a Tris-base 25 mM, glycin 200 mM solution) in the case of BRCA1. Blocking of the membrane and incubation with the antibody were performed in a solution containing 10 mM Tris-HCl pH 8.0, 150 mM NaCl, 0.05% Tween 20 and 5% dry low-fat milk. The mouse monoclonal anti-HA antibody was a gift from Dr. S. Leibovitch (Institute Gustave Roussy, Villejuif, France). Anti-E1A (SC430) and anti-CtBP1 (SC11390) rabbit polyclonal antibodies were obtained from Santa-Cruz Biotechnology, Inc.
Authors' contributions
CD carried out the protein biochemistry studies, sequence analysis and part of the molecular genetic studies and drafted the manuscript. PP carried out part of the molecular genetic studies and helped to draft the manuscript. ML carried out part of the molecular genetic studies. OB conceived the study, participated in its design, coordinated it and helped to draft the manuscript.
Acknowledgements
We thank Dr R.M. Evans and Dr M. Fuchs for the kind gift of many constructs coding for E1A mutants. We are grateful to Dr C. Svensson for providing us with CtBP1 plasmids, to Dr J. Feunteun and Dr N. Dalla Venezia for the BRCA1 constructs, to Dr S. Leibovitch for the anti-HA antibody and to Dr S. Khochbin for HDAC constructs and antibodies. We are greatly indebted to V. Pignot for her efficient help with western blotting. We also thank A. Trousson and J.-F. Buquet for help with some experiments, Dr N. Modjtahedi and B. Dubourg for helpful discussions and Dr C. Lavialle for critical reading of the manuscript. C.D. was supported by fellowships from the Ligue Nationale Française contre le Cancer and from the Fondation pour la Recherche Médicale. P.P. was supported by a fellowship from the Ligue Nationale Française contre le Cancer.
Figures and Tables
Figure 1 Effect of E1A-13S or E1A-d30-76 proteins on the activity of the K18 promoter. SW613-3 (T) or SW613-B3 (NT) cells were transfected with plasmid pK18(80)luc (A) or plasmid pK18(41)luc (B) alone (none) or in the presence of plasmid pE1A-13S or pE1A-d30-76, as indicated. Luciferase activities are expressed relative to that obtained with the reporter plasmid alone for each cell clone. Error bars represent the standard deviation of the mean of two experiments. (C) Structure of the promoter of the human K18 gene. The binding site of the Sp1 transcription factor, the initiation site (arrow at position -11; we have shown that the initiation is located 11 bp upstream of previously published +1 position [22]) and the TATA box are indicated. Numbers indicate positions in nucleotides. K18(80) and K18(41), the two forms of the promoter inserted into luciferase expression vectors are shown.
Figure 2 Effect of E1A-13S or E1A-d30-76 proteins on the activity of mutated versions of the K18 promoter. SW613-3 (T) or SW613-B3 (NT) cells were transfected with each of the reporter plasmids carrying the indicated mutated form of the K18(80) promoter, alone or in the presence of plasmid pE1A-13S (A and C) or plasmid pE1A-d30-76 (B and D). Luciferase activities are expressed relative to that obtained with the promoter construct alone in each cell line. Since the activities of all promoter constructs alone were adjusted to 100, they are all represented by a single pair of columns in each panel. The activity of these constructs relative to that of the wild type K18(80) promoter has been reported previously [22]. The structure of the various mutant forms of the K18(80) promoter is schematized in E. Error bars represent the standard deviation of the mean of duplicate experimental points.
Figure 3 Effect of C-terminal and N-terminal deletion mutants of the E1A protein on the activity of the K18 promoter. (A) Schematic representation of the structure of the E1A-13S protein and of the deletion mutants. Know domains of E1A-13S (CR1, CR2, CR3 and CR4 and the N-terminal region) are shown as boxes. Numbers indicate positions in aa. (B and C) SW613-3 (T) or SW613-B3 (NT) cells were transfected with plasmid pK18(80)luc alone (none) or in the presence of the indicated E1A constructs. Luciferase activities are expressed relative to that obtained with the pK18(80)luc plasmid alone. Error bars represent the standard deviation of the mean of at least two experiments.
Figure 4 Expression level of the E1A constructs. SW613-3 cells were transfected with plasmid constructs coding for the indicated wild-type or mutant E1A proteins. Cellular extracts were analyzed by Western blotting with an anti-E1A antibody. (A) N- and C-terminal deletion mutants. The E1A-C29, -C79 and -C109 polypeptides cannot be detected with this antibody. (B) Internal deletion mutants. Note that E1A and some of its mutants are very sensitive to proteolysis (white asterisks) in spite of the presence of protease inhibitors. Similar results were obtained in SW613-B3 cells (not shown).
Figure 5 Effect of internal deletion mutants of the E1A protein on the activity of the K18 promoter. (A) Schematic representation of the structure of the mutants. Symbols are the same as in figure 3. (B) SW613-3 (T) and SW613-B3 (NT) cells were transfected with plasmid pK18(80)luc alone (none) or in the presence of the indicated E1A constructs. Results are presented as in figure 3. Error bars represent standard deviation of the mean of at least two experiments.
Figure 6 CtBP1 has a role in the stimulatory effect of E1A mutants on the K18 promoter. (A) CtBP1 interacts with E1A-N190 but not with the mutant E1A-N190-mut 281-282. GST pull-down assays were performed on extracts of SW613-3 (3) or SW613-B3 (B3) cells using GST alone (empty vector) or the fusion proteins GST-E1A-N190 or GST-E1A-N190-mut281-282. Bound proteins were analyzed by Western blotting with an anti-CtBP1 antibody. (B) Effect of the E1A-N190 or E1A-N190-mut281-282 proteins on the activity of the K18 promoter. SW613-3 (T) or SW613-B3 (NT) cells were transfected with plasmid pK18(80)luc alone (none) or in the presence of plasmid pE1A-N190 or pE1A-N190-mut281-282 as indicated. Results are presented as in figure 3. Error bars represent standard deviation of the mean of duplicate experimental points. (C) Reversion of the stimulatory effect of E1A-N190 by CtBP1. SW613-3 (T) or SW613-B3 (NT) cells were transfected with plasmid pK18(80)luc alone (none) or in the presence of plasmid pE1A-N190 (2 μg) or pcDNA3-CtBP (24 μg) or with the three plasmids together, as indicated. Results are presented as in figure 3. Error bars represent standard deviation of the mean of triplicate experimental points. (D) The CtBP1 protein has no direct effect on the activity of the K18 promoter. SW613-3 (T) or SW613-B3 (NT) cells were transfected with plasmid pK18(80)luc alone (none) or in the presence of various amounts of plasmid pcDNA3-CtBP, as indicated. Results are presented as in figure 3. Error bars represent standard deviation of the mean of triplicate experimental points.
Figure 7 Expression level of the E1A-N190 and CtBP1 constructs. (A) SW613-B3 or SW613-3 cells were transfected with plasmid pE1A-N190 (wt) or pE1A-N190-mut281-282 (mut). Cellular extracts were analyzed by Western blotting with an anti-E1A antibody. (B) SW613-3 (3) or SW613-B3 (B3) cells were transfected with plasmid pcDNA3-CtBP or untransfected (mock). Cellular extracts were analyzed by Western blotting with an anti-CtBP1 antibody.
Figure 8 The BRCA1 protein has a role in the repression of the K18 promoter in NT cells. (A) SW613-3 (T) or SW613-B3 (NT) cells were transfected with plasmid pK18(80)luc alone (none) or in the presence of 24 μg of plasmids pCBRCA1C (BRCA1wt), pcDNA3-BRCA1(Y1853→STOP) or pcDNA3-BRCT, as indicated. Results are presented as in figure 3. Error bars represent standard deviation of the mean of three experiments. (B) and (C) Expression level of the BRCA1 constructs. SW613-3 (3) or SW613-B3 (B3) cells were untransfected (mock) or transfected with plasmid pCBRCA1C (wt), pcDNA3-BRCA1(Y1853→STOP) (mut) or pcDNA3-BRCT, as indicated. Cellular extracts were analyzed by Western blotting with an anti-HA antibody.
Figure 9 A model for the putative mechanism responsible for the differential activity of the K18 promoter and for the effect of the different E1A mutants. The E1A full-length protein is represented at the top with emphasis on the two domains of interest: aa 12 to 25 and the PLDLS motif. The horizontal line represents the DNA and the initiation site is symbolized by an arrow, whose thickness is an indication of the activity of the promoter. (A) Functioning of the K18 promoter in T and NT cells. The F factor is an as yet unidentified protein which has a HAT/FAT activity and is supposedly more abundant or more active in T cells (bold) than in NT cells. F acetylates (Ac) the substrate protein S, a non-histone protein whose acetylation level controls the activity of the K18 promoter. The acetylation level of S is also under the control of HDAC proteins. The CtBP1 protein interacts with the PLDLS motif of the CtIP protein. CtBP1 associates with some HDAC proteins by a domain that is different from the one recognizing the PLDLS motif [4, 48]. The recruited HDAC proteins deacetylate the S substrate in NT cells where the acetylase activity is already weak and this down-regulates the promoter. CtIP is an interaction partner of BRCA1 which is represented here as part of the preinitiation complex (PIC). (B) Effects of C-terminal deletion mutants of E1A on the activity of the K18 promoter in T and NT cells. The binding of these mutants to the F factor through the 12–25 domain results in a low HAT/FAT activity, deacetylation of S and inhibition of the promoter in T cells. These mutants are expected to have little effect in NT cells since the level of acetylation of S is already low in these cells. (C) Effects of N-terminal deletion mutants of E1A on the activity of the K18 promoter. The C-terminal part of E1A contains the PLDLS motif which permits its interaction with CtBP1. The displacement of CtBP1 from CtIP prevents the recruitment of HDAC proteins to the preinitiation complex. This is of no consequence in T cells where the S protein is maintained in an hyperacetylated state by the high HAT/FAT activity. In contrast, in NT cells removal of HDAC proteins allows the weak HAT/FAT activity to acetylate S to a high level which results in a stimulation of the promoter.
==== Refs
Gallimore PH Turnell AS Adenovirus E1A: remodelling the host cell, a life or death experience Oncogene 2001 20 7824 7835 11753665 10.1038/sj.onc.1204913
Ben Israel H Kleinberger T Adenovirus and cell cycle control Front Biosci 2002 7 d1369 d1395 11991831
Frisch SM Mymryk JS Adenovirus-5 E1A: paradox and paradigm Nat Rev Mol Cell Biol 2002 3 441 452 12042766 10.1038/nrm827
Chinnadurai G CtBP, an unconventional transcriptional corepressor in development and oncogenesis Mol Cell 2002 9 213 224 11864595 10.1016/S1097-2765(02)00443-4
Schaeper U Boyd JM Verma S Uhlmann E Subramanian T Chinnadurai G Molecular cloning and characterization of a cellular phosphoprotein that interacts with a conserved C-terminal domain of adenovirus E1A involved in negative modulation of oncogenic transformation Proc Natl Acad Sci U S A 1995 92 10467 10471 7479821
Schaeper U Subramanian T Lim L Boyd JM Chinnadurai G Interaction between a cellular protein that binds to the C-terminal region of adenovirus E1A (CtBP) and a novel cellular protein is disrupted by E1A through a conserved PLDLS motif J Biol Chem 1998 273 8549 8552 9535825 10.1074/jbc.273.15.8549
Wong AK Ormonde PA Pero R Chen Y Lian L Salada G Berry S Lawrence Q Dayananth P Ha P Tavtigian SV Teng DH Bartel PL Characterization of a carboxy-terminal BRCA1 interacting protein Oncogene 1998 17 2279 2285 9811458 10.1038/sj.onc.1202150
Yu X Wu LC Bowcock AM Aronheim A Baer R The C-terminal (BRCT) domains of BRCA1 interact in vivo with CtIP, a protein implicated in the CtBP pathway of transcriptional repression J Biol Chem 1998 273 25388 25392 9738006 10.1074/jbc.273.39.25388
Li S Chen PL Subramanian T Chinnadurai G Tomlinson G Osborne CK Sharp ZD Lee WH Binding of CtIP to the BRCT repeats of BRCA1 involved in the transcription regulation of p21 is disrupted upon DNA damage J Biol Chem 1999 274 11334 11338 10196224 10.1074/jbc.274.16.11334
Lavialle C Modjtahedi N Cassingena R Brison O High c-myc amplification level contributes to the tumorigenic phenotype of the breast carcinoma cell line SW 613-S (correction) Oncogene 1988 3 335 339 3060796
Cherif D Lavialle C Modjtahedi N Le Coniat M Berger R Brison O Selection of cells with different chromosomal localizations of the amplified c-myc gene during in vivo and in vitro growth of the breast carcinoma cell line SW 613-S Chromosoma 1989 97 327 333 2707103 10.1007/BF00371974
Lavialle C Modjtahedi N Lamonerie T Frebourg T Landin RM Fossar N Lhomond G Cassingena R Brison O The human breast carcinoma cell line SW 613-S: an experimental system to study tumor heterogeneity in relation to c-myc amplification, growth factor production and other markers (review) Anticancer Res 1989 9 1265 1279 2686529
Lamonerie T Lavialle C Haddada H Brison O IGF-2 autocrine stimulation in tumorigenic clones of a human colon-carcinoma cell line Int J Cancer 1995 61 587 592 7759165
Donzelli M Bernardi R Negri C Prosperi E Padovan L Lavialle C Brison O Scovassi I Apoptosis-prone phenotype of human colon carcinoma cells with a high level amplification of the c-myc gene Oncogene 1999 18 439 448 9927200 10.1038/sj.onc.1202309
Lavialle C Modjtahedi N Cassingena R Brison O High c-myc amplification level contributes to the tumorigenic phenotype of the human breast carcinoma cell line SW 613-S Oncogene 1988 3 335 339 3060796
Modjtahedi N Lavialle C Poupon MF Landin RM Cassingena R Monier R Brison O Increased level of amplification of the c-myc oncogene in tumors induced in nude mice by a human breast carcinoma cell line Cancer Res 1985 45 4372 4379 4028021
Modjtahedi N Frebourg T Fossar N Lavialle C Cremisi C Brison O Increased expression of cytokeratin and ferritin-H genes in tumorigenic clones of the SW 613-S human colon carcinoma cell line Exp Cell Res 1992 201 74 82 1377134 10.1016/0014-4827(92)90349-D
Modjtahedi N Haddada H Lavialle C Lazar E Lamonerie T Brison O TGF-alpha production correlates with tumorigenicity in clones of the SW613-S human colon carcinoma cell line Int J Cancer 1992 52 483 490 1399125
Lamonerie T Lavialle C de Galle B Binoux M Brison O Constitutive or inducible overexpression of the IGF-2 gene in cells of a human colon carcinoma cell line Exp Cell Res 1995 216 342 351 7531153 10.1006/excr.1995.1043
Galdemard C Brison O Lavialle C The proto-oncogene FGF-3 is constitutively expressed in tumorigenic, but not in non-tumorigenic, clones of a human colon carcinoma cell line Oncogene 1995 10 2331 2342 7784081
Gunther M Frebourg T Laithier M Fossar N Bouziane-Ouartini M Lavialle C Brison O An Sp1 binding site and the minimal promoter contribute to overexpression of the cytokeratin 18 gene in tumorigenic clones relative to that in nontumorigenic clones of a human carcinoma cell line Mol Cell Biol 1995 15 2490 2499 7537848
Prochasson P Gunther M Laithier M Fossar N Lavialle C Brison O Transcriptional mechanisms responsible for the overexpression of the keratin 18 gene in cells of a human colon carcinoma cell line Exp Cell Res 1999 248 243 259 10094831 10.1006/excr.1999.4402
Prochasson P Delouis C Brison O Transcriptional deregulation of the keratin 18 gene in human colon carcinoma cells results from an altered acetylation mechanism Nucleic Acids Res 2002 30 3312 3322 12140315 10.1093/nar/gkf462
Sang N Caro J Giordano A Adenoviral E1A: everlasting tool, versatile applications, continuous contributions and new hypotheses Front Biosci 2002 7 d407 d413 11815298
Quinlan MP Whyte P Grodzicker T Growth factor induction by the adenovirus type 5 E1A 12S protein is required for immortalization of primary epithelial cells Mol Cell Biol 1988 8 3191 3203 2974923
Sum EY Peng B Yu X Chen J Byrne J Lindeman GJ Visvader JE The LIM domain protein LMO4 interacts with the cofactor CtIP and the tumor suppressor BRCA1 and inhibits BRCA1 activity J Biol Chem 2002 277 7849 7856 11751867 10.1074/jbc.M110603200
Scully R Anderson SF Chao DM Wei W Ye L Young RA Livingston DM Parvin JD BRCA1 is a component of the RNA polymerase II holoenzyme Proc Natl Acad Sci U S A 1997 94 5605 5610 9159119 10.1073/pnas.94.11.5605
Chiba N Parvin JD The BRCA1 and BARD1 association with the RNA polymerase II holoenzyme Cancer Res 2002 62 4222 4228 12154023
Avvakumov N Kajon AE Hoeben RC Mymryk JS Comprehensive sequence analysis of the E1A proteins of human and simian adenoviruses Virology 2004 329 477 492 15518825
Boyd JM Loewenstein PM Tang Qq QQ Yu L Green M Adenovirus E1A N-terminal amino acid sequence requirements for repression of transcription in vitro and in vivo correlate with those required for E1A interference with TBP-TATA complex formation J Virol 2002 76 1461 1474 11773419
Wang HG Rikitake Y Carter MC Yaciuk P Abraham SE Zerler B Moran E Identification of specific adenovirus E1A N-terminal residues critical to the binding of cellular proteins and to the control of cell growth J Virol 1993 67 476 488 8416379
Reid JL Bannister AJ Zegerman P Martinez-Balbas MA Kouzarides T E1A directly binds and regulates the P/CAF acetyltransferase EMBO J 1998 17 4469 4477 9687513 10.1093/emboj/17.15.4469
Chakravarti D Ogryzko V Kao HY Nash A Chen H Nakatani Y Evans RM A viral mechanism for inhibition of p300 and PCAF acetyltransferase activity Cell 1999 96 393 403 10025405 10.1016/S0092-8674(00)80552-8
Gedrich RW Bayley ST Engel DA Induction of AP-1 DNA-binding activity and c-fos mRNA by the adenovirus 243R E1A protein and cyclic AMP requires domains necessary for transformation J Virol 1992 66 5849 5859 1326646
Lipinski KS Esche H Brockmann D Amino acids 1-29 of the adenovirus serotypes 12 and 2 E1A proteins interact with rap30 (TF(II)F) and TBP in vitro Virus Res 1998 54 99 106 9660075 10.1016/S0168-1702(98)00003-3
Severino A Baldi A Cottone G Han M Sang N Giordano A Mileo AM Paggi MG De Luca A RACK1 is a functional target of the E1A oncoprotein J Cell Physiol 2004 199 134 139 14978742 10.1002/jcp.10448
Grooteclaes M Deveraux Q Hildebrand J Zhang Q Goodman RH Frisch SM C-terminal-binding protein corepresses epithelial and proapoptotic gene expression programs Proc Natl Acad Sci U S A 2003 100 4568 4573 12676992 10.1073/pnas.0830998100
Schuierer M Hilger-Eversheim K Dobner T Bosserhoff AK Moser M Turner J Crossley M Buettner R Induction of AP-2alpha expression by adenoviral infection involves inactivation of the AP-2rep transcriptional corepressor CtBP1 J Biol Chem 2001 276 27944 27949 11373277 10.1074/jbc.M100070200
Sundqvist A Bajak E Kurup SD Sollerbrant K Svensson C Functional knockout of the corepressor CtBP by the second exon of adenovirus E1a relieves repression of transcription Exp Cell Res 2001 268 284 293 11478854 10.1006/excr.2001.5280
Li H Lee TH Avraham H A novel tricomplex of BRCA1, Nmi, and c-Myc inhibits c-Myc-induced human telomerase reverse transcriptase gene (hTERT) promoter activity in breast cancer J Biol Chem 2002 277 20965 20973 11916966 10.1074/jbc.M112231200
Zheng L Pan H Li S Flesken-Nikitin A Chen PL Boyer TG Lee WH Sequence-specific transcriptional corepressor function for BRCA1 through a novel zinc finger protein, ZBRK1 Mol Cell 2000 6 757 768 11090615 10.1016/S1097-2765(00)00075-7
Fossar N Chaouche M Prochasson P Rousset M Brison O Deregulated expression of the keratin 18 gene in human colon carcinoma cells Somat Cell Mol Genet 1999 25 223 235 11586789 10.1023/A:1019231926567
Choi CH Hiromura M Usheva A Transcription factor IIB acetylates itself to regulate transcription Nature 2003 424 965 969 12931194 10.1038/nature01899
Choi CH Burton ZF Usheva A Auto-acetylation of transcription factors as a control mechanism in gene expression Cell Cycle 2004 3 114 115 14712067
Fuchs M Gerber J Drapkin R Sif S Ikura T Ogryzko V Lane WS Nakatani Y Livingston DM The p400 complex is an essential E1A transformation target Cell 2001 106 297 307 11509179 10.1016/S0092-8674(01)00450-0
Magnard C Bachelier R Vincent A Jaquinod M Kieffer S Lenoir GM Venezia ND BRCA1 interacts with acetyl-CoA carboxylase through its tandem of BRCT domains Oncogene 2002 21 6729 6739 12360400 10.1038/sj.onc.1205915
Speicher DW Coligan JE, Dunn BM, Ploegh HL, Speicher DW and Wingfield PT Electrophoresis Current Protocols in Protein Science 1999 New York, John Wiley & Sons 10.0.1 10.13.16
Zhang CL McKinsey TA Lu JR Olson EN Association of COOH-terminal-binding protein (CtBP) and MEF2-interacting transcription repressor (MITR) contributes to transcriptional repression of the MEF2 transcription factor J Biol Chem 2001 276 35 39 11022042 10.1074/jbc.M007364200
| 15831101 | PMC1087485 | CC BY | 2021-01-04 16:22:25 | no | BMC Mol Biol. 2005 Apr 14; 6:8 | utf-8 | BMC Mol Biol | 2,005 | 10.1186/1471-2199-6-8 | oa_comm |
==== Front
BMC Med EducBMC Medical Education1472-6920BioMed Central London 1472-6920-5-111580436010.1186/1472-6920-5-11DebateDeveloping essential professional skills: a framework for teaching and learning about feedback Henderson Penny [email protected] Anne C [email protected] Martin H [email protected] Department of Anatomy, University of Cambridge, Downing Street, Cambridge, CB2 3DY, UK2005 1 4 2005 5 11 11 20 12 2004 1 4 2005 Copyright © 2005 Henderson et al; licensee BioMed Central Ltd.2005Henderson 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 ability to give and receive feedback effectively is a key skill for doctors, aids learning between all levels of the medical hierarchy, and provides a basis for reflective practice and life-long learning. How best to teach this skill?
Discussion
We suggest that a single "teaching the skill of feedback" session provides superficial and ineffective learning in a medical culture that often uses feedback skills poorly or discourages feedback. Our experience suggests that both the skill and the underlying attitude informing its application must be addressed, and is best done so longitudinally and reiteratively using different forms of feedback delivery. These feedback learning opportunities include written and oral, peer to peer and cross-hierarchy, public and private, thereby addressing different cognitive processes and attitudinal difficulties.
Summary
We conclude by asking whether it is possible to build a consensus approach to a framework for teaching and learning feedback skills?
==== Body
Background
Notwithstanding many recent changes to the medical curriculum, medical teaching retains a strong apprentice-based element, in which experienced senior doctors pass on their knowledge and skills to students and juniors. The requirements of the profession demand both extensive acquisition of knowledge and a high level of specialist skill development. Multiple academic and qualification hurdles have to be surmounted and there is a highly structured promotion system. This hierarchy of skills and knowledge has consequences for the profession's ethos. First, juniors inevitably know and can do less than seniors. If ineffective or inadequate negative or positive feedback is given, juniors may either develop "false confidence" or become demoralised and fearful of making mistakes. This fear can easily lead to a culture of criticism or blame and so to defensiveness, closing down the junior's openness to learning. Either outcome can be ultimately dispiriting, especially in a highly stressful profession [1,2]. Second, the hierarchical nature of the profession discourages feedback from junior to senior doctors making for a uni-directional "them and us" educational and professional experience. The consequences for the profession of this difficulty in effectively challenging authority figures can be adverse for both the reputation of the profession and the welfare of the patient [3,4]. Given the importance placed on continuing professional development at all levels of the profession, including specifically the requirement for new doctors to be competent in a range of teaching and learning skills [5], the skills of giving and receiving feedback effectively both across and within the hierarchy assume particular importance.
Learning effectively from feedback requires that it is given in a way that helps the recipient to listen to it, receive it constructively, reflect on it, and consider how to take action as a result. Technical aspects of how to give or receive effective feedback after a teaching session have been discussed extensively (see for example [6,7]) and tools are being developed for assessing competence to inform the content of feedback [8,9]. This paper is not concerned with these aspects, which we accept as essential prerequisites for an effective teaching session. We are concerned with how to develop a work culture into which is built the learning and application of the skills required for giving and receiving both positive and negative feedback, regardless of the position of colleagues in the hierarchy, such that the use of these skills becomes second nature and not "an act of bravery" or something to avoid if at all possible. We base our discussion on 14 years of experiential development of a strategic approach to the learning and assessment of the processes and skills of giving and receiving feedback by pre-clinical medical students at Cambridge University (details of the general nature and structure of the course, or of components of it, have been published previously [10-15]). Our objective here is to illustrate some of the key features we have found to be important, as a stimulus for others to share different or similar experiences, and so perhaps to build a consensus approach to a framework for teaching and learning feedback skills?
Discussion
Overall structure of the feedback learning process
Initially, feedback was taught as a "technical skill" in a single session and the students were then expected to apply it – the "immunisation" approach. This approach was ineffective and now feedback is taught both conceptually and practically as a reiterative developmental strand throughout a whole year. Table 1 illustrates those course components during which elements of the skills of feedback giving and receiving are addressed explicitly. Key developmental steps in the learning sequence are described below (with reference to Table section in parenthesis).
Table 1 Feedback components of the course*
Structures Skills Attitudes Awareness
1. "Group contract" for workshops (week 1) Negotiation Self & mutual respect Uses student's 'hopes and fears' to create group safety & self responsibility
2. Group task involving cooperation in a physical activity (week 2) Distinguish descriptive and evaluative comments Respect for value of different roles in a group Awareness of the differing power of types of words and of one's own value assumptions
3. Skill building pairs to develop and share listening/consultation skills (weeks 2–16) Direct, honest communication; application of theory about communication skills Trust Openness Empathy develops awareness of difference; e.g. experiences of being heard influence listening styles
4. Assertiveness exercises (week 4) Getting to a mutually acceptable outcome; assertiveness Mutual and self respect One's own natural style and the impact of stress on it; how to change it or tailor it to situations
5. Explicit feedback exercise (week 7) (see Additional file 1) Observation Reflection Assertiveness Respect
Resilience Willingness to take risks; clarify own professional style;
mistakes as opportunities for learning
6. Written self evaluation and feedback to one other student after giving a dissertation based seminar to the Department Synthesis of content and style Value of clear oral communication Awareness of strengths and weaknesses of own written and oral communication skills
7. Feedback structure to review video role plays (weeks 12–13) Identifying specifics; creativity about alternatives Curiosity about what works and why Capacity to receive, acknowledge, value & use appreciation and criticism
8. Feedback to course organisers through:
8a. feedback forms about each seminar/talk;
8b. weekly workshop evaluations;
8c. student representation at course management meeting (throughout) Verbal and written feedback
Distinguish opinion from evidence-based description Risk taking Feeling entitled to express both appreciation and suggestions to course team
9. Feedback to external examiner (after course end) Synthesis of observation, evaluation, theory & experience Self respect Sensitivity about giving feedback to a senior
10. Summative assessment in practical exam (giving written feedback to a consultant on their consultation skills) Synthesis of observation, evaluation, theory & experience Self respect Sensitivity about giving feedback to a senior
*The course described is taken by third-year pre-clinical students between the completion of their basic science courses and their entry into clinical school. It has 16 weeks of teaching, plus two vacations of 5 weeks each and a period of 5 weeks for independent study including research on a dissertation plus follow up reading. The course is assessed summatively by four written papers, a practical examination and two oral examinations, one based around the dissertation and one centred on the demonstration of acquired skills, one of which is "giving and receiving feedback". The course seeks to develop the integration of scientific and evidential knowledge with personal and professional skills. The emphasis throughout is on a sound understanding of theory grounded in practical experimentation and focussed through structured reflection. The aim is to foster integrated and enduring personal and professional development.
Developing useful feedback tools
(i) Building respect, responsibility and empathy (1)
The value of effective feedback is exemplified at their first session, when the students build their own guidelines for working as a "professional development group". This process is structured to encourage identification of their hopes for learning on the course and their concerns about what might prevent learning. From these shared hopes and concerns, they negotiate a working list of behaviours to maximise hopes and manage concerns. A crucial element that emerges during this process is a requirement to treat each other with respect regardless of their individual level of experience, knowledge and difference (ethnicity, gender, religion, sexuality etc). Respect is a crucial element underpinning the ability to give and to receive feedback. Because this process is shared and student-driven at the outset, they take responsibility for the culture and outcome of course sessions. It also encourages students from the outset to address their own behaviour and attitudes and those of their colleagues and teachers critically, constructively and empathically within an agreed learning framework, and serves as a basis for the later specific development of feedback skills.
(ii) Distinguishing description from evaluation (2)
Whilst students understand the distinction between observation and interpretation when looking at, for example, pathological sections or clinical signs, they have more difficulty doing so when observing human behaviour. The conflation of the two in their every day experiences makes an analytical approach to human behaviour a key learning step to effective feedback. An early exercise, in which groups of students undertake a videoed physical group activity, which is then played back to them, encourages each student to observe identified peers and record what they see descriptively. They then report back in turn and other students, including those observed, challenge non-descriptive "interpretive" statements and offer alternative suggestions. This exercise raises awareness of the differing power of types of words and of one's own value assumptions and in the process builds respect for the value of different types of role in a group activity.
(iii) Assertiveness (4)
Students engage in three exercises to explore attitudes and behaviour appropriate for taking a difficult situation to an effective outcome through direct communication. During the exercises, the students are also encouraged to identify their own "natural" attitudes and behaviour in potentially difficult situations and to determine what might assist them to develop and apply different and even more effective ones. The outcome from this exercise is a list of useful strategies to facilitate effective direct communication. These exercises do not address directly the giving of feedback among group members, but do build skills and encourage attitudes that are useful for doing this such as self and mutual respect, and empathy.
Feedback across a hierarchy
(i) Building confidence and skills interactively in writing (8b)
Each week, students write evaluations of their workshop learning experiences (details in [14]). In these, students include suggestions to the facilitators for different or more effective ways of achieving the workshop objectives or to make explicit to the facilitators when a workshop worked well for them and why and how it worked or did not. Facilitators respond to these evaluations to encourage such inclusions and to develop a good evidence-based evaluation. This procedure illustrates the legitimacy and value of giving feedback to "superiors" and also allows teachers to challenge and encourage students who are reluctant to engage in this risky process. This weekly exercise therefore builds the concept of feedback gently and promotes development of the skills to do so progressively. Facilitators will offer comments to the students about ways they can give suggestions in a manner which is not offensive or patronising and how to distinguish opinion from evidence-based description. This is particularly valuable when students have had a difficult or uncomfortable experience and are conveying both their emotional state (upset) and their suggestions for how the exercise could have been done differently.
(ii) Feedback to help organisers and teachers with their own skills (8a & 9)
Further opportunities for feeding back "upwards" are provided in the confidential course critique to the external examiner, completed by all students, and used by course organisers to plan changes for the following year. In addition, all seminars and lectures are followed by completion of a short standard feedback form (occupying 1/4 of an A4 sheet) issued after each teaching session. The form includes a 1–5 score on overall value plus a space for qualitative comments. This small format was developed to address 'feedback fatigue' and overuse of paper, and has resulted this year in feedback rates of 72.5% for term one and 68% for term two (number of students = 18). Comments of great value to both organizers and teachers have been received.
One-to-one peer-to-peer feedback (3 & 6)
In week 2, the students initiate a weekly "homework" session in the form of pair-work to build and practice a new communication micro-skill (see [13] for details). The relevant point is that, in these pairings, students are also asked to give each other feedback on how their skills are developing and what is more and less effective about their listening and responding skills. Thus, built into the course is a weekly opportunity for mutual feedback with a clear and valued purpose. The feedback function is re-emphasised at the debriefings of the pair work in the subsequent workshop [13]. Each student also gives a seminar about their dissertation, and another is nominated to give them written feedback about their presentation, clearly distinguishing content and presentation issues. Each student submits a self-evaluation, which is submitted for summative examination (see later).
Giving feedback in public
(i) The first group feedback exercise (5)
The group exercise in week 7 represents a critical point in the feedback learning strategy, which most students find demanding. Up to this point, students have functioned in a culture in which inter-individual feedback is encouraged as a useful tool to help learning. In this exercise, we "go public" and ask for the giving and receiving of feedback in small groups of 6–8 plus a facilitator. We first run a group problem-solving exercise after which each student gives and receives feedback on their "performance" in the task to and from group peers. Each then comments how the feedback process has affected them (see Additional file 1). Resulting from this process, students draw up a check-list of behaviours that aid the giving and receiving of feedback. They compare their own lists with published examples. Throughout, and as previously, they are encouraged to be "allies for each other" in the process of learning by describing and praising achievement and by pointing out alternative possibilities to peers for their less successful actions or comments, but this exercise is a critical test of any mismatch between intention and skill acquisition. In particular, the distinction between description and evaluation becomes critically important. One student, who had a difficult experience in making this distinction effectively, wrote about it thus:
"I did not have a particularly enjoyable time and felt uncomfortable for part of it. ....... in the small group feedback exercise in which I was the observer. I will be the first to admit that the way in which I gave my feedback, despite my best efforts to spare people's feelings, was not good. At times, I felt unsupported whilst giving feedback. It was clear to everyone that it was not going well and to further matters, I felt that the facilitator, herself, could have been more supportive. I will be the first to acknowledge that this perception could have been distorted by my own insecurities regarding my performance. Still, the fact remains that the exercise affected my mood for the rest of the day. ....... there were many things that came out of the workshop that I found useful and thought-provoking such as the difficulty of giving feedback and ways in which the destructive capacity of negative criticism can be minimised. I have tried to be honest [in my feedback] and hope that I have not caused offence – that was certainly not my intention. I feel it to be more beneficial if I express my thoughts in this way than if I were to write an evaluation along the lines of "this is what we did and this is why it was useful."
This feedback demonstrates how hard it can be for students to engage in this process, and also how effectively they can learn from it as demonstrated by the way this feedback was given to the senior person who facilitated the session.
(ii) The second group feedback exercise (7)
A more extended public opportunity occurs towards the end of the course, when a day is spent working with actor patients with whom students take turns to practise brief consultations, each of which is videoed and then replayed. Based broadly but not rigidly on 'Pendleton's rules' for feedback [6,7], each student first assesses their own performance for what they did well, and then the rest of the group add their comments. The same sequential process is repeated to garner ideas for what they could do differently. The process emphasises that criticism, which comes with a specific suggestion for alternative behaviour can, under the right conditions, feel more like a gift than an attack. In this session, students display the steady build of giving/receiving feedback skills that has occurred throughout the year as reflected in their feedback evaluations.
"....... the 'doctor' was given a chance for self-analysis first and then was able to see how realistic they were in their praise and criticism of their interview. In most cases we found that we were overly critical of our own performances and that some skills we thought we were carrying out badly or over-using were in fact deemed good or useful by the others." (Value of feedback as an independent appraisal of own skills).
"When I make mistakes I become very aware of the thing I did wrong, so the constructive criticism I received after my consultation was invaluable to me. Since beginning this course I find that I welcome feedback, negative as well as positive, whereas before I was scared of it. Being able to improve after mistakes or listening to other suggestions on your current approach, I think, is a very important skill to have."(Ability to use criticism constructively).
"I was happy to see an improvement in my "performance" throughout the day and this was much assisted by having specific feedback. Such feedback allows us to learn from our mistakes as well as from others, and could be seen as mirroring the reflective practice that I hope to achieve in my career. Again the way in which we should give and accept feedback was reinforced in the exercises. We should offer it as suggestions of things that could be done differently as opposed to better, (the latter is purely opinion), and we should be as specific as possible. In fact, I often find that receiving feedback that addresses the process of how to change is particularly effective. We need only accept feedback that we agree with and that we can comfortably change, unless it detracts from the care of our patients. The initial comments I received regarded the speed of my speech, phrasing of some question and fidgeting. Just having an awareness of this allowed me to correct these the second time I had a go." (Recognition of different types and values of feedback and how to use them rapidly to make and monitor change).
"I've realised just how useful examples can be, in convincing me of the validity of both complimentary and critical comments, so I make sure to try and use them when commenting on others."(Clear distinction between observation and opinion, and value thereof).
Summative assessment (10)
During the year, teaching sessions are not subject to summative assessment and thus provide a space to take risks and develop skills. However, it is made clear at the course outset that the ability to give and receive feedback will be examined summatively. This is achieved in two ways. First, a practical exam consists of a video of a consultation. The students are required to analyse the overall structure of the consultation and the effectiveness with which communication skills were used within it. They are also asked to write feedback to the Dr to help them to develop their skills. The feedback skills demonstrated are assigned a mark equal in value to the analytic section. In 2004, the average mark was 70.4% (range 66% to 80%), where 70% is first class, suggesting that students are performing at a high level. Second, examiners read each student's written self-evaluation of their own seminar performance to assess how well students can distinguish content and presentational skills. Some of these points may be picked up in the oral examination. A final opportunity to assess feedback skills, should this seem necessary, occurs within the oral examinations, students may be offered feedback on their performance from the examiners and some are asked to give oral feedback to the examiners on their examining skills.
Concluding comments for debate
This paper suggests key criteria for the effective teaching of feedback, with qualitative reports of learning outcomes. Key features are:
(i) early introduction of elements of the skills required, with one-to-one guidance from the teacher, including clear distinction of descriptive from evaluative comments, building of self respect and responsibility for self, and building of mutual respect;
(ii) reiterated opportunities for private peer practice, with generalised group support at intervals;
(iii) a gradual move to more structured written feedback;
(iv) "public" giving and receiving of feedback in groups. This process creates a more developed skill in the communication of both positive and challenging comments to colleagues and superiors. It develops the student capacity to listen attentively and willingly to suggestions and to use them creatively and positively. It encourages self-awareness, since this state promotes the capacity to take responsibility for how feedback is given or received.
(v) By developing an attitude of self respect allied to the values and skills of assertive communication, students build a sense of entitlement to receive the same from others and the skills to ask for it. They may thus later take the risks inherent in offering feedback to colleagues in the health care profession with care and clarity [16], or even to become a whistleblower should the need be unavoidable [17,18].
These conclusions embody many of the principles identified by educational psychologists as key to adult learning, in which the encouragement of specific social feedback from teachers and peers is used to extend and build the students' own feedback skills [19]. It is our contention that, by being open to feedback and used to receiving suggestions, student learning through their remaining professional life should be richer. Additionally, since effective feedback is essentially an aid to change, it may be a particularly useful therapeutic skill for use with patients who are embarking on behaviour or life style change, such as change to dietary, drug or tobacco use [20]. Likewise, patients or their relatives receiving difficult news that requires them to make adaptive changes to accommodate its consequences in their lives may find the process facilitated by clinical students or doctors who have developed their capacity for effective and supportive feedback.
Have the key features we have found to be important shared by others from their experiences, and can we build a consensus approach to a framework for teaching and learning feedback skills?
Summary
1. Feedback should be taught both conceptually and practically to address attitudes as well as skills, an approach best achieved through longitudinal and reiterative teaching sessions incorporating formative assessment from teachers and peers.
2. The concept of working in mutually respectful partnership towards a shared beneficial outcome, regardless of relative position in the power hierarchy, informs the attitudes underlying feedback giving and thereby the integrity of the practical process.
3. This concept of investment in a shared outcome also reduces the emotional load on both the giver and receiver of feedback.
4. A thorough understanding of the value of distinguishing description from evaluation/interpretation of observed behaviours helps effective technical application of feedback skills.
5. Assessment of feedback skills summatively should be flagged from the course outset and contribute significantly to final grades.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
All authors contributed equally to this work.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Supplementary Material
Additional File 1
"Outline of the group exercise used as a prelude to social feedback – as an example". This file describes briefly the exercise referred to in Table 1 as Exercise 5. It describes a group activity that is then used by students to give structured feedback to each other in a public setting.
Click here for file
Acknowledgements
We thank our former students for their permission to use quotations from their evaluations, and referee Debra Nestel for her helpful suggestions.
==== Refs
Hawton K Clements A Sakarovitch C Simkin S Deeks JJ Suicide in doctors: a study of risk according to gender, seniority and specialty in medical practitioners in England and Wales, 1979–1995 J Epidemiol Community Health 2001 55 296 300 11297646 10.1136/jech.55.5.296
Robinson G Stresses on women physicians: consequences and coping techniques Depress Anxiety 2003 17 180 189 12768652 10.1002/da.10069
MacLean M Bainham A, Sclater SD, Richards M Letting go: parents, professionals and the law in the retention of human material after post mortem Body lores and law 2002 Oxford and Portland: Hart Publishing Ltd
Stern V Gynaecological examinations post-Ledward – a private matter Lancet 2001 358 1896 1898 11741649 10.1016/S0140-6736(01)06892-1
GMC Tomorrow's Doctors: The Duties of a Doctor registered with the General Medical Council 2003 London: General Medical Council
Pendleton D Schofield T Tate P The consultation: An approach to learning and teaching 1984 Oxford: Oxford University Press
Kurtz S Silverman J Draper J Teaching and Learning Communication Skills in Medicine 1998 Oxford: Radcliffe Medical Press
Misch D Evaluating physicians' professionalism and humanism: the case for humanism "connoisseurs" Academic Medicine 2002 77 489 495 12063192
Musick D McDowell S Clark N Salcido R Pilot study of a 360-degree assessment instrument for physical medicine & rehabilitation residency programs Am J Phys Med Rehabil 2003 82 394 402 12704281 10.1097/00002060-200305000-00016
Johnson MH Henderson P Acquiring and demonstrating attitudes in medical education: attitudes to homosexuality as a case study Medical Teacher 2000 22 586 592
Johnson MH Exploring masculinity with male medical students Sexual and Relationship Therapy 2001 16 165 171 10.1080/14681990120040087
Johnson MH Bainham A, Sclater SD, Richards M Male medical students and the male body Body lores and law 2002 Oxford and Portland: Hart Publishing Ltd 91 103
Henderson P Johnson MH Barnett M Weaver S Supporting medical students to take responsibility for developing their communication skills Medical Teacher 2001 23 86 87 11260748 10.1080/0142159002005686
Henderson P Johnson MH An innovative approach to developing the critical evaluation skills of medical students BMC Medical Education 2002 2 4 12003640 10.1186/1472-6920-2-4
Henderson P Johnson MH Assisting medical students to conduct empathic conversations with patients from a sexual medicine clinic Sexually Transmitted Infections 2002 78 246 249 12181460 10.1136/sti.78.4.246
Ginsburg S Regehr G Lingard L The disavowed curriculum: understanding student's reasoning in professionally challenging situations J Gen Intern Med 2003 18 1015 1022 14687260 10.1111/j.1525-1497.2003.21247.x
Burrows J Telling tales and saving lives: whistleblowing – the role of professional colleagues in protecting patients from dangerous doctors Medical Law Review 2001 9 110 129 14982078
Frais A Whistleblowing heroes – boon or burden? Bull Med Ethics 2001 170 13 19 11881660
Palincsar A Social constructivist perspectives on teaching and learning Annual Reviews in Psychology 1998 49 345 375 10.1146/annurev.psych.49.1.345
Rollnick S Mason P Butler C Health behaviour change. A guide for practitioners 1999 Edinburgh, UK: Churchill Livingstone
| 15804360 | PMC1087486 | CC BY | 2021-01-04 16:30:56 | no | BMC Med Educ. 2005 Apr 1; 5:11 | utf-8 | BMC Med Educ | 2,005 | 10.1186/1472-6920-5-11 | oa_comm |
==== Front
BMC Med Inform Decis MakBMC Medical Informatics and Decision Making1472-6947BioMed Central London 1472-6947-5-81578413410.1186/1472-6947-5-8Research ArticleDeveloping optimal search strategies for detecting clinically sound and relevant causation studies in EMBASE Haynes R Brian [email protected] Monika [email protected] Nancy L [email protected] Hedges Team 1 Health Information Research Unit, Department of Clinical Epidemiology and Biostatistics, McMaster University, Hamilton, Ontario, L8N 3Z5, Canada2 Department of Medicine, Faculty of Health Sciences, McMaster University, Hamilton, Ontario, L8N 3Z5, Canada2005 22 3 2005 5 8 8 2 9 2004 22 3 2005 Copyright © 2005 Haynes et al; licensee BioMed Central Ltd.2005Haynes 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
Evaluating the existence and strength of an association between a putative cause and adverse clinical outcome is complex and best done by assessing all available evidence. With the increasing burden of chronic disease, greater time demands on health professionals, and the explosion of information, effective retrieval of best evidence has become both more important and more difficult. Optimal search retrieval can be hampered by a number of obstacles, especially poor search strategies, but using empirically tested methodological search filters can enhance the accuracy of searches for sound evidence concerning etiology. Although such filters have previously been developed for studies of relevance to causation in MEDLINE, no empirically tested search strategy exists for EMBASE.
Methods
An analytic survey was conducted, comparing hand searches of journals with retrievals from EMBASE for candidate search terms and combinations. 6 research assistants read all issues of 55 journals indexed in EMBASE. All articles were rated using purpose and quality indicators and categorized into clinically relevant original studies, review articles, general papers, or case reports. The original and review articles were then categorized as 'pass' or 'fail' for scientific merit according to explicit criteria in the areas of causation (etiology) and other clinical topics. Candidate search strategies were developed for causation, then run in a subset of 55 EMBASE journals, the retrievals being compared with the hand search data. The sensitivity, specificity, precision, and accuracy of the search strategies were calculated.
Results
Of the 1489 studies classified as causation, 14% were methodologically sound. When search terms were combined, sensitivity reached 92%. Compared with the best single-term strategy, the best combination of terms resulted in an absolute increase in sensitivity (19%) and specificity (5.2%). Maximizing specificity for combined terms resulted in an increase of 7.1% compared with the single term but this came at an expense of sensitivity (39% absolute decrease). A search strategy that optimized the trade-off between sensitivity and specificity achieved 81.9% for sensitivity and 81.4% for specificity.
Conclusion
We have discovered search strategies that retrieve high quality studies of causation from EMBASE with high sensitivity, high specificity, or an optimal balance of each.
==== Body
Background
Clinical problems encountered by clinicians often involve examining questions about harm that involve genes, treatments, or environmental exposures [1,2]. Knowledge of a causal relationship is important to clinicians, as it guides their approach to better patient management, and provides recommendations for future research on modifiable environmental risk factors or genetically determined characteristics [3]. With the increasing burden of chronic disease and greater time demands on clinicians and the explosion of research information, effective retrieval of the best evidence has become difficult. Clinicians seldom know of the relevant and rigorous evidence that is available on a particular topic and most often do not attempt to retrieve it even when pertinent to a clinical problem at hand [4,5].
Large biomedical databases such as MEDLINE and EMBASE provide online access to the medical literature [6], and conducting searches in these databases has been recommended as a basic skill for evidence-based practitioners [7]. To make better clinical decisions with the potential of positively affecting the care of their patients, clinicians need ways to optimize their retrieval of the best evidence [8-10]. However, clinicians face a number of obstacles that inhibit optimum search retrieval. The overwhelming amount of available information, coupled with the over 2 million new articles that get published each year [10-12], makes keeping up-to-date challenging and difficult [10,13]. In EMBASE, the European biomedical database counterpart to MEDLINE, clinicians must search through more than 9 million citations from over 4000 journals to narrow their search for best evidence [14].
For clinicians, increasing time demands restrict the practice of evidence-based medicine [4,7], despite the strong belief in its implementation [6,15]. Lack of time is also a major barrier to conducting searches [4,9,15]. Even though the evidence is readily available, clinicians are more likely to seek answers from colleagues [5] or other easily accessible resources than to search for answers with evidence and evaluate the results of original research [16]. As a result, most clinicians do not find answers to their clinical questions or do not pursue them because they have doubt about the existence of useful information in available resources [4,9].
The very low concentration of rigorous studies also limits clinicians' awareness and detection of key articles [13]. Furthermore, clinicians use less than optimal strategies because they lack search skills; do not know how to narrow their searches without missing relevant information; and have uncertainties about when to stop searching, which articles to read, and how thoroughly to read them [16,17].
Methodologic search filters (which capture relevant articles while eliminating those that are not of interest) are one way of improving the retrieval of scientifically sound and clinically relevant studies from biomedical literature databases [18]. Search strategies are useful tools and have been developed for causation studies as well as for studies in other categories (e.g., treatment) for MEDLINE [19,20]. For EMBASE however, very few search strategies have been developed [21]. In fact, we are unable to find an empirically tested search strategy for the retrieval of causation studies in EMBASE.
In this paper, we report on the evaluation and comparison of the retrieval performance of causation search strategies in EMBASE with a manual review ("gold standard") of each article for each issue of 55 journals in 2000. Compared with previous strategies developed for MEDLINE in 1991, the methods we applied for selecting articles for EMBASE were tighter and the calibration database larger (55 journals for EMBASE compared with 10 for MEDLINE in 1991). In addition, we tested many more search strategies, which for MEDLINE resulted in the development of search strategies that work better than the ones previously reported. The focus of the strategies is to help clinicians and researchers retrieve methodologically sound study reports on causation, to assist with evidence-based patient care decisions based on the best quality evidence available. To our knowledge, no approach exists that applies such rigorous standards to EMBASE.
Methods
The study compared the retrieval performance of methodologic search terms and phrases in EMBASE with a manual review of each article for each issue of 55 journal titles for the year 2000. Index terms and text words related to research design features were run as search strategies. The search strategies were treated as "diagnostic tests" for sound studies and the manual review of the literature was treated as the "gold standard." The sensitivity, specificity, precision, and accuracy of EMBASE search strategies were determined. Sensitivity for a given topic is defined as the proportion of high quality articles for that topic that are retrieved; specificity is the proportion of low quality articles not retrieved; precision is the proportion of retrieved articles that are of high quality; and accuracy is the proportion of all articles that are correctly classified.
Individual search terms with sensitivity > 25% and specificity > 75% for causation studies were incorporated into the development of search strategies that included a combination of 2 or more terms. All combinations of terms used the Boolean OR, for example, "risk.tw. OR cohort.tw.". The Boolean AND was not used because this strategy invariably compromised sensitivity. For the development of multiple-term search strategies to either optimize sensitivity or specificity, we tested all 2-term search strategies with sensitivity at least 75% and specificity at least 50%. For optimizing accuracy, 2-term search strategies with accuracy > 75% were considered for multiple-term development. 13,901 search strategies were tested.
We did not attempt to use logistic regression to improve search performance in this study because our previous development of regression strategies for retrieving studies of treatment [unpublished observation] and prognosis [22] showed no benefit.
Six research assistants hand searched 170 journals titles in total for the year 2000, and applied methodologic criteria to each item in each issue to determine if the article was methodologically sound for 7 purpose categories (e.g., causation, treatment, diagnosis; two other types of articles, cost and qualitative studies, were also classified but had no rigor criteria). All purpose category definitions and corresponding methodologic rigor were outlined in a previous paper [23]. The methodologic criteria applied for studies of causation are in Table 1. Research staff were rigorously calibrated before reviewing the 2000 literature and inter-rater agreement for application of all criteria exceeded 80% beyond chance [23].
Table 1 Methodologic Rigor Applied for Studies of Causation
Purpose Category Methodologic Rigor
Causation Observation concerned with the relationship between exposures and putative clinical outcomes;
Data collection is prospective;
Clearly identified comparison group(s);
Blinding of observers of outcome to exposure;
and Analysis consistent with study design.
The 170 journal titles reviewed were chosen based on recommendations of clinicians and librarians, Science Citation Index Impact Factors provided by the Institute for Scientific Information, and ongoing assessment of their yield of studies and reviews of scientific merit and clinical relevance for the disciplines of internal medicine, general medical practice, mental health, and general nursing practice (list of journals provided by the authors upon request). 135 of the 170 journals were indexed in EMBASE. We previously developed search strategies in MEDLINE using the 161 hand-searched journals that were indexed in MEDLINE but found that search strategies developed in much smaller journal subsets are equally robust [24] and that computation time is substantially decreased. We also found that when strategies were developed in 60% of the database and validated in the remaining 40% there were no statistical differences in performance [19]. Thus, for EMBASE we developed search strategies using a 55 journal-subset chosen based on those journals, which had the highest number of methodologically sound studies. This selection somewhat enriches the sample of target articles (those that "pass" for scientific merit) thereby improving the estimates of the sensitivity and specificity search term performance and simplifying data processing. Enriching the prevalence of qualified articles, however, results in overestimates of precision and, to a lesser extent, accuracy. This problem is universal in using a diagnostic testing approach, and is also true for any other classification approach of which we are aware, including machine learning models.
To identify candidate search terms and strategies, we compiled an initial list of index terms and text words by selecting words that related to etiology (eg, etiology, cause, causation) and to research methods for establishing causation (see examples below). We then sought input from clinicians and librarians in the United States and Canada through interviews of known searchers, and requests at meetings and conferences. Individuals were asked to identify terms or phrases they used when searching for studies of causation, prognosis, diagnosis, treatment, economics, clinical prediction guides, reviews, costs, and studies of a qualitative nature. We compiled a list of 5385 terms of which 4843 were unique and 3524 returned results in the 55-journal subset in EMBASE (list of terms tested provided by the authors upon request). Examples of the search terms tested are 'adverse drug reaction', 'risk ratio', 'cohort study', and 'harm', all as text words; 'risk', the index term, and the index term 'exposure', exploded.
Results
Indexing information was downloaded from EMBASE for 27,769 articles from the 55 hand searched journals. Of these, 1489 were classified as causation, of which 215 (14.4%) were judged methodologically sound. Search strategies were developed using all 27,769 articles. Thus, the strategies were tested for their ability to retrieve articles about higher quality causation studies from all other articles, including both lower quality causation studies and all non-causation studies.
The operating characteristics of the best single-term for high-sensitivity, high-specificity, and best optimization of sensitivity and specificity are displayed in Table 2. When specificity was maximized (87.5%), the most noticeable, but expected trade-off was the decrease in sensitivity (21.9% absolute decrease), but there was a slight increase in precision (1.8% absolute increase).
Table 2 Single Term with the Best Sensitivity (keeping Specificity ≥50%), Best Specificity (keeping Sensitivity ≥50%), and Best Optimization of Sensitivity and Specificity (based on the lowest possible absolute difference between sensitivity and specificity) for Detecting Studies of Causation in EMBASE in 2000
Search term OVID search* Sensitivity (%) (n = 215) Specificity (%) (n = 27,554) Precision (%)† Accuracy (%) (n = 27,769)
Best sensitivity
exp general aspects of disease
72.6 (66.6 to 78.5)
55.7 (55.1 to 56.3)
1.3 (1.1 to 1.5)
55.8 (55.2 to 56.4)
Best specificity
exp risk
50.7 (44.0 to 57.4)
87.5 (87.1 to 87.8)
3.1 (2.5 to 3.6)
87.2 (86.8 to 87.6)
Best Optimization of Sensitivity & Specificity
control:.mp.
60.9 (54.5 to 67.5)
66.4 (65.9 to 67.0)
1.4 (1.2 to 1.6)
66.4 (65.8 to 66.9)
*The search strategy is reported using Ovid's search engine syntax for EMBASE.
†Denominator varies by row.
exp = explode, a search term that automatically includes closely related indexing terms; : = truncation; mp = multiple posting – term appears in title, abstract, or subject heading.
Combinations of terms with the best results for sensitivity, specificity, and optimization of sensitivity and specificity are shown in Table 3. As expected, combining terms increased sensitivity. The 3-term combination strategy, "risk:.mp OR exp methodology OR exp epidemiology" yielded the best sensitivity (91.6%) with specificity 60.9%. Compared with the best sensitivity single-term strategy, "exp general aspects of disease", the combination strategy resulted in an absolute increase in both sensitivity (19%) and specificity (5.2%).
Table 3 Combination of Terms with the Best Sensitivity (keeping Specificity ≥50%), Best Specificity (keeping Sensitivity ≥50%), and Best Optimization of Sensitivity and Specificity (based on abs [sensitivity-specificity]<1%) for Detecting Studies of Causation in EMBASE in 2000
Search Strategy OVID search* Sensitivity (%) (n = 215) Specificity (%) (n = 27,554) Precision (%)‡ Accuracy (%) (n = 27,769)
Best Sensitivity
risk:.mp.
OR exp methodology
OR exp epidemiology
91.6 (87.9 to 95.3)
60.9 (60.3 to 61.4)
1.8 (1.6 to 2.0)
61.1 (60.5 to 61.7)
Best Specificity
cohort.tw.
OR relative risk:.tw.
53.0 (46.4 to 59.7)
94.6 (94.4 to 94.9)
7.1 (5.9 to 8.4)
94.3 (94.0 to 94.6)
Small decrease in best specificity with a substantive increase in sensitivity
cohort.tw.
OR relative risk:.tw.
OR adjusted OR†.tw.
61.4 (54.9 to 67.9)
92.9 (92.6 to 93.2)
6.3 (5.3 to 7.3)
92.6 (92.3 to 92.9)
Best Optimization of Sensitivity & Specificity
risk.tw.
OR mortalit:.tw.
OR cohort.tw.
81.9 (76.7 to 87.0)
81.4 (80.9 to 81.8)
3.3 (2.8 to 3.8)
81.4 (80.9 to 81.8)
*Search strategies are reported using Ovid's search engine syntax for EMBASE.
†OR = odds ratio.
‡Denominator varies by row.
: = truncation; mp = multiple posting – term appears in title, abstract, or subject heading; exp = explode, a search term that automatically includes closely related indexing terms; tw = textword (word or phrase appears in title or abstract).
The two-term strategy, "cohort.tw. OR relative risk:.tw." yielded the best specificity (94.6%) but with an expected trade-off in sensitivity, which was lowered to 53% (38.6% absolute decrease). However, maximizing specificity improved both precision (5.3% absolute increase) and accuracy (33.2% absolute increase). The combination of 3 terms, "cohort.tw. OR relative risk:.tw. OR adjusted OR.tw" (where "adjusted OR" is not the Boolean OR but rather refers to adjusted odds ratio) achieved a substantive increase in sensitivity (8.4% absolute increase) with a small decrease in specificity (1.7% absolute decrease) (Table 3). The combination of search terms, "risk.tw. OR mortalit:.tw. OR cohort:.tw." (81.9% sensitivity, 81.4% specificity) led to the best optimization of sensitivity and specificity (Table 3).
Table 4 shows the 3 top-performing search strategies for best sensitivity, best specificity, and best balance between sensitivity and specificity. Because the accuracy of search terms is driven by their specificity, the 3 top-performing search strategies with the best accuracy were similar to those with best specificity. In addition, two 2-term strategies slightly outperformed all the 3-term strategies for best specificity.
Table 4 Top 3-performing Combination of Terms with the Best Sensitivity (keeping Specificity ≥50%), Specificity (keeping Sensitivity ≥50%), and Best Optimization of Sensitivity and Specificity (based on abs [sensitivity-specificity]<1%) for Detecting Studies of Causation in EMBASE in 2000
Search Strategy OVID search* Sensitivity (%)
(n = 215) Specificity (%)
(n = 27,554) Precision (%)† Accuracy (%)
(n = 27,769)
Top 3-performing combination of terms with best Sensitivity
risk:.mp.
OR exp methodology
OR exp epidemiology 91.6 (87.9 to 95.3) 60.9 (60.3 to 61.4) 1.8 (1.6 to 2.0) 61.1 (60.5 to 61.7)
risk:.tw.
OR exp methodology
OR exp epidemiology 91.2 (87.4 to 95.0) 63.0 (62.4 to 63.6) 1.9 (1.6 to 2.2) 63.2 (62.6 to 63.8)
risk:.mp.
OR exp methodology
OR exp mortality 90.7 (86.8 to 94.6) 65.1 (64.5 to 65.7) 2.0 (1.7 to 2.3) 65.3 (64.7 to 65.8)
Top 3-performing combination of terms with best Specificity
cohort.tw.
OR relative risk:.tw. 53.0 (46.4 to 59.7) 94.6 (94.4 to 94.9) 7.1 (5.9 to 8.4) 94.3 (94.0 to 94.6)
confidence interval.tw.
OR relative risk:.tw. 50.7 (44.0 to 57.4) 94.5 (94.2 to 94.7) 6.7 (5.4 to 7.9) 94.1 (93.8 to 94.4)
OR relative risk:.tw.
OR cohort:.tw. 53.5 (46.8 to 60.2) 94.4 (94.1 to 94.6) 6.9 (5.7 to 8.1) 94.1 (93.8 to 94.3)
Top 3-performing combination of terms with best optimization of Sensitivity and Specificity
risk.tw.
OR mortalit:.tw,
OR cohort.tw. 81.9 (76.7 to 87.0) 81.4 (80.9 to 81.8) 3.3 (2.8 to 3.8) 81.4 (80.9 to 81.8)
risk:.tw.
OR cohort:.mp.
OR confidence interval:.tw. 81.9 (76.7 to 87.0) 81.2 (80.8 to 81.7) 3.3 (2.8 to 3.8) 81.3 (80.8 to 81.7)
risk.tw.
OR mortalit:.tw.
OR cohort:.tw 81.9 (76.7 to 87.0) 81.2 (80.8 to 81.7) 3.3 (2.8 to 3.8) 81.2 (80.8 to 81.7)
*Search strategies are reported using Ovid's search engine syntax for EMBASE.
†Denominator varies by row.
: = truncation; mp = multiple posting – term appears in title, abstract, or subject heading; exp = explode, a search term that automatically includes closely related indexing terms; tw = textword (word or phrase appears in title or abstract).
Discussion
We developed causation search filters for EMBASE that provide for highly sensitive, highly specific, and highly accurate searches in EMBASE for high quality studies concerning etiology. The utility of these 3 types of search filters will vary according to the needs of end users or the clinical question that is being sought.
For example, a person conducting a search to find original articles for constructing a systematic review will have different retrieval needs than the clinician who is looking for quick answers to manage a patient. The best sensitive search would be more beneficial for a systematic review. Although it is time consuming to search through 270 citations that may include some irrelevant articles, key studies that are needed to conduct a robust systematic review would not be missed. In contrast, for quick answers, the narrower yield of a specific search takes less time, and will likely provide a sufficient number of relevant articles to answer the clinical question sought, but with somewhat higher potential for missing key studies. The trade-off between time investment and consequences of missing useful evidence is important to consider [21].
Our results indicate that combination-term strategies generally perform better than single-term strategies. However, in our previous research, "risk:.mp", yielded close to best sensitivity in developing causation search filters in MEDLINE [19]. The resulting test characteristics were surprising, as this search resulted in a substantial gain in specificity (26.5% absolute increase) at a very low cost to sensitivity (0.5% absolute decrease). An end user who doesn't have adequate time for a lengthy search will sacrifice only a small decrease in best sensitivity in exchange for a much higher specificity. To test if a similar benefit could be achieved in sensitivity, we also tested the best specificity single-term strategy from our previous MEDLINE strategy, "Risk factor:.mp" in EMBASE. Unfortunately, the small gain in specificity (5.3% absolute increase) was at a very high cost to sensitivity, which was lowered to 35.8% – well below our acceptable prespecified sensitivity at ≥ 50%. Unfortunately, we were only able to do limited comparisons between EMBASE and MEDLINE search strategies, as the two databases do not support the same index terms.
A logistic regression approach to developing search strategies was done when deriving search filters for MEDLINE [22]. The analysis did not improve on search strategies developed using the Boolean approach described above.
Another expected result from our study was that precision was generally low. For a large, multipurpose biomedical database such as EMBASE, it was not surprising to find a low proportion of relevant, high quality causation studies. Although a slight improvement in precision was seen when specificity was maximized, the overall low precision in our study will still require physicians to invest time eliminating irrelevant articles. However, improving precision may be possible by combining search strategies with content-specific terms using the Boolean "AND" or "AND NOT". Our future research will focus on enhancing precision by developing more sophisticated search filters, and by using the strategies above.
Conclusion
We developed several search strategies that can enhance the retrieval of causation articles in EMBASE. The needs of end users play an important role in determining the most beneficial trade-off between sensitivity and specificity.
Competing interests
The authors declare that they have no competing interests.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This research was funded by the National Library of Medicine, USA. Ovid Technologies Inc provided access to the EMBASE database. The Hedges Team includes Angela Eady, Brian Haynes, Susan Marks, Ann McKibbon, Doug Morgan, Cindy Walker-Dilks, Stephen Walter, Stephen Werre, Nancy Wilczynski, and Sharon Wong.
==== Refs
Levine M Walter S Lee H Users' Guide to the Medical Literature. IV. How to Use An Article about Harm JAMA 1994 271 1615 19 8182815 10.1001/jama.271.20.1615
McKibbon A Eady A Marks S PDQ: Evidence-Based Principles and Practice 1999 Hamilton: BC Decker Inc
Fletcher RH Fletcher SW Wagner EH Clinical Epidemiology: The Essentials 1996 3 Baltimore: Williams and Wilkins
McColl A Smith H White P Field F General practitioners' perceptions of the route to evidence based medicine: a questionnaire survey BMJ 1998 316 361 7 9487174
Tomlin Z Humphrey C Rogers S General practitioners' perceptions of effective health care BMJ 1999 318 1532 35 10356011
Suarez-Almazor ME Belseck E Homik J Identifying Clinical Trials in the Medical Literature with Electronic Databases: MEDLINE alone is not enough Control Clin Trials 2000 21 476 87 11018564 10.1016/S0197-2456(00)00067-2
Oxman AD Sackett DL Guyatt GH Users' Guides to the Medical Literature I. How to Get Started JAMA 1993 270 2093 7 8411577 10.1001/jama.270.17.2093
Evidence-Based Medicine Working Group. Evidence-Based Medicine A New Approach to Teaching the Practice of Medicine JAMA 1992 268 2420 5 1404801 10.1001/jama.268.17.2420
McKibbon KA Walker-Dilks CJ The Quality and Impact of MEDLINE searches performed by end users Health Libraries Review 1995 12 191 200 10159236 10.1046/j.1365-2532.1995.1230191.x
Haynes RB Sackett DL Tugwell P Problems in the Handling of Clinical and Research Evidence by Medical Practitioners Arch Intern Med 1983 143 1971 5 6625783 10.1001/archinte.143.10.1971
Bennett NL Casebeer LL Kristofco RE Strasser SM Physicians' Internet information-seeking behaviours J Contin Educ Health Prof 2004 24 31 8 15069910 10.1002/chp.1340240106
Lee KP Schotland M Bacchetti P Bero LA Association of Journal Quality Indicators With Methodological Quality of Clinical Research Articles JAMA 2002 287 2805 8 12038918 10.1001/jama.287.21.2805
Sackett DL Haynes RB Guyatt GH Tugwell P Clinical Epidemiology: A Basic Science for Clinical Medicine 1991 2 Boston: Little, Brown
Elsevier Science Embase [electronic database] Amsterdam, The Netherlands: Elsevier Science [producer] 1974 New York: Ovid Technologies, Inc
Doig GS Simpson F Efficient literature searching: a core skill for the practice of evidence-based medicine Intensive Care Med 2003 29 2119 27 12955188 10.1007/s00134-003-1942-5
Ely JW Osheroff JA Ebell MH Obstacles to answering doctors' questions about patient care with evidence: qualitative study BMJ 2002 324 710 11909789 10.1136/bmj.324.7339.710
Balas EA Stockham MG Mitchell JA In search of controlled evidence for health care quality improvement J Med Syst 1997 21 21 32 9172067 10.1023/A:1022887224126
Haynes RB McKibbon KA Fitzgerald D How to keep up with the Medical Literature: V. Access by Personal Computer to the Medical Literature Ann Int Med 1986 105 810 24 3532898
Wilczynski NL Haynes RB for the Hedges Team Developing optimal search strategies for detecting clinically sound causation studies in MEDLINE Proc AMIA Annu Symp 2003 719 23
Haynes RB Wilczynski N McKibbon KA Walker CJ Sinclair JC Developing optimal search strategies for detecting clinically sound studies in MEDLINE J Am Med Inform Assoc 1994 1 447 458 7850570
Bachmann LM Estermann P Kronenberg C Identifying diagnostic accuracy studies in EMBASE J Med Libr Assoc 2003 91 341 6 12883560
Wilczynski NL Haynes RB The Hedges Team Developing optimal search strategies for detecting clinically sound prognostic studies in MEDLINE: an analytic survey BMC Med 2004 2 23 15189561 10.1186/1741-7015-2-23
Wilczynski NL McKibbon KA Haynes RB Enhancing retrieval of best evidence for health care from bibliographic databases: calibration of the hand search of the literature Medinfo 2001 10 390 93 11604770
Wilczynski NL Haynes RB Robustness of Empirical Search Strategies for Clinical Content in MEDLINE Proc AMIA Symp 2002 904 8 12463956
| 15784134 | PMC1087487 | CC BY | 2021-01-04 23:52:12 | no | BMC Med Inform Decis Mak. 2005 Mar 22; 5:8 | utf-8 | BMC Med Inform Decis Mak | 2,005 | 10.1186/1472-6947-5-8 | oa_comm |
==== Front
BMC NeurosciBMC Neuroscience1471-2202BioMed Central London 1471-2202-6-241581999310.1186/1471-2202-6-24Research ArticleHsp27 and axonal growth in adult sensory neurons in vitro Williams Kristy L [email protected] Masuma [email protected] Karen M [email protected] Division of Basic Medical Sciences, Memorial University of Newfoundland, St. John's, NL, A1B 3V6, Canada2005 8 4 2005 6 24 24 13 11 2004 8 4 2005 Copyright © 2005 Williams et al; licensee BioMed Central Ltd.2005Williams 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
Neurite growth can be elicited by growth factors and interactions with extracellular matrix molecules like laminin. Among the targets of the signalling pathways activated by these stimuli are cytoskeletal elements, such as actin, tubulin and neurofilaments. The cytoskeleton can also be modulated by other proteins, such as the small heat shock protein Hsp27. Hsp27 interacts with actin and tubulin in non-neuronal cells and while it has been suggested to play a role in the response of some neurons to injury, there have been no direct studies of its contribution to axonal regeneration.
Results
We have investigated neurite initiation and process extension using cultures of adult dorsal root ganglion (DRG) sensory neurons and a laminin stimulation paradigm. Employing confocal microscopy and biochemical analyses we have examined localization of Hsp27 at early and later stages of neurite growth. Our results show that Hsp27 is colocalized with actin and tubulin in lamellopodia, filopodia, focal contacts and mature neurites and growth cones. Disruption of the actin cytoskeleton with cytochalasin D results in aberrant neurite initiation and extension, effects which may be attributable to alterations in actin polymerization states. Inhibition of Hsp27 phosphorylation in our cultures results in an atypical growth pattern that may be attributable to an effect of pHsp27 on the stability of the actin cytoskeleton.
Conclusion
We observed colocalization of the phosphorylated and non-phosphorylated forms of Hsp27 with actin and tubulin in both very early and later stages of neurite growth from cultured adult DRG neurons. The colocalization of Hsp27 and pHsp27 with actin in lamellopodia and focal contacts at early stages of neurite growth, and in processes, branch points and growth cones at later stages, suggests that Hsp27 may play a role in neuritogenesis and subsequent neurite extension, and potentially in the patterning of this growth. Hsp27 has been reported to play a key role in modulating actin cytoskeletal dynamics as an actin-capping protein in non-neuronal cells. Our results suggest that this may also be the case in neurons and support a role for Hsp27 in neurite outgrowth via its phosphorylation state-dependent interactions with actin.
==== Body
Background
We know that various factors can influence and promote regeneration of peripheral axons. In addition to soluble factors (neurotrophins, cytokines and other growth factors), the extracellular environment in which growth occurs is critically important. Axonal regeneration does not occur to any great extent in the CNS, and while this is due to a number of factors, the most prominent is a non-permissive growth environment as well as an unavailability of appropriate growth-promoting factors. In the PNS, on the other hand, peripheral axons (both motor and sensory) generally regenerate quite well.
Growth factors and extracellular matrix (ECM) molecules like laminin act through cell surface receptors that activate often convergent signalling pathways to elicit neurite growth in sensory neurons [1]. Among the targets of these pathways are the cytoskeletal elements responsible for initiating and maintaining the structure of growing processes. Actin, tubulin and intermediate filaments all play a part in growth processes [2-4]. There are also a variety of other molecules that interact with these components to modulate or protect the cytoskeleton from deleterious stresses.
One class of molecules known to act as chaperones include the small heat shock protein family, of which heat shock protein 27 is a member. Hsp27, in addition to its roles in regulating apoptosis and protein folding, interacts with different cytoskeletal elements [5-9]. Much of this work has been carried out using non-neural cells, particularly fibroblast and epithelial derived cells. Part of its protective role in stressed cells has been attributed to its actions as an actin-capping protein [10,11]. Hsp27 has been reported to be a component of focal contacts, play an important role in smooth muscle contraction and be important for cellular migration in endothelial cells (reviewed in [12]). Rodent Hsp27 can be phosphorylated on 2 sites, Ser15 and Ser 86, although human Hsp27 has 3 serine phosphorylation sites (S15, S78 and S82) [13,14]. MAPKAP-K2, via its activation by p38 MAPK, is reported to be the Hsp27 kinase, although there are recent reports that PKC α,δ and cAMP-dependent kinase can also phosphorylate Hsp27 [15,16]. In terms of its influence on actin, pHsp27 acts to promote actin polymerization and stress fibre formation. It also has a role in protecting or stabilizing the actin cytoskeleton, although this appears to depend upon the nature of the pHsp [6,8,10]. Monomeric and non phospho-Hsp27 inhibit actin polymerization in vitro, while phosphorylated monomers and non-phosphorylated multimers have no effect on actin polymerization [10].
Prior reports and our own observations have suggested a role for Hsp27 in axonal growth or regeneration, in addition to its role in promoting neuronal survival. Hsp27 is upregulated after injury in DRG neurons in vivo and after dissociation in vitro ([17]; Dodge and Mearow, unpublished observations). Other injury models have shown increases in Hsp27 in Schwann cells and white matter columns [18] and it has been speculated that Hsp27 might be important in the neuronal response to injury and regeneration [17,19]. Of direct relevance to a potential role of Hsp27 in axonal growth are the recent reports indicating that Hsp27 and the related Hsp22 gene deletions are responsible for familial peripheral axonopathies [20,21].
In vitro models have been widely used to study the growth behaviour of neurite initiation and extension in both CSN and peripheral neurons. In many models, neurotrophin stimulation is required for neurite growth, although in most of these models neurotrophins are also required for survival. Another widely used paradigm involves the stimulation of plated neurons with soluble laminin or extracellular matrix preparations (eg., Matrigeltm), both of which elicit neurite initiation [22-24]. This approach is particularly useful in mature DRG neurons, where not all cells will respond to a given neurotrophin (for example, NGF). Regardless of how process formation is evoked, there appear to be several general stages that can be identified including the formation of lamellopodia, filopodia, and the eventual emergence of immature neurites with growth cones [2,4]. The cellular mechanisms responsible for these behaviours are not fully elucidated.
In our cultures of adult DRG neurons we have observed robust expression and distribution of Hsp27 in dissociated DRG neurons, particularly in neuritic networks and growth cones. These observations, along with the reported role of Hsp27 in modulating the actin cytoskeleton in other cells types, led us to investigate the potential role of Hsp27 in interacting with cytoskeletal elements in different stages of neurite initiation and extension. Our hypothesis was that Hsp27 associates with the cytoskeleton in neurons and plays a key role in regulating or fine-tuning the observed ability of the cells to initiate and extend processes in response to the appropriate stimuli.
Results
Laminin induces several identifiable stages of neurite initiation and growth
In order to investigate stages of neurite initiation and subsequent growth, we employed a laminin stimulation paradigm. As these neurons are adult, they do not require any added trophic factors for their survival, and therefore neither NGF nor any other neurotrophin was required to initiate growth in these experiments. Similar stimulation experiments using laminin or matrigel have been carried out using sympathetic neurons [22].
Neurons were dissociated and plated on poly-lysine coated 16-well slides and allowed to adhere overnight (approx 18 hrs). Subsequently, the plating medium was removed and 50 μl of medium containing soluble laminin (40 μg/ml) was added to the cells. Control wells consisted of mock stimulation (eg., removal and replacement of laminin-free medium). Slides were fixed at 5, 15, 30 min and 1, 6 and 24 hrs after stimulation and subsequently processed for detection of actin, tubulin and Hsp27. Various distinctive stages in neuronal membrane expansion and neurite growth were observed and are summarized in Figure 1. One of the first steps is the appearance of a membranous expansion either around the whole soma or only from a particular portion of the cell body (A, 5 min). These lamellae are positively stained for actin (using phalloidin). Within 15–30 min, small sprouts extend from the lamellae and there are clear examples of focal contacts forming around the periphery of a lamellopodium (B, C, arrows). At later time points (1–6 hrs) some of the sprouts have elongated into filopodia and often have small growth cones associated with them (D, 1 hr; E, 6 hrs). Subsequently, neurites form and some are selected for extension by a process that is not well characterized (F, 24 hrs).
Figure 1 Laminin stimulation elicits lamellopodia and process formation in adult sensory neurons. DRG neurons plated on polylysine were stimulated with laminin in solution for 5 min, 15 min, 30 min, 1 hr, 6 hrs, 24 hrs. After fixation, neurons were stained with rhodamine-phalloidin to detect actin and images obtained using confocal microscopy. Panels A-F provide representative examples of the various stages of lamellopodia formation and eventual process protrusion, and show various distinctive stages in neuronal membrane expansion and neurite growth. At the earliest stages, lamellopodia are formed (A- 5 min, B- 15 min, C- 30 min) with evidence of focal contacts (arrows) at the leading edge of the lamellopodia (B, C). In D (1 hr) and E (6 hrs) filopodia begin to protrude from the lamellopodium around the circumference of the neuron. Eventually, these processes appear to coalesce into one or more neurites that continue to extend (F- 24 hrs, arrow). Scale bar – 20 μm.
Hsp27 colocalizes with actin and tubulin in the early stages of process initiation
Based on our hypothesis that Hsp27 may play a role in process initiation or neurite growth, we examined the localization of Hsp27 in neurons in various stages of process formation using immunocytochemistry and confocal microscopy. Here, examples of the different stages as defined in the previous section (eg., lamellopodia, focal contacts, neurite emergence) were selected from cells stimulated with laminin for 1 or 6 hrs. In addition, because of the association of Hsp27 with actin and tubulin in non-neuronal cells [25-32], we also examined whether Hsp27 would colocalize with actin and/or tubulin in neurons. Representative results are presented in Figure 2. Panels 2A and D show actin (red) in contact points (arrows) located at the periphery of an lamellopodium at one end of the neuron in A and around the circumference of the lamellum of the neuron in D. Panels 2B and 2E show the corresponding images for Hsp27 (green). The merged images (2C, F) show that Hsp27 and actin appear to be colocalized in focal contacts.
Figure 2 Hsp27 co-localizes with actin and tubulin at early stages of neurite growth. Neurons were plated on polylysine and stimulated with laminin for 1–6 hrs. Following fixation, neurons were labelled with rhodamine-phalloidin (red-A, D) or immunostained with antibodies directed against total tubulin (red – G, J) or Hsp27 (green -B, E, H, K). Images were obtained with confocal microscopy and panels C, F, I, L represent the merged images of the single channel images. Note colocalization of Hsp27 and actin in the lamellopodium (A-C, arrow) and in focal contacts observed in D-F (arrow). In panels G-I, there is some colocalization of the staining for tubulin and Hsp27 in the cortical region (large arrowhead) and in small processes emerging from the soma (arrow). In panels J-L, there is a more distinct colocalization of tubulin and Hsp27 in the cortical area (arrow, J-L) as well as in an obvious process that seems to be wrapping around the cell and finally extending (arrowhead, J-L). Scale bar – 20 μm
In panels G-L, the cells were costained with antibodies for Hsp27 (green) and total tubulin (red). The neuron in Figure 2G and 2I is beginning to show progress from the lamellar stage toward the formation of small filopodia (G-I, arrow). Tubulin staining shows some concentration in the cortical area (G, large arrowhead). There is colocalization with Hsp27 in the cortical area (I, large arrowhead) and filopodia (I, arrow), although there are areas where there is little or no overlap with Hsp27 staining (G-I, small arrowhead). The neuron shown in Figure 2J and 2L displays a pattern that was seen consistently in several different experiments, with colocalization of tubulin and Hsp27 at the cortical area (arrow) and the emergence of a more discrete process (J-L, arrowhead).
Phosphorylated Hsp27 is also localized with actin and tubulin at the early stages of process formation
Hsp27 can be phosphorylated on 2 sites of rat Hsp27 (ser15 and ser 86), and this phosphorylation is reported to be important in the role of Hsp27 in its interactions with actin [31]. Using an antibody that recognizes Hsp27 phosphorylated on the ser15 site (pHsp27 S15, ABR), we costained neurons at early stages (as defined above) of process formation for pHsp27 and actin (Fig 3A and 3F) and pHsp27 and tubulin (Fig 3G and 3L). Actin (A, D) and pHsp27 (B, E) show overlap in the lamellopodium (arrow) and in focal contacts (arrow, D-F). However, this is not complete, as noted by the exclusion of the pHsp27 from the leading edge of the lamellopodium (arrow, B, C). Tubulin (Fig 3G, J) and pHsp27 (Fig 3H, K) also colocalize in focal contacts (arrow, G-I), emerging filopodia (arrow, J-L) and in the cortical area (small arrowhead, J-L).
Figure 3 pHsp27 also co-localizes with actin and tubulin at early stages of neurite growth. Neurons plated on polylysine and stimulated with laminin for 1–6 hrs were also immunostained with antibodies directed against phosphorylated Hsp27 (pHsp27Ser15) to examine colocalization with actin or tubulin. A, D – rhodamine-phalloidin (red); B, E, H, K – pHsp27 (green); G, J – tubulin (red). The respective merged images are presented in panels C, F, I, L. Actin and pHsp27 appear to be colocalized in the body of the lamellopodium in A-C, but actin seems to be excluded from the leading edge (arrows). There is also localization of pHsp27 and actin in focal contacts (D-F, arrows). pHsp27 also colocalizes with tubulin in focal contacts (G-I, arrow), in a cortical ring (arrowhead) and in processes emerging from the cell body (J-L, arrow). Scale bar – 20 μm.
Colocalization of Hsp27 and cytoskeletal elements in neurites and growth cones at later stages of neurite growth and extension
Our initial observations indicated that the majority of adult DRG neurons in culture display robust expression of Hsp27, not only in the cell bodies but throughout the neurites when present. Hsp27 expression in sensory neuron cell bodies as well as dendritic and axonal networks has been previously reported for in vivo expression [17,19,33]. In the present study, we further examined this distribution, particularly in terms of co-expression with actin and tubulin. Rather than using the soluble laminin stimulation paradigm employed in the experiments examining early events, in these experiments we plated the neurons directly onto laminin-coated slides and then fixed the cultures 24 hrs after plating. We have previously reported that when adult DRG neurons are cultured on surfaces coated with diluted growth factor-free Matrigel, or laminin, a relatively high percentage of the neurons display significant amount of neurite outgrowth by 24 hrs after plating [34].
Figure 4A, F show representative neurons stained for actin (A, D) and pHsp27 (B) or Hsp27 (E) with the merged images displaying colocalization (C, F). The bottom panels show staining for total tubulin (G, J), pHsp27 (H) and Hsp27 (K) and the corresponding merged images in (I, L). As can be seen from the figures, Hsp27 is expressed throughout the neurons and associated neurites. As with the early stages of growth, tubulin strongly stains the cortical aspect of the cell soma as well as being present through out the processes. One interesting feature of the Hsp27 localization is the presence or local accumulation of Hsp27 and pHsp27 along with actin (but apparently not with tubulin) in branch or nodal points (arrowheads), suggesting a potential role in the pattern of neurite growth and branching. A previous publication reports beading of Hsp27 staining in dendrites of motor neurons and sensory neurons in sectioned material, although there was little discussion of the significance of this staining, other than to indicate that it was not associated with degenerating fibres [33].
Figure 4 Hsp27 continues to be expressed and localized with cytoskeletal elements in neurons and neuritic networks. In these experiments, neurons were plated on laminin (no added neurotrophins) and fixed 24 hrs after plating. As shown in the images, many neurons exhibit extensive neuritic growth under these conditions. A-F: Neurons were labelled with rhodamine-phalloidin (A, D, red), and immunostained for pHsp27 (B, green) and Hsp27 (E, green); C, F – merged images. G-L: Neurons were immunostained for tubulin (G, green; J, red), pHsp27 (H, red), Hsp27 (K, green); I, L – merged images. Hsp27 and pHsp27 are expressed throughout the neuritic network, and there is colocalization of these with actin (C, F) and less so with tubulin (I, L). Note the accumulation of pHsp27 and Hsp27 at point of branching of neurites (arrowheads- B, C, E, F, H, I, K, L). The cortical colocalization of tubulin with pHsp27 and Hsp27 is still evident at this stage of neurite growth (arrows – I, L). Scale bar – 50 μm.
We also noted that the Hsp27 and pHsp27 were strongly expressed in growth cones, further supporting an important role for Hsp27 not only in neurite initiation but also continued neurite extension. Figure 5 presents typical growth cones seen in the cultures described above. Different types of growth cones were observed with pHsp27 (A, B) and Hsp27 (C-F) being present in the core (arrowheads) of more expanded growth cones as well as in the filopodia (arrows). The growth cones in Figure 5C, F resemble the branch points noted in Figure 4, with an accumulation of an Hsp27 core and filopodia showing both Hsp27 and tubulin (merged images in D and F; tubulin, green and Hsp27, red). While the significance of this localization is not entirely clear, it is possible that one role of Hsp27 is to stabilize the cytoskeleton at these points where branching may occur (see below).
Figure 5 Co-localization of Hsp27 and actin in growth cones of growing neurites. Growth cones from neurons plated on laminin as outlined for Figure 4 were observed to express both pHsp27 and Hsp27. pHsp27 (A) and Hsp27 (C, E, red), shown together with tubulin (green-yellow) in the merged images (B, D, F), are present in growth cones and filopodia extending from the growth cones (arrows). There is also an accumulation in the core of growth cones and at points of neurite branching (arrowheads- A-F). Note that the tubulin staining does not completely overlap with pHsp27 or Hsp27, particularly in some of the extending filopodia (B, arrow) and the core of the growth cones in D, F (arrowheads). Scale bar – 10 mm.
Disruption of actin cytoskeleton with cytochalasin D results in aberrant neurite growth
Hsp27 has been suggested to play a key role in modulating actin cytoskeletal dynamics by acting as an actin-capping protein. In order to understand the role of Hsp27 in neuritic growth we decided to first examine the effects of disrupting the actin cytoskeleton integrity using cytochalasin D (CytD). Neurons were plated on laminin-coated slides and CytD was added to the medium 3 hrs post-plating (2 μM final concentration). Cultures were fixed 24 hrs later and examined for changes in neurite growth patterns and expression of Hsp27 and actin or tubulin.
Representative examples of the effects of CytD on neurons are presented in Figure 6. There was no discernible distinction between different sizes of neurons in their response to CytD; small, medium and large sized neurons displayed atypical process formation. Compared to the usual patterns of neuritic growth (Fig 4), neurons treated with CytD showed aberrant growth (Fig 6) including multiple processes emerging from the cell body (A-C), as well as stunted and disoriented neurites (D-L). In the cell displayed in Figure 6A, C, the processes show accumulation of actin (red) and pHsp27 (green) in their tips (arrowheads). Another example (D-F) shows several neurites that appear to have a disorganized internal structure resulting in the lack of the normal radial neurite extension and branching (arrowheads). In these examples, we used an antibody against actin, rather than phalloidin, in order to see total actin. In the bottom panels of Figure 6, two more examples are presented showing tubulin (green, G and J), pHsp27 (red, H) and Hsp27 (red, K) and the corresponding merged images (I, L). The cytoskeleton is more apparent in these latter examples, where the tubulin (and Hsp27) staining is clearly fibrillar in nature (arrows). Again, the disorganized and looping growth of neurites is apparent (arrowheads). In panels A-F, the actin antibody recognizes total actin, so even though CytD should disrupt the F-actin network, the antibody still detects G-actin.
Figure 6 Disruption of the actin cytoskeleton results in aberrant neurite growth. Neurons plated on LN were treated with cytochalasin D (2 mM, added 3 hrs after plating), fixed 24 hrs later and stained for pHsp27 (B, H), Hsp27 (E, K), actin (A, D) or tubulin (G, J). The respective merged images are presented in panels C, F, I and L. The cytochalasin D treatment resulted in various atypical patterns of growth. One phenotype was the elaboration of numerous processes or microspikes as seen in panels A-C, with obvious accumulation of actin and pHsp27 especially at the tips of the microspikes (arrowheads); pHsp27, but not actin, accumulates in the nucleus (B, C, arrow). Abnormal process extension was also observed. In the neuron shown in D-F, some extension was observed although there was now less colocalization of actin with the Hsp27 (arrowheads). Panels G-L show tubulin staining along with either pHsp27 or Hsp27; the nature of the cytoskeletal network is clearer in these examples (arrows). Arrowheads point to atypical neurite growth, eg., lacking the usual radial branching pattern as seen in Fig 4. Scale bar – 20 mm.
Inhibition of Hsp27 phosphorylation also results in aberrant neurite growth
Because of the reported role of Hsp27 phosphorylation in modulating the actin cytoskeleton, we wished to determine the effects of inhibiting p38 MAPK. p38 MAPK activity leads to the phosphorylation and activation of MAPKAP-K2, which acts as an Hsp27 kinase [35,36]. Inhibition of p38MAPK activity has been used to block phosphorylation of Hsp27 in the absence of direct inhibitors of MAPKAP-K2. We have thus used a combination of 2 commercially available p38 MAPK inhibitors (SB, SB203580 and SB202190, 10 μM) to investigate the potential contribution of phosphorylated Hsp27 to neurite growth.
We initially determined whether the inhibitors were effective in preventing Hsp27 phosphorylation. Using larger scale cultures, neurons were plated on LN-coated 12-well plates and after 3 hrs the inhibitors were added; 24 hrs after SB addition, cell lysates were prepared as described in the Methods. For these experiments, we used a commercially available protocol to fractionate the cells into cytosolic, membrane, nuclear and cytoskeleton fractions. Following electrophoresis, the resulting blots were probed with pHsp27 S15 and total Hsp27 antibodies. The results of a representative experiment presented in Figure 7 show that inhibitors do indeed attenuate the phosphorylation of Hsp27. The blots also show that Hsp27 is found in the cytosolic, membrane and cytoskeletal fractions, while the pHsp27 is associated primarily with the soluble fraction.
Figure 7 p38 MAPK inhibition blocks phosphorylation of Hsp27. Neurons plated on laminin were exposed to p38 MAPK inhibitors, SB203580 and SB202190 (10 mM each). Cells were sampled at 24 hrs post SB addition, using cellular subfractionation (as described in the Methods). The resulting protein from cytosol, membrane, nucleus and cytoskeleton fractions was electrophoresed and the blot subsequently probed for pHsp27 and Hsp27. Inhibition of p38 MAPK activity (laminin+SB) results in attenuation of the Hsp27 phosphorylation.
Having determined that the inhibitors had the expected effects on pHsp27, we then plated the neurons on laminin-coated slides as for the previous experiments, and treated the cultures with SB 3 hrs after plating, fixed the cells 24 hrs later and carried out immunostaining for pHsp27, Hsp27, actin and tubulin as before.
Neurons treated with SB displayed clearly atypical neurite growth. The examples presented are representative of the various patterns of neurite growth observed. As with the CytD treatment, there was no discernible distinction between different sizes of neurons in their response to SB; small, medium and large sized neurons displayed aberrant process formation. In the neuron shown in Figure 7A, C, the neurites emerged from the cell body but wrapped around the soma (arrowheads) and appeared unable to undergo appropriate extension. Another common observation was the appearance of relatively short but flattened and expanded processes and growth cones. The example in Figure 8D, F is stained for tubulin (D, green) and Hsp27 (E, red), with the merged image (F) showing the disorganized nature of the cytoskeletal elements (arrows). In this example, note that tubulin does not have complete overlap with Hsp27 staining, particularly at the tips of the growth cones (F, arrowheads).
Figure 8 Aberrant neurite growth following inhibition of Hsp27 phosphorylation. Neurons plated on laminin were treated with p38 MAPK inhibitors (SB203580 and SB202190, 10 μM each, added 3 hrs after plating) and fixed 24 hrs later. Representative results are presented. Some neurons showed abortive extension, with neurites wrapping around the cell body, such as the example in panels A-C (arrowheads, A, tubulin, B, Hsp27, C, merged image). In another example, numerous processes were observed, but these terminated in large, flattened and splayed growth cones, as shown in panel D-F (D, tubulin, E, Hsp27, F, merged image). The fibrillar nature of the Hsp27 (E, arrows) and tubulin (D, arrows) is evident and the sites of colocalization with tubulin are also apparent (F, arrows). Also note that there is not a complete overlap of Hsp27 and tubulin at the tips of the growth cones (F, arrowheads). Scale bar – 20 μm.
In addition, some neurons displayed extensive neurite growth, although this was again generally characterized by flattened and expanded processes and growth cones. Figure 9 presents such an example. This neuron has at least 7–8 processes extending from the cell body, all of which show process expansion. In panels A-C, Hsp27 (red) can clearly be observed colocalized with tubulin (green) in the processes emerging from the cell body (arrows). Areas of the fibrillar nature and overlap of Hsp27 and tubulin are also noted (arrowheads). In Figure 9D, F, larger magnification of the area generally noted by the arrowheads in A-C is shown. Here the splaying of the growth cones (arrows) and loss of cytoskeletal bundling is more apparent (arrowheads; compare processes observed in Figure 4 or 5 with those in Figure 9).
Figure 9 Flattened growth cones and processes show co-localization of tubulin and Hsp27. This figure shows another example of a neuron treated with the p38 MAPK inhibitors as outlined in Figure 8. Colocalization of Hsp27 (B, red) with tubulin (A, green) is apparent in the emerging processes (C, arrows), and in the flattened and splayed growth cones (A-C, arrowheads); scale bar – 50 μm. At a higher magnification (D-F), loss of microtubule bundling is observed (arrowheads) along with the fibrillar nature of Hsp27 and colocalization with tubulin (arrows); scale bar – 20 μm.
These results suggest that attenuation of the phosphorylation of Hsp27 can have adverse effects on the neuritic cytoskeleton, similar to those observed with Cyt D. Although our assumption (based on previous reports in the literature) is that the SB compounds block p38 MAPK activity, its downstream effects on MAPKAP-K2 and the subsequent inhibition of Hsp27 phosphorylation, it is possible that these compounds may have other inhibitory influences, or that they may be influencing the cytoskeletal elements through actions not involving Hsp27. While our data show that the SB compounds do inhibit phosphorylation of Hsp27, we cannot completely rule out effects on other signalling components, although at the concentrations we have used, the effects are reported to be specific for p38 MAPK inhibition, rather than any other additional kinases.
Discussion
We describe early events in adult DRG neuron process formation in response to stimulation with the extracellular matrix protein laminin. Our data show that Hsp27 appears to associate with actin and tubulin in structures found at all stages of neurite initiation. Lamellopodia, filopodia, microspikes and focal contacts all displayed a colocalization of Hsp27 and actin or tubulin. The filamentous nature of the Hsp27 was quite clear in neurites and growth cones supporting the hypothesis that Hsp27 is associating with cytoskeletal elements.
Our results are similar to those described previously for neurite growth initiation and process extension in embryonic cultured CNS neurons. Culture studies of early neuritogenesis events in hippocampal neurons have provided information that demonstrates that events after initial cellular attachment to the substrate are quite similar among different cell types and indeed events in neurons are very similar to those in migratory fibroblasts [2,4,37]. The cells attach and are surrounded by a thin lamellopodium from which small extensions sprout. These extensions often have growth cones and display dynamic back and forth movements. At some point, one or more of these processes elongates, while the others remain stationary or retract. All the stages described by DaSilva and Dotti [2] and Dehmelt and Halpain [4] could be identified in our cultures of adult DRG neurons, suggesting that this process is intrinsic to all neurons.
Neurite protrusion requires the actin cytoskeleton, with lamellopodia being filled with an actin meshwork necessary for the appropriate adhesion and filopodia having actin bundles with the rapidly growing ends oriented towards the tips. Studies have shown that actin polymerizes at the leading edge of the lamellopodia, and then disassembles and recedes from the peripheral area [4,38]. This phenomenon influences growth cone advance and could likely play a role in neurite initiation as well. Microtubules may play a mechanical role in this since they invade the actin cytoskeleton in lamellopodia of various cell types. [2,4,38].
In cell-free assay systems, Hsp27 can act as an actin-capping protein which prevents the polymerization of actin and the assembly of F-actin [10,11]. Phosphorylation of Hsp27 leads to the loss of its ability to inhibit actin polymerization, and thus increases the rate and extent of actin polymerization and the formation of F-actin [6,8,10,11,39,40]. In addition to modulating the actin cytoskeleton, Hsp27 has also been reported to interact with both neurofilaments and microtubules in a phosphorylation-dependent manner [41,42]. Hsp27 has been inferred to stabilize not only actin, but also neurofilament and microtubules [31].
Phosphorylation of Hsp27 promotes the polymerization of actin and stress fibre formation [6,10,43]. Hsp27 is phosphorylated on 3 serines in the human Hsp27 (S15, S78, S82) and 2 in the rodent Hsp27 (S15 and S86 in mouse or S90 in hamster Hsp27). Hsp27 in unstressed cells exists as large oligomers, while upon phosphorylation Hsp27 dissociates in smaller species, including dimers and monomers [44,45]. In cell free assays, the unphosphorylated monomers of Hsp27 blocked actin polymerization, while the unphosphorylated oligomers and the phosphorylated monomeric form were ineffective. While the evidence based primarily on structural studies supports a role for phosphorylation of Hsp27 in stabilization of the actin cytoskeleton, a recent study has provided direct functional evidence that this is indeed the case [46].
Hsp27 phosphorylation is regulated by activity of the p38MAPK pathway, whereby p38 MAPK activation of MAP-kinase-activated protein-kinase 2/3 (MAPKAP-K2) leads to the phosphorylation of Hsp27 [36,44]. However, PKCδ and cGMP-dependent kinase have also been reported to phosphorylate Hsp27 in smooth muscle [47,48]. While the classical stress-activated signalling pathway activation of p38 MAPK regulates Hsp27 after heat-shock and other stresses, it is more likely that activation of p38 MAPK is downstream of the Cdc-42 and Rac activation of Pak1 with respect to neurite initiation and growth. For example, laminin can lead to p38 MAPK activation and Hsp27 phosphorylation, as previously reported for Schwann cells [49], and this is likely via Cdc-42 and Rac activation downstream of integrin-dependent signalling cascades [31]. Given the role of Rac and Rho in regulating actin dynamics in growth cones and the observations that inhibition of Rho promotes axonal growth on inhibitory substrates [50,51], the interactions of Hsp27 with Rho observed in smooth muscle cells [52-54] suggests an intriguing interplay among these components. Whether a similar interaction occurs in neurons or axons is not known.
Treatment of neurons with agents that disrupt the actin cytoskeleton result in aberrant neurite initiation and growth. Neurons treated with Cyt D, which caps existing actin filaments at barbed ends, consistently show rapid emergence of numerous neurites that elongate in a disoriented fashion [2,4,37]. Treatment of DRG neurons in our cultures resulted in similar aberrant growth, particularly at the early stages examined where numerous projections emerged from the neurons as early as 1 hr after laminin stimulation (data not shown); neurons examined at 24 hrs showed disoriented growth of processes (eg., Fig 6).
Since Cyt D acts to cap barbed actin filaments and non-phophorylated Hsp27 has been suggested to do the same, we reasoned that if pHsp27 was important for normal neurite initiation and extension, if we inhibited the phosphorylation of Hsp27 we might observe similar effects on neurite initiation. As shown in our results, attenuation of Hsp27 phosphorylation using the p38MAPK inhibitors, does indeed result in atypical growth patterns. At the early stages, results were similar to what we had observed with Cyt D (data not shown), and at later stages, neurite growth was again quite clearly aberrant (Fig 8, 9). Some neurons showed neurites that tended to wrap around the cell soma or extend in a disoriented fashion. Another consistent characteristic of the relatively short processes that did extend was the flattened and splayed nature of the neurites and growth cones. There appeared to be a lack of the appropriate actin and microtubular bundling that would result in normal neurite extension and growth cone dynamics (eg., compare Fig 4, 5 with Fig 6 and 8) [4,38]. We have inferred that effects of p38 MAPK inhibition on neurite growth were due to the inhibition of Hsp27 phosphorylation. A similar inhibition of neurite initiation by SB has been reported in PC12 cells [55]; interestingly, in this study induction of Hsp27 by heat shock promoted neuritogenesis. However, there may be effects on other cytoskeletal elements. Ackerley et al [56] have reported that p38 MAPK also phosphorylates neurofilaments in transfected COS cells, although they did not find any effect of p38 MAPK inhibition on neurofilament phosphorylation in cortical neurons.
There are relatively few reports of the interaction of Hsp27 with cytoskeletal elements other than actin. Hsp27 has been reported to associate with microtubules in HeLa cells [57] and in CHO cells [58]. In the latter report, overexpression of Hsp27 was shown to protect microtubules from heat shock and pH-induced collapse, although the contribution of pHsp27 to this effect was not reported [58].
pHsp27 also appears to be required for the migration of several cell types [29,59-61]. A recent study concluded that p38MAPK activation and Hsp27 phosphorylation played a key role in the regulation of actin polymerization, possibly by regulating the spatial organization of the lamellopodia by promoting branch formation at the leading edge and stability at the base [29]. They suggest that at the dynamic leading edge of lamellopodia, Hsp27 might promote branching by its actin-capping activity, while at the base p38MAPK remains active and Hsp27 is phosphorylated and might stabilize actin filaments.
Mutations of the small Hsp (Hsp22 and Hsp25/27) genes have been linked to axonal Charcot-Marie-Tooth disease and distal hereditary motor neuropathy (DHMN) [20,21]. This appears to be related to the disruption of the neurofilament networks by the aggregation of neurofilament proteins and collapse of neurofilament networks [20]. This study and recent commentaries [42,62] point to the importance of the small heat shock proteins like Hsp27 in regulating or modulating the function of cytoskeletal elements other than actin. However, the mechanisms underlying the function of Hsp27 and its regulation remain essentially unknown in neuronal cells.
Our results suggest that Hsp27 is necessary for the initiation of neurite outgrowth in DRG neurons. The data also suggest that phosphorylation of Hsp27 plays a key role in modulating the dynamic interactions of Hsp27 with cytoskeletal elements such as actin and tubulin to regulate the response of DRG neurons to environmental cues that mediate growth.
Conclusion
Using immunocytochemistry, we observed colocalization of the phosphorylated and non-phosphorylated forms of Hsp27 with actin and tubulin in both very early and later stages of neurite growth from cultured adult DRG neurons. The colocalization of Hsp27 and pHsp27 with actin in lamellopodia and focal contacts at early neurite initiation stages, and in processes, branch points and growth cones in later stages suggests that Hsp27 may play a role in neurite initiation and extension and potentially in the patterning of this growth. While the mechanisms of action require further investigation, it is possible that one role of Hsp27 is to stabilize the cytoskeleton at potential sites of branching or sprouting. Hsp27 has been reported to play a key role in modulating actin cytoskeletal dynamics as an actin-capping protein in non-neuronal cells and our results suggest that this may also be the case in neurons. Neurons treated with cytochalasin D showed aberrant neurite growth patterns. Neurons treated with p38 MAPK inhibitors, which inhibit the downstream phosphorylation of Hsp27, also displayed either lack of neurite growth or failure of appropriate neurite extension. The similar results from the CytD and inhibition of Hsp27 phosphorylation support a role for Hsp27 in neurite outgrowth via its phosphorylation state-dependent interactions with actin.
Methods
Neuronal cultures
Dorsal root ganglia (DRG) from young adult (5–6 wk) Sprague-Dawley rats (Memorial University of Newfoundland Vivarium and Charles River Canada, Montreal, QC) were dissected and dissociated using modifications to techniques described previously [34,63]. Briefly, ganglia from all spinal levels were removed and the roots trimmed, and subsequently incubated in 0.25% collagenase for 45 min, followed by 0.25% trypsin for 20 min (Invitrogen/ Gibco BRL, Burlington, Ont). Dissociated neurons were suspended in serum-free Neurobasal medium (NB, Invitrogen) supplemented with 100 U penicillin/streptomycin, B27 supplement (Invitrogen), and 20 μM cytosine arabinoside (modified NB). This suspension was then layered on top of a 28% Percoll solution (Amersham Bioscience, Baie d'Urfe, QC) in 15 ml conical tubes, centrifuged at 400 g for 20 min at room temperature. Pellets were then carefully extracted with a sterile pasture pipette, placed in a fresh tube, washed with the previous suspension media and centrifuged to remove any remaining Percoll. Neurons were plated in Lab-Tek 16-well chamber slides (Nunc International, Naperville, NC) for neurite growth assessment or 12-well plates for Western blotting and incubated at 37°C, 95% O2 and 5% CO2. Slides and culture plates were coated with poly-lysine (PL, 1 μg/ml, BD Bioscience, Bedford, MA) or laminin, (LM, 20–40 μg/ml, Invitrogen) where appropriate. The neurons were cultured in modified serum-free NB alone with no added growth factors.
Immunocytochemistry
Neurons were fixed in 4% paraformaldehyde (pH 7–7.4) in PBS for 20 minutes, permeabilized with 0.1% Triton-X-100 and blocked with 5% normal goat serum in PBS. Antibodies used were as follows: Hsp27 (SPA-801, Stressgen Corp, Victoria, BC) and phospho-Hsp27S15 (PAI-018, Affinity BioReagents, Golden, CO), total tubulin (Sigma-Aldrich, St. Louis, MO), actin (Sigma-Aldrich). It should be noted that the Hsp27 antibody recognizes both the non-phosphorylated and phosphorylated Hsp27, while the pHsp27 antibody only recognizes the phosphorylated form. We have also tested two other pHsp27 antibodies (UBI and Santa Cruz, see [63]), but have found the Affinity Bioreagents Antibody to be better for immunostaining.
Cells were incubated with the primary antibodies at 4°C for 16–20 hrs, followed by Cy2 or Cy5-tagged secondary antibodies (Jackson Immunoresearch Labs, West Grove, PA). In some experiments, cells were also labelled with rhodamine-phalloidin after antibody incubation (Sigma-Aldrich). The slides were coverslipped with glycerol and imaged with confocal laser scanning microscopy using z-stage scanning and image stacking. Stacked digital images were imported into Adobe Photoshop for compilation into the final composite figures.
Laminin stimulation and neurite growth initiation
Neurite initiation was assessed in two ways. The first series of experiments employed neurons plated on laminin-coated slides, with the cells being fixed and analyzed for outgrowth parameters (lamellopodia, filopodia and neurite initiation and extension) at 24 hrs after plating. In a second series of experiments, the neurons were first plated on polylysine coated slides and allowed to stabilize overnight prior to being stimulated with soluble laminin (20 μg/ml in basal medium). Following the addition of the laminin solution, cells were then fixed at 5, 15, 30 min, 1 hr, 6 hr, and 24 hr. After fixation, cells were immunostained and analyzed as described above.
Inhibitor experiments
SB 203580 and SB 202190 (10 μM, Calbiochem/EMD Biosciences, San Diego, CA) were used to inhibit p38 MAPK activity, in order to assess the contribution of phosphorylated Hsp27. Inhibitors were added 1 hr prior to laminin stimulation. For the 24 hr cultures, the inhibitors were added 2–3 hrs after plating the cells on laminin-coated slides and retained in the medium for the extent of the experiment (usually 24 hrs). Cytochalasin D (2 μM, Sigma) was also used in longer term experiments, and was added 3 hrs after plating and maintained in the medium for the extent of the experiment (24 hrs)
Immunoblotting
For Western analyses, neurons were plated in 12-well plates that had been coated with polylysine alone or with laminin, depending on which experimental paradigm was used (see above). Neurons were subsequently processed according to our established procedures [34,63]. Cellular fractionation was carried out using a subcellular protein extraction kit (ProteoExtract, Calbiochem/EMD Biosciences, San Diego, CA) to isolate cytoplasmic, membrane, nuclear and cytoskeletal fractions. This protocol involves sequential isolation of these fractions using specific buffer systems (proprietary, as supplied by the manufacturer) to lyse cells in situ in the tissue culture plates. Subsequently, protein concentrations were determined for the fractions using the BCA protein assay (Pierce Chemicals, Rockford, IL.). Equivalent amounts of protein (40 μg protein) were loaded in each lane. Following transfer to nitrocellulose, the blots were first stained with Ponceau Red to assess protein loading, and subsequently probed with the following antibodies: phospho-Hsp27S15 (PAI-016, Affinity Bioreagents) and Hsp27 (SPA-801, Stressgen),. Blots were cut and reprobed sequentially, and visualized with ECL reagents (NEN, Boston, MA) and exposure to X-ray film (Cronex MRF Clear base, Agfa Corp, Greenville, SC). Developed films were subsequently digitized and densitometrically analyzed with a cyclone ChemiImager and AlphaEase software. Digital images of the blots were used to make composite figures with Adobe Photoshop graphics software (Adobe Corp, MountainView CA).
Authors' contributions
KLW carried out neurite growth and confocal analyses, as well as the Western blotting experiments. MR provided technical assistance with aspects of the work, most notably preparing neuronal cultures. KMM wrote the manuscript and was responsible for the experimental concept, design and overall supervision of the experiments, as well as carrying out some of the confocal analyses.
Acknowledgements
Funding for this work was through an NSERC Discovery grant (KMM). The authors thank Dr. Daniel MacPhee, Elaine Dodge, Sherri Rankin and Budd Tucker for their invaluable assistance and discussion.
==== Refs
Tucker BA Rahimtula M Mearow KM Integrin activation and neurotrophin signaling cooperate to enhance neurite outgrowth in sensory neurons J Comp Neurol 2005 486
da Silva JS Dotti CG Breaking the neuronal sphere: regulation of the actin cytoskeleton in neuritogenesis Nat Rev Neurosci 2002 3 694 704 12209118 10.1038/nrn918
Dehmelt L Smart FM Ozer RS Halpain S The role of microtubule-associated protein 2c in the reorganization of microtubules and lamellipodia during neurite initiation J Neurosci 2003 23 9479 9490 14573527
Dehmelt L Halpain S Actin and microtubules in neurite initiation: are MAPs the missing link? J Neurobiol 2004 58 18 33 14598367 10.1002/neu.10284
Charette SJ Landry J The interaction of HSP27 with Daxx identifies a potential regulatory role of HSP27 in Fas-induced apoptosis Ann N Y Acad Sci 2000 926 126 131 11193028
Guay J Lambert H Gingras-Breton G Lavoie JN Huot J Landry J Regulation of actin filament dynamics by p38 map kinase-mediated phosphorylation of heat shock protein 27 J Cell Sci 1997 110 ( Pt 3) 357 368 9057088
Huot J Houle F Spitz DR Landry J HSP27 phosphorylation-mediated resistance against actin fragmentation and cell death induced by oxidative stress Cancer Res 1996 56 273 279 8542580
Lavoie JN Lambert H Hickey E Weber LA Landry J Modulation of cellular thermoresistance and actin filament stability accompanies phosphorylation-induced changes in the oligomeric structure of heat shock protein 27 Mol Cell Biol 1995 15 505 516 7799959
Lavoie JN Gingras-Breton G Tanguay RM Landry J Induction of Chinese hamster HSP27 gene expression in mouse cells confers resistance to heat shock. HSP27 stabilization of the microfilament organization J Biol Chem 1993 268 3420 3429 8429018
Benndorf R Hayess K Ryazantsev S Wieske M Behlke J Lutsch G Phosphorylation and supramolecular organization of murine small heat shock protein HSP25 abolish its actin polymerization-inhibiting activity JBiolChem 1994 269 20780 20784
Miron T Vancompernolle K Vandekerckhove J Wilchek M Geiger B A 25-kD inhibitor of actin polymerization is a low molecular mass heat shock protein J Cell Biol 1991 114 255 261 2071672 10.1083/jcb.114.2.255
Gerthoffer WT Gunst SJ Invited review: focal adhesion and small heat shock proteins in the regulation of actin remodeling and contractility in smooth muscle J Appl Physiol 2001 91 963 972 11457815
Landry J Lambert H Zhou M Lavoie JN Hickey E Weber LA Anderson CW Human HSP27 is phosphorylated at serines 78 and 82 by heat shock and mitogen-activated kinases that recognize the same amino acid motif as S6 kinase II J Biol Chem 1992 267 794 803 1730670
Mehlen P Arrigo AP The serum-induced phosphorylation of mammalian hsp27 correlates with changes in its intracellular localization and levels of oligomerization Eur J Biochem 1994 221 327 334 8168520 10.1111/j.1432-1033.1994.tb18744.x
Meier M King GL Clermont A Perez A Hayashi M Feener EP Angiotensin AT(1) receptor stimulates heat shock protein 27 phosphorylation in vitro and in vivo Hypertension 2001 38 1260 1265 11751700
Bitar KN HSP27 phosphorylation and interaction with actin-myosin in smooth muscle contraction Am J Physiol Gastrointest Liver Physiol 2002 282 G894 903 11960785
Costigan M Mannion RJ Kendall G Lewis SE Campagna JA Coggeshall RE Meridith-Middleton J Tate S Woolf CJ Heat shock protein 27: developmental regulation and expression after peripheral nerve injury J Neurosci 1998 18 5891 5900 9671676
Murashov AK Ul Haq I Hill C Park E Smith M Wang X Goldberg DJ Wolgemuth DJ Crosstalk between p38, Hsp25 and Akt in spinal motor neurons after sciatic nerve injury Brain Res Mol Brain Res 2001 93 199 208 11589997 10.1016/S0169-328X(01)00212-1
Benn SC Perrelet D Kato AC Scholz J Decosterd I Mannion RJ Bakowska JC Woolf CJ Hsp27 upregulation and phosphorylation is required for injured sensory and motor neuron survival Neuron 2002 36 45 56 12367505 10.1016/S0896-6273(02)00941-8
Evgrafov OV Mersiyanova I Irobi J Van Den Bosch L Dierick I Leung CL Schagina O Verpoorten N Van Impe K Fedotov V Dadali E Auer-Grumbach M Windpassinger C Wagner K Mitrovic Z Hilton-Jones D Talbot K Martin JJ Vasserman N Tverskaya S Polyakov A Liem RK Gettemans J Robberecht W De Jonghe P Timmerman V Mutant small heat-shock protein 27 causes axonal Charcot-Marie-Tooth disease and distal hereditary motor neuropathy Nat Genet 2004 36 602 606 15122254 10.1038/ng1354
Irobi J Van Impe K Seeman P Jordanova A Dierick I Verpoorten N Michalik A De Vriendt E Jacobs A Van Gerwen V Vennekens K Mazanec R Tournev I Hilton-Jones D Talbot K Kremensky I Van Den Bosch L Robberecht W Van Vandekerckhove J Broeckhoven C Gettemans J De Jonghe P Timmerman V Hot-spot residue in small heat-shock protein 22 causes distal motor neuropathy Nat Genet 2004 36 597 601 15122253 10.1038/ng1328
Slaughter T Wang J Black MM Microtubule transport from the cell body into the axons of growing neurons J Neurosci 1997 17 5807 5819 9221779
Tang D Goldberg DJ Bundling of microtubules in the growth cone induced by laminin Mol Cell Neurosci 2000 15 303 313 10736206 10.1006/mcne.1999.0820
Yu W Ling C Baas PW Microtubule reconfiguration during axogenesis J Neurocytol 2001 30 861 875 12373094 10.1023/A:1020622530831
Tessier DJ Komalavilas P Panitch A Joshi L Brophy CM The small heat shock protein (HSP) 20 is dynamically associated with the actin cross-linking protein actinin J Surg Res 2003 111 152 157 12842460 10.1016/S0022-4804(03)00113-6
Tomasovic SP Simonette RA Wolf DA Kelley KL Updyke TV Co-isolation of heat stress and cytoskeletal proteins with plasma membrane proteins Int J Hyperthermia 1989 5 173 190 2926184
Liang P MacRae TH Molecular chaperones and the cytoskeleton J Cell Sci 1997 110 ( Pt 13) 1431 1440 9224761
Panasenko OO Kim MV Marston SB Gusev NB Interaction of the small heat shock protein with molecular mass 25 kDa (hsp25) with actin Eur J Biochem 2003 270 892 901 12603322 10.1046/j.1432-1033.2003.03449.x
Pichon S Bryckaert M Berrou E Control of actin dynamics by p38 MAP kinase - Hsp27 distribution in the lamellipodium of smooth muscle cells J Cell Sci 2004 117 2569 2577 15128872 10.1242/jcs.01110
Kindas-Mugge I Rieder C Frohlich I Micksche M Trautinger F Riedler C Characterization of proteins associated with heat shock protein hsp27 in the squamous cell carcinoma cell line A431 Cell Biol Int 2002 26 109 116 11779227 10.1006/cbir.2001.0822
Mounier N Arrigo AP Actin cytoskeleton and small heat shock proteins: how do they interact? Cell Stress Chaperones 2002 7 167 176 12380684 10.1379/1466-1268(2002)007<0167:ACASHS>2.0.CO;2
Aquino DA Padin C Perez JM Peng D Lyman WD Chiu FC Analysis of glial fibrillary acidic protein, neurofilament protein, actin and heat shock proteins in human fetal brain during the second trimester Brain Res Dev Brain Res 1996 91 1 10 8821474 10.1016/0165-3806(95)00146-8
Plumier JC Hopkins DA Robertson HA Currie RW Constitutive expression of the 27-kDa heat shock protein (Hsp27) in sensory and motor neurons of the rat nervous system J Comp Neurol 1997 384 409 428 9254036 10.1002/(SICI)1096-9861(19970804)384:3<409::AID-CNE7>3.0.CO;2-7
Jones DM Tucker BA Rahimtula M Mearow KM The synergistic effects of NGF and IGF-1 on neurite growth in adult sensory neurons: convergence on the PI 3-kinase signaling pathway J Neurochem 2003 86 1116 1128 12911620
Huot J Lambert H Lavoie JN Guimond A Houle F Landry J Characterization of 45-kDa/54-kDa HSP27 kinase, a stress-sensitive kinase which may activate the phosphorylation-dependent protective function of mammalian 27-kDa heat-shock protein HSP27 Eur J Biochem 1995 227 416 427 7851416 10.1111/j.1432-1033.1995.tb20404.x
Larsen JK Yamboliev IA Weber LA Gerthoffer WT Phosphorylation of the 27-kDa heat shock protein via p38 MAP kinase and MAPKAP kinase in smooth muscle Am J Physiol 1997 273 L930 40 9374719
Bradke F Dotti CG The role of local actin instability in axon formation Science 1999 283 1931 1934 10082468 10.1126/science.283.5409.1931
Gallo G Letourneau PC Regulation of growth cone actin filaments by guidance cues J Neurobiol 2004 58 92 102 14598373 10.1002/neu.10282
Landry J Huot J Modulation of actin dynamics during stress and physiological stimulation by a signaling pathway involving p38 MAP kinase and heat-shock protein 27 Biochem Cell Biol 1995 73 703 707 8714691
Landry J Huot J Regulation of actin dynamics by stress-activated protein kinase 2 (SAPK2)-dependent phosphorylation of heat-shock protein of 27 kDa (Hsp27) BiochemSocSymp 1999 64 79 89
Perng MD Cairns L van den IJP Prescott A Hutcheson AM Quinlan RA Intermediate filament interactions can be altered by HSP27 and alphaB-crystallin J Cell Sci 1999 112 ( Pt 13) 2099 2112 10362540
Der Perng M Quinlan RA Neuronal diseases: small heat shock proteins calm your nerves Curr Biol 2004 14 R625 6 15296780 10.1016/j.cub.2004.07.047
Huot J Houle F Rousseau S Deschesnes RG Shah GM Landry J SAPK2/p38-dependent F-actin reorganization regulates early membrane blebbing during stress-induced apoptosis J Cell Biol 1998 143 1361 1373 9832563 10.1083/jcb.143.5.1361
Rouse J Cohen P Trigon S Morange M Alonso-Llamazares A Zamanillo D Hunt T Nebreda AR A novel kinase cascade triggered by stress and heat shock that stimulates MAPKAP kinase-2 and phosphorylation of the small heat shock proteins Cell 1994 78 1027 1037 7923353 10.1016/0092-8674(94)90277-1
Lambert H Charette SJ Bernier AF Guimond A Landry J HSP27 multimerization mediated by phosphorylation-sensitive intermolecular interactions at the amino terminus J Biol Chem 1999 274 9378 9385 10092617 10.1074/jbc.274.14.9378
An SS Fabry B Mellema M Bursac P Gerthoffer WT Kayyali US Gaestel M Shore SA Fredberg JJ Role of heat shock protein 27 in cytoskeletal remodeling of the airway smooth muscle cell J Appl Physiol 2004 96 1701 1713 14729728 10.1152/japplphysiol.01129.2003
Butt E Immler D Meyer HE Kotlyarov A Laass K Gaestel M Heat shock protein 27 is a substrate of cGMP-dependent protein kinase in intact human platelets: phosphorylation-induced actin polymerization caused by HSP27 mutants J Biol Chem 2001 276 7108 7113 11383510 10.1074/jbc.M009234200
Maizels ET Peters CA Kline M Cutler REJ Shanmugam M Hunzicker-Dunn M Heat-shock protein-25/27 phosphorylation by the delta isoform of protein kinase C Biochem J 1998 332 ( Pt 3) 703 712 9620873
Fragoso G Robertson J Athlan E Tam E Almazan G Mushynski WE Inhibition of p38 mitogen-activated protein kinase interferes with cell shape changes and gene expression associated with Schwann cell myelination Exp Neurol 2003 183 34 46 12957486 10.1016/S0014-4886(03)00101-8
Lehmann M Fournier A Selles-Navarro I Dergham P Sebok A Leclerc N Tigyi G McKerracher L Inactivation of Rho signaling pathway promotes CNS axon regeneration J Neuro Sci 1999 19 7537 7547
Ellezam B Dubreuil C Winton M Loy L Dergham P Selles-Navarro I McKerracher L Inactivation of intracellular Rho to stimulate axon growth and regeneration Prog Brain Res 2002 137 371 380 12440379
Patil SB Tsunoda Y Pawar MD Bitar KN Translocation and association of ROCK-II with RhoA and HSP27 during contraction of rabbit colon smooth muscle cells Biochem Biophys Res Commun 2004 319 95 102 15158447 10.1016/j.bbrc.2004.04.159
Patil SB Pawar MD Bitar KN Phosphorylated HSP27 essential for acetylcholine-induced association of RhoA with PKCalpha Am J Physiol Gastrointest Liver Physiol 2004 286 G635 44 14592945 10.1152/ajpgi.00261.2003
Wang P Bitar KN Rho A regulates sustained smooth muscle contraction through cytoskeletal reorganization of HSP27 Am J Physiol 1998 275 G1454 62 9843784
Kano Y Nakagiri S Nohno T Hiragami F Kawamura K Kadota M Numata K Koike Y Furuta T Heat shock induces neurite outgrowth in PC12m3 cells via the p38 mitogen-activated protein kinase pathway Brain Res 2004 1026 302 306 15488493 10.1016/j.brainres.2004.07.053
Ackerley S Grierson AJ Banner S Perkinton MS Brownlees J Byers HL Ward M Thornhill P Hussain K Waby JS Anderton BH Cooper JD Dingwall C Leigh PN Shaw CE Miller CC p38alpha stress-activated protein kinase phosphorylates neurofilaments and is associated with neurofilament pathology in amyotrophic lateral sclerosis Mol Cell Neurosci 2004 26 354 364 15207859 10.1016/j.mcn.2004.02.009
Hino M Kurogi K Okubo MA Murata-Hori M Hosoya H Small heat shock protein 27 (HSP27) associates with tubulin/microtubules in HeLa cells Biochem Biophys Res Commun 2000 271 164 169 10777697 10.1006/bbrc.2000.2553
Hargis MT Storck CW Wickstrom E Yakubov LA Leeper DB Coss RA Hsp27 anti-sense oligonucleotides sensitize the microtubular cytoskeleton of Chinese hamster ovary cells grown at low pH to 42 degrees C-induced reorganization Int J Hyperthermia 2004 20 491 502 15277022 10.1080/02656730410001699100
Piotrowicz RS Hickey E Levin EG Heat shock protein 27 kDa expression and phosphorylation regulates endothelial cell migration FASEB J 1998 12 1481 1490 9806757
Rousseau S Houle F Landry J Huot J p38 MAP kinase activation by vascular endothelial growth factor mediates actin reorganization and cell migration in human endothelial cells Oncogene 1997 15 2169 2177 9393975 10.1038/sj.onc.1201380
Hedges JC Dechert MA Yamboliev IA Martin JL Hickey E Weber LA Gerthoffer WT A role for p38(MAPK)/HSP27 pathway in smooth muscle cell migration J Biol Chem 1999 274 24211 24219 10446196 10.1074/jbc.274.34.24211
Lee S Leavitt B Turning up the heat on hereditary neuropathies Clin Genet 2004 66 294 296 10.1111/j.1399-0004.2004.00347b.x
Mearow KM Dodge ME Rahimtula M Yegappan C Stress-mediated signaling in PC12 cells - the role of the small heat shock protein, Hsp27, and Akt in protecting cells from heat stress and nerve growth factor withdrawal J Neurochem 2002 83 452 462 12423255 10.1046/j.1471-4159.2002.01151.x
| 15819993 | PMC1087488 | CC BY | 2021-01-04 16:39:09 | no | BMC Neurosci. 2005 Apr 8; 6:24 | utf-8 | BMC Neurosci | 2,005 | 10.1186/1471-2202-6-24 | oa_comm |
==== Front
BMC NeurosciBMC Neuroscience1471-2202BioMed Central London 1471-2202-6-261582630610.1186/1471-2202-6-26Research ArticleProliferation dynamics of germinative zone cells in the intact and excitotoxically lesioned postnatal rat brain Faiz Maryam [email protected] Laia [email protected] Bernardo [email protected] Berta [email protected] Unit of Histology, Faculty of Medicine, Autonomous University of Barcelona, Campus UAB, 08193 Bellaterra, Spain2005 12 4 2005 6 26 26 10 12 2004 12 4 2005 Copyright © 2005 Faiz et al; licensee BioMed Central Ltd.2005Faiz 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 forebrain subventricular zone (SVZ)-olfactory bulb pathway and hippocampal subgranular zone (SGZ) generate neurons into adulthood in the mammalian brain. Neurogenesis increases after injury to the adult brain, but few studies examine the effect of injury on neural and glial precursors in the postnatal brain. To characterize the spatio-temporal dynamics of cell proliferation in the germinative zones, this study utilized a model of postnatal damage induced by NMDA injection in the right sensorimotor cortex at postnatal day 9.
Dividing cell populations were labeled with 5-Bromodeoxyuridine (BrdU) in the intact and damaged postnatal brain. Identity of proliferating cells was determined by double immunolabeling with nestin, GFAP, NeuN and tomato lectin (TL).
Results
In the control brain, grouped BrdU+ cells were observed in the Rostral Migratory Stream (RMS), SVZ and SGZ. Maximal proliferation was seen at P12, persisted until P23 and diminished by P49. After injury, a striking reduction in the number of BrdU+ cells was observed in the ipsilateral SVZ from 10 hours (58% decrease) until 14 days post-lesion (88% decrease). In contrast, an increase in grouped BrdU+ cells was seen in the striatum adjacent to the depleted SVZ. Significantly reduced numbers of BrdU+ cells were also seen in the RMS until 3 days post-lesion. No changes were noted in the SGZ. Both in controls and lesioned hemispheres, BrdU+ cells located in the germinal zones were mostly nestin positive and negative for GFAP, NeuN, and TL. In the SVZ area lining the ventricle, BrdU+/nestin+ cells were mainly located between TL+ ependyma and parenchymal GFAP+ astrocytes. After excitotoxicity, a decrease in the number and orientation of GFAP/nestin+ prolongations leaving the SVZ to the cortex, corpus callosum and striatum was noted until 5 days post-lesion.
Conclusion
Postnatal excitotoxic injury differentially affects proliferating cells in the germinative zones: no change is observed in the dentate gyrus whereas excitotoxicity causes a significant decrease in proliferating cells in the SVZ and RMS. Depletion of BrdU+ cells in the postnatal SVZ and RMS differs from previous studies after adult brain injury and may affect the SVZ-RMS migration and is suggestive of progenitor recruitment to injured areas.
==== Body
Background
Progenitor cells of the central nervous system (CNS) are a population of undifferentiated cells of neuroectodermal origin with high proliferative capacity that can differentiate into neuronal cells or macroglial cells (i.e. astrocytes and oligodendrocytes).
These progenitors for neurons and glia, which can proliferate throughout life in the CNS, are located in at least 3 germinal zones: the ventricular wall and adjacent subventricular zone (SVZ), the subgranular zone of the hippocampal dentate gyrus (SGZ) and the olfactory bulb (OB) [1-3]. The fate of progenitor cells depends highly on their location; whereas progenitor cells of the SGZ migrate into the granule cell layer of the hippocampus [3], SVZ progenitors migrate tangentially in chains surrounded by astrocytic tube structures forming the rostral migratory stream (RMS) towards the OB [1]. Furthermore, progenitor cells of a more restricted fate have been described, in vitro, to reside within the cerebral cortex and provide a local source of new neurons [4]. Thus, multiple brain areas could function as reservoirs of brain precursor cells. However, the normal CNS environment is likely to limit would-be uncommitted progenitor reservoirs to a glial fate [5-7] and the role of these progenitors both in the normal and injured CNS is not well established.
After damage to the adult brain, several studies have reported different profiles of progenitor cell proliferation and differentiation, but generally show an increase of 5-Bromodeoxyuridine (BrdU)+ cells in the germinative zones of the brain [8,9]. In the SVZ, enhanced neurogenesis has been shown after aspiration lesions, inflammatory demyelization, and cortical transection lesions [10-13], although most studies have focused on the changes following focal ischemic injury, where consistent increases in the numbers of SVZ progenitor cells have been reported [13-16]. Obviously of interest is the fate and life span of newly generated progenitor cells. These cells are able to express the appropriate neuronal phenotype; however, the majority disappear by 5 weeks following the lesion [14,17]. Thus, the capacity of newly generated progenitors for CNS repair is still under debate [18]. Nevertheless, SVZ neural precursors are seen to migrate towards the primarily degenerated injury core where they are able to differentiate into neurons and glia [15]. In the SGZ of the dentate gyrus, brain insults also promote enhanced neurogenesis; increased numbers of progenitor cells have been seen after adult brain mechanical lesions, epileptic seizures, excitotoxicity and stroke [19-21]. Moreover, after global ischemia, treatment with growth factors can increase CA1 neuronal renewal and subsequent functional improvements [22].
In addition to a general injury-induced increase in progenitor cells, there are key differences in the regional proliferation and differentiation dynamics after distinct types of injury and at different ages. In the immature rat brain, it seems reasonable that postnatal dynamics of the germinative zones would be unique. A plasticity window period in the early postnatal rat brain has been defined, between days 6 and 10, and extensively studied by Kolb and his colleagues, showing that the immature cortex is capable of dendritic sprouting and cortical reorganization after injury. This plasticity results in a lack of abnormalities in cortico-cortical and subcortico-cortical connections; moreover, normal behavioral abilities are retained [23-27]. In this sense, the postnatal rat brain provides an optimal experimental model for the study of neuronal injury and the concurrent cell response to degenerative and regenerative processes. Specifically, the contribution of progenitor pools to the enhanced recovery during the plasticity window remains unknown.
To our knowledge, in spite of the fact that the post-injury environment in the developing brain differs greatly from the adult brain, there are only a few recent studies on the developmental changes of progenitor pools in the early postnatal brain and their response to brain damage, which provide contradictory data. Whereas two studies have shown decreased numbers of SVZ oligodendrocyte precursors after hypoxic-ischemic injury and a lasting decrease of neural stem cells [28,29], a more recent study has showed hypoxic-ischemic induced stimulated cell proliferation and neurogenesis in the SVZ and peri-infarct striatum [30]. Therefore, further studies based on the early postnatal stage are necessary to understand the dynamics of neurogenic and gliogenic proliferating cell populations in the developing brain and to clarify the links between developmental plasticity and neurogenesis after injury. In the most hopeful of scenarios these cells would be able to proliferate, differentiate, and integrate into injured tissue to facilitate neurological recovery from damage.
Accordingly, the aim of our study was to evaluate the temporal and spatial dynamics of cell proliferation in the three main germinative zones of the brain during normal postnatal development and following excitotoxic injury caused by an injection of N-methyl-D-aspartate (NMDA) in 9-day old rat pups. Moreover, proliferating cells in these areas were characterized by using double labeling with specific cell markers for progenitors, mature glia and neurons.
Results
Injection of 27 ηmol of NMDA into the right sensimotor cortex of 9-day old rats caused neuronal degeneration accompanied by a glial response in the entire cortex at the level of the injection site and surrounding tissue, extending to the septum, striatum and rostral hippocampus. No affectation was seen in the contralateral hemisphere, as has been previously described in detail [31,32].
In the intact postnatal brain, maximal proliferation is seen at P12
In the control postnatal brain, maximal proliferation was observed at postnatal day (P) 12, persisted until P16-P23 and was clearly diminished by P49. Proliferative activity was observed in the RMS, SVZ and the SGZ as well as in the cortex, striatum, thalamus, hippocampus, capsula interna, corpus callosum, and the lateral septal nuclei. In these postnatal controls, two patterns of BrdU-positive cell distribution were observed: grouped BrdU+ cells and scattered BrdU+ cells, homogeneously distributed within the brain parenchyma. Grouped BrdU+ cells, which are the focus of our study, were seen at all time points mainly in the RMS, the SVZ, and the SGZ, but were also seen until P16 in several white matter tracts (the corpus callosum, hippocampal fimbria, internal capsula, and striatal white matter). Scattered BrdU+ cells were seen in the cortex and thalamus until P16.
In general, in the germinative zones, BrdU+ cell numbers in control brains reached maximal numbers at P12 and then gradually decreased as the animal aged (Fig. 1). In the SVZ, a significant increase in proliferating cell number was seen from P9 to P12, numbers remained statistically constant until P23, after which a significant decrease was seen from P23 to P49 (Fig. 2a). In the RMS, a slight increase in proliferating cell numbers was seen between P9 and P12, cell numbers significantly decreased from P12 to P14, remained statistically constant until P16, and then decreased again between P16 and P23, after which no change was noted (Fig. 2b). In the DG, cell numbers significantly increased from P9 to P12, to reach maximum levels, where after numbers decreased gradually until P49 (Fig. 2c).
Figure 1 Camera lucida drawings of anterior coronal brain sections. Dots represent BrdU+ cells. No difference in number or spatial and temporal distribution of BrdU+ cells was seen in the contralateral and control hemispheres at corresponding time points (corresponding time points: 10 h = P9, 3d = P12, 5d = P14, 7d = P16, 14d = P23, 40d = P49). Ipsilateral (ip) hemispheres show BrdU+ cell distribution following NMDA-induced excitotoxicity in coronal brain sections at the level of the neurodegenerating area (outlined in black) and SVZ (BrdU+ cells of the SVZ are represented in red). Proliferation was observed at 10 h (P9), peaked at 3d (P12) and diminished by 40d (P49).
Figure 2 Temporal dynamics of proliferation in the germinative zones of control brains. Trends of cell proliferation in the postnatal rat brain, from P9 to P49, in the SVZ (a), RMS (b), and DG (c) are shown. Peak proliferation was always seen at P12. Curves were calculated using a polynomial regression line (order = 5, period = 2).
Excitotoxic damage results in a reduction of BrdU+ cells in the ipsilateral SVZ and in the RMS at early survival times
In the lesioned postnatal brain, the contralateral hemisphere showed no differences in number and distribution when compared to the control postnatal brain at each corresponding time point. However, the ipsilateral damaged hemisphere showed changes in the amount and distribution of BrdU+ cells (Fig. 1). Noteworthy, were the increased amounts of BrdU+ cells observed in the neurodegenerating areas: the lesioned cortex, septum, rostral thalamus and striatum, where it was interesting to find grouped BrdU+ cells approximately at 80–100μn away from the SVZ.
In the ipsilateral germinal zones, changes in the number of BrdU+ cells were strongly dependent on location. In the SVZ, a striking decrease (58%) in the number of BrdU+ cells was already observed at 10 hours post lesion (Fig. 1 and Figs. 3a). At 3 days post lesion, the number of BrdU+ cells in the ipsilateral SVZ showed a 72% reduction when compared to the contralateral side, and this decrease was sustained until 14 days post lesion where an 88% decrease was seen (Fig. 3a, 4a–d). In the lesioned RMS, a significant reduction in the number of BrdU+ cells was only observed at 10 hours post lesion and 3 days post lesion whereas no changes in BrdU+ cell numbers were seen at later time points (Figs. 3b and 4e–h). In the DG of lesioned animals, although a higher density of BrdU+ cells was observed, there was notable ipsilateral hippocampal degeneration and shrinkage caused by the lesion. Because the cells were counted in the entire dentate gyrus, irrespective of area, there was no significant difference in the number of BrdU+ cells in the ipsilateral DG when compared to the contralateral control side or saline-injected controls (Figs. 3c and 4i–j).
Figure 3 Quantitative analysis of excitotoxicity-induced changes in the germinative zones. Mean values of the number of BrdU+ cells in the SVZ (a), RMS (b), and DG (c) are shown and compared between intact controls, and lesioned (NMDA injected) contralateral (cl) and ipsilateral (ip) hemispheres. No difference was seen between intact controls and contralateral hemispheres at any time. In the SVZ, there is a significant decrease (asterisk) between the total number of BrdU+ cells in ip hemispheres when compared to cl and control hemispheres at all time points except 40 days post lesion. In the RMS, a significant decrease (asterisk) was seen at 10 hours and 3 days post lesion in NMDA-injected brains when compared to controls. In the DG, no significant difference in BrdU+ cell number was observed at any time point in ip hemispheres were compared to cl or control hemispheres.
Figure 4 Immunohistochemical staining for BrdU in the SVZ, RMS and SGZ. The significant decrease in BrdU+ cell number in the ipsilateral (ip) SVZ following NMDA-induced excitotoxicity is shown in a coronal brain section at 3 days post lesion (3d; compare a, b). BrdU+ cells at all time points in the ip SVZ showed a sharp decrease in number when compared to control hemispheres until 40 days post lesion (40d), the last time point studied, where no difference was seen (c, d). In the ip RMS, a decrease in BrdU+ cells was seen at 10 hours post lesion (10 h) and at 3 days post lesion (e) when compared to the corresponding control RMS (P9 and P12 brains, respectively; f). Thereafter, the ip RMS (7–40 days post lesion) when compared to controls (P16-P49, respectively), showed no difference in BrdU+ cell numbers (g, h). The dentate gyrus of control and NMDA-injected hemispheres showed no differences in cell number at any time point, as shown in a coronal section of the DG at 3 days post lesion (i, j). Note the hippocampal shrinkage in the ipsilateral hemisphere (j).
Most BrdU+ cells in the germinative zones express nestin. GFAP+/nestin+ prolongations leaving the SVZ in the intact postnatal brain display changes in arrangement following damage
The study of double stained sections showed that most scattered BrdU+ cells found throughout the control brains until P16 were tomato lectin (TL)+ endothelial cells lining blood vessels (data not shown). Both in the controls and lesioned hemispheres, grouped BrdU+ cells in the germinal zones were mostly double labeled for nestin intermediate filament protein (nestin) and were negative for glial fibrillary acidic protein (GFAP) (Figs. 5a–g) and neuronal nuclear antigen (NeuN) (data not shown).
Figure 5 Phenotypes of proliferating cells in the SVZ. Optical microscope studies of coronal brain sections after double-labeled immunofluorescence showed that there is a characteristic placement of cell phenotypes along the entire length of the ventricular wall (vw). Photographs show TL+ ventricular ependymal cells (arrow, a) bordered BrdU+/nestin+ cells (arrows, b), which were found next to GFAP+ cells (arrow, c). Photographs of confocal imaging in the SVZ (more specifically, in zone a of the SVZ) revealed a similar placement where TL+ ependymal cells (d) of the ventricular wall are seen next to BrdU+/nestin+ progenitor cells (e; arrow, f) and GFAP+ cells (g). BrdU+/nestin+ cells outlined in white in photograph e can be seen at greater magnification in f. This patterning was seen at all time points except at postnatal day 9 (P9). At this time, in both control (P9) and lesioned brains (10 h), BrdU+/GFAP+ cells were also seen next to the ventricular wall (asterisk, h). At 10 hours post lesion, a decrease in the number of GFAP+ filaments in the ipsilateral (ip; i) hemisphere was also noted when compared to contralateral (cl; h) and control hemispheres. BrdU (red); nestin, TL, GFAP (green).
The SVZ area lining the ventricle showed a layered type distribution of the different cell markers. BrdU+/nestin+ cells were located between TL+ ependymal cells lining the ventricle and GFAP+ astrocytes in the parenchyma at all time points (Figs. 5a–c), except at P9, when less nestin expression was seen and GFAP+ structures were located alongside the TL+ ependymal cells of the lateral ventricle and within BrdU+ cells (Fig. 5h). In addition, until P14, GFAP+ and nestin+ prolongations were seen to extend radially from the top of the ventricle to the cortex, (Fig. 6a), curve from the SVZ to the corpus callosum (Fig. 6e), and extend from the ventricular wall to the striatum (Fig. 5b, c). These cells occasionally had a BrdU+ nucleus, and demonstrated either a bipolar morphology, or had a single prolongation leaving from the BrdU+ nucleus.
Figure 6 Nestin+ and GFAP+ Projections leaving the SVZ. In control brains at P12 and P14 GFAP+ nestin+ projections were seen leaving the SVZ extending towards the cortex (cx; arrows, a). After NMDA-induced excitotoxicity at 3 and 5 days post lesion, the presence of mature astrocytes with a star-like morphology were noticed next to the prolongations (b; asterisks, d). There was also a decrease observed in the number of prolongations (c, d) and orientation seemed disturbed such that many prolongations did not extend towards the cortex or were shorter (arrows, c) and did not reach the outer layers of the cortex (b-d). At 3 and 5 days post lesion note that the GFAP+ and nestin+ prolongations were also seen to extend into the corpus callosum (cc; e) and towards the striatum (see figure 5, b-c) in control brains at P12 and P14. BrdU (red); nestin, GFAP (green).
In the lesioned hemisphere, there was a decrease in the amount of prolongations leaving the SVZ when compared to the contralateral hemisphere (compare Figs. 5h and 5i). Mainly at 10 hours post lesion, but also at 3 and 5 days post lesion, there were disturbances in the orientation of GFAP+ and nestin+ prolongations (Fig. 6c, d). Moreover, at 3 and 5 days post lesion, a number of GFAP+ astrocytes were also seen alongside these prolongations (Fig. 6b, d).
As mentioned above, grouped BrdU+ cells were observed in the striatum of lesioned animals, close to the SVZ. These grouped BrdU+ cells were mainly BrdU+/GFAP+ (Fig. 7a–c) and BrdU+/nestin+ (Fig. 7d-–g). It was possible to locate mature astrocytes, with a typical star-like morphology, but also cells with a bipolar morphology and cells with few projections (Fig. 7). Additionally, a small population of BrdU+ cells in this area were TL+ (Fig. 7h).
Figure 7 Double immunohistochemical staining of BrdU+ cell clusters in the striatum. In lesioned animals, from 3 days post lesion until 7 days post lesion grouped BrdU+ cells were found in the parenchyma a short distance away from the ventricular wall. Double labeling immunofluorescence studies followed by confocal imaging revealed that striatal BrdU+ cells colocalized with GFAP (arrow, a, b-c) and nestin (d, e-g). Within these groups it was possible to located cells with unipolar (arrow, a), bipolar (arrows, b), and typical mature astrocyte star-like morphology (b, c). Few BrdU+ cells were seen to colocalize with TL (h). BrdU (red); nestin, TL, GFAP (green).
In the RMS, most of the BrdU+ cells were again seen to colocalize with nestin (Fig. 8a). No TL+, GFAP+, or NeuN+ cells were seen to colocalize with BrdU+ cells in the RMS. A tube-like migratory system was evident, as the BrdU+/nestin+ GFAP+ filaments surrounded cells. Especially spectacular, was the RMS at 10 hours post lesion, where GFAP+ filaments, surrounding BrdU+ cells were seen to curve and follow the route to the olfactory bulb (Fig. 8b).
Figure 8 Phenotypes of proliferating cells in the RMS and DG. Double-labeled immunofluoresence studies showed that in the RMS most cells were BrdU+/nestin+ (arrow, a) and revealed the presence of GFAP+ filaments (arrow, b) surrounding BrdU+ cells (asterisk, b). In the DG, BrdU+/nestin+ cells (c) were seen and a few BrdU+/GFAP+ cells could also be found (arrow, d, e). BrdU (red); nestin, GFAP (green).
In the dentate gyrus, most BrdU+ cells were nestin+ and negative for GFAP, a number located in the granule cell layer and others in the germinative sub-granular zone (Figs. 8c, d).
Discussion
This study shows that the number of proliferating cells in the three main germinative zones: SVZ, RMS and hippocampal SGZ vary during the postnatal development from P9 and until adulthood, reaching maximal proliferating cell numbers at P12. The majority of these germinative zone BrdU+ cells showed nestin labeling and some also coexpressed GFAP. Following excitotoxic damage to the immature brain there is a significant decrease in the number of proliferating cells in the ipsilateral SVZ at all time points studied and a notable decrease within the first 3 days post lesion in the RMS, whereas no post-lesion changes are seen in the dentate gyrus. Total cell numbers were counted in the germinative zones of the brain to obtain a representative sample of cells proliferating in these areas. Noteworthy, is the significant decrease in cell number despite observed tissue shrinkage and ventricular swelling, which affect the areas analyzed.
Changes in progenitor cell number
Proliferation studies have generally reported an increase of BrdU+ cells in the germinative zones of the brain, following injury to the adult rat brain [8,16,17]. However, the specific conclusions of the studies suggest that the response of neural progenitors to damage is variable and transient and its characteristics depend greatly on the specific germinative region, the time post-lesion and the time when BrdU is administered. Accordingly, proliferation dynamics of the germinative zones in the early postnatal brain are unique, especially during the plasticity window period, when the higher capacity for cortical reorganization and recovery could be related to a distinct behavior of progenitor pools.
Decreased SVZ proliferation in the damaged postnatal brain could be due to an increased capacity of progenitors for migration, possibly at a faster rate when compared to adults, thus leaving the SVZ "empty" or showing an apparent decrease of cells. In this sense, high-levels of polysialylated neural cell adhesion molecule (PSA-NCAM) have been shown to be present in the developing brain in progenitors of the prenatal cortex and in the developing SVZ [33]. And, high levels of PSA-NCAM are expressed in the adult cortex, subcortex and the SVZ after ischemia, suggesting a re-expression of a developmental phenotype in the injured adult brain [34]. Thus, it is likely that the postnatal brain has the capacity to upregulate migration in the germinative zones, and more so than the adult, since it has a greater capacity for regeneration and plasticity. Interestingly, TdT mediated d-UTP-biotin Nick End Labelling (TUNEL) staining shows no signs of apoptotic cell death in the SVZ (L. Acarin, unpublished results), which may suggest that BrdU immunoreactive cells of the SVZ are not dying following excitotoxic insult. This contrasts a recent study of neonatal hypoxia/ischemia in the perinatal brain where it was shown that injury resulted in depletion of the SVZ by death of neural stem cells and oligodendrocyte progenitors [28]. Migration of SVZ progenitors could also explain the results of this study showing that at 5 days post lesion and 7 days post lesion, the BrdU+ cells were found a short distance (a few nuclear diameters) away from the middle of the lateral ventricle, putatively indicating movement of progenitor cells away from the SVZ towards the striatum. In this regard, migration of progenitors towards damaged CNS parenchyma has been previously seen after chemoconvulsant-induced status epilepticus, where forebrain SVZ neurogenesis increases and a number of the neuroblasts destined for the olfactory bulb leave the RMS prematurely and migrate to the injured forebrain [13]. Likewise, it has been shown that neuroblasts having undergone division before injury can be recruited to the damaged striatum and new neurons can even be found in unaffected striatal areas [14,17]. In this study, in agreement with the hypothesis of premature deviation of progenitors from the RMS to repopulate other brain areas, a decrease of BrdU immunoreactive cells was noted in the RMS at 10 hours post lesion and 3 days post lesion, also in the absence of TUNEL staining in this area. This decrease could be due to a curtailed normal radial migration of a large quantity of cells from the SVZ to the RMS by chemoattractant molecules, which induce SVZ progenitor redirection to other brain areas that are excitotoxically damaged. Moreover, it should also be mentioned that the decrease in SVZ BrdU+ cells is no longer observed at 40 days post lesion, when the number of progenitors in the ipsilateral SVZ was not significantly different from contralateral control hemispheres. This finding further strengthens the migration or redirection theory versus cell depletion.
However, another possible explanation for the decrease in cell number seen in the SVZ, the timeframe in which gliogenesis takes place includes the first weeks of the postnatal period [35]. Glial cells generated in the SVZ migrate via radial fibers towards the cortex, striatum and dorsal white matter [36]. It is possible that NMDA-induced excitotoxic cortical damage disrupts these radial fiber directed migrational routes, preventing glial cell migration to their destined cortical zones, and inducing migration routes towards the less severely affected striatum. This could explain the disruption in nestin+ and GFAP+ prolongations leaving the SVZ and the presence of clusters of BrdU+ cells in the striatum and thalamus from 5–7 days post lesion.
The fact that no changes in the number of BrdU+ cells are observed in the SGZ of the dentate gyrus needs further elaboration. In contrast with the results of this study, insults to the adult CNS result in increases in DG neurogenesis. In cerebral ischemia, these increases are transitory, whereas a large increase in proliferating cell number is observed when BrdU is injected at 4–6 days after MCAO, while no changes are seen after 11–13 days [14,20,21,37,38]. In addition to the differences between postnatal and adult DG progenitor response to damage, it is plausible that precursors of the SVZ-RMS which are headed throughout postnatal life to give rise, for example, to late born neurons and astrocytes, and later to oligodendrocytes, would be promptly activated when cortical and striatal tissue damage is induced (as in the case of our study), whereas precursors of the SGZ would only be mobilized if massive granule cell death occurs in the DG.
Characterization of progenitor BrdU+ cells
In general, most BrdU+ cells located in the germinative zones express the intermediate filament nestin. This protein is a typical component of the cytoskeleton of immature cells and has previously been described as a marker of non-differentiated precursor cells in the three main germinative zones [39,40]. Accordingly, the BrdU+/nestin+ population of cells seen at all time points in the SVZ may correspond to the previously described resident stem cell population that turn into rapidly proliferating precursors (most of the nestin+ cells) which express PSA-NCAM and migrate along the RMS into the olfactory cortex where they proceed to renew populations of granular and periglomular neurons and astrocytes [41,42]. In a number of recent studies, upregulation of nestin has been described after injury by ischemic insult [39,43], and traumatic brain injury [44]. It has been suggested that nestin re-expression after injury recapitulates the scenario found in the developing brain, where nestin expression commits progenitor cells to differentiate into neurons and glia, and thus plays a role in brain remodeling and recovery [39,43,44]. In addition, a population of pre-oligodendrocytes, multipolar mitotically-active late oligodendrocyte progenitors have been described to originate in the SVZ of the lateral ventricles [45] and could account for a minority of BrdU+ nuclei in the SVZ that are not labeled by nestin, GFAP or TL.
Finally, it should be noted that although the main finding of this study is the decreased number of precursor cells in the damaged SVZ, changes in the nestin+/GFAP+ prolongations were also noted. First of all, the identity of these structures is not clear. It is known that the ventricular zone (VZ) in the early postnatal brain is largely composed of radial glial cell bodies [46] and that most dividing progenitor cells in the VZ express radial glial proteins and maintain a vimentin-positive radial fiber throughout each stage of cell division [47], which plays a role in determining progenitor migration towards the cortex. Therefore, in relation to these studies, the nestin positive prolongations seen in this investigation, to be leaving the SVZ, and either crossing the corpus callosum into the cortex, along the axis of the corpus callosum, or into the striatum, are suggestive of radial glial prolongations. Furthermore, the same types of prolongations were GFAP+ and some contained GFAP+ astrocytic cells in close association with typical star shaped morphologies, further suggesting that they could be radial glia like structures directing astrocytic migration. In this sense, it has been proposed that radial glia scaffolding that directs neuronal precursor migration during embryonic development, is responsible for generating and directing late neuronal precursor migration in the neocortex [48]. Recent studies have provided in vitro and in vivo evidence that radial glia are able to generate cortical neurons [46,48-53] and that these newly generated cells take on a bipolar shape and migrate along the radial fiber of the mother cell, which remains attached to the ventricular surface [48]. These prolongations, described as radially arranged vimentin-positive processes, have also been seen to interact with cells of unipolar and bipolar morphology just above the SVZ in the corpus callosum. In more dorsal white matter, these cells showed more complicated morphologies and could be early differentiating glial cells [54]. Accordingly, some studies have also suggested that subpopulations of radial glial cells, apart from their role in neuronal guidance may be GFAP+ neural progenitor cells [55], giving rise to astrocytes, ependymal cells and also GFAP+ precursor cells. Therefore, if indeed radial glial has precursor potential, the nestin+ prolongations noted in could be radial glial progenitor cells that differentiate into the GFAP+ associated astrocytes upon leaving the SVZ.
Alternatively, because nestin is known to be re-expressed in deafferented target areas [56] and by reactive astrocytes, this finding could also suggest that progenitors are undergoing lesion-induced gliogenesis, differentiating into astrocytes that become reactive as they reach the lesioned area, as has been previously described in a study of the spinal cord, where nestin+ periventricular cells migrate to the injury site and become GFAP+ [57].
Conclusion
The number of proliferating cells in the three main germinative zones: SVZ, RMS, and hippocampal SGZ, vary during postnatal development from P9 until adulthood, reaching maximal cell numbers at P12 and decreasing thereafter. Postnatal excitotoxic damage induces a significant decrease in the number of proliferating cells in the ipsilateral SVZ at all time points studied. A notable decrease is also observed within the first 3 days post lesion in the RMS, whereas no post-lesion changes are seen in the dentate gyrus. The majority of germinative zone BrdU+ cells are labeled for nestin and some of them coexpress GFAP. Understanding the proliferation dynamics of proliferating cells in the germinative zones of the brain in normal and pathological states is a key first step in harnessing any future potential for recovery that this endogenous population may provide.
Methods
Long Evans black hooded rats of both sexes were used in this study. All procedures were conducted in compliance with Spanish legislation and according to the European Union directives on this subject. The animal care committee of the Autonomous University of Barcelona approved protocols.
NMDA injections
Postnatal day 9 (P9) pups were anesthetized with isofluorane and then placed in a stereotaxic frame adapted for newborns (Kopf). The skull was opened using a surgical blade and 27 ηmoles of N-methyl-D-aspartate (NMDA; Sigma M-3262) diluted in 0.15 μl of saline solution (0.9% NaCl) were injected into the right sensorimotor cortex at the level of the coronal suture (2 mm lateral from bregma and 0.5 mm deep) with a 0.5 μl Hamilton microsyringe (needle gauge 25). After suture, the pups were placed in an incubator and maintained at 36° for 2 hours before being returned to their mothers, in order to maintain normothermia since NMDA-induced lesions are highly dependent on body temperature. In control animals the same procedure was followed but instead of a NMDA injection, pups received an injection of 0.15 μl of saline solution.
5'Bromodeoxyuridine (BrdU) injections
In order to examine the pattern and time course of proliferating cells in the brain BrdU, the thymidine analog which incorporates into the DNA of dividing cells during S-phase, was used to label actively proliferating cells 10 hours before sacrifice (Sigma Chemical, St Louis, MO, USA). NMDA injected, saline injected and intact controls were administered with BrdU. Pulse labelling was carried out by intraperitoneally injecting BrdU (50 mg/kg) diluted in TB (0.05 M Trizma base, pH 7.4) every 2 hours for 10 hours before sacrifice. Rats were killed at 10 hours, 3 days, 5 days, 7 days, 14 days and 40 days post-lesion. Five NMDA injected animals and two saline controls were used for each survival time. In addition, two non-lesioned rats for each survival time (aged P9, P12, P14, P16, P23 and P49 corresponding to 10 hours, 3, 5, 7, 14 and 40 days post lesion, respectively) received the same injection of BrdU and were used as control groups, in order to study changes in proliferation during development.
Tissue processing
Rats were anesthetized and then sacrificed by intracardial perfusion for 10 minutes with 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4). Brains were immediately removed and immersed in the same fixative for 2 hours and then cryoprotected in a 30% sucrose solution in 0.1 M phosphate buffer. Brains were then frozen with dry CO2 and 30 μm thick sections were cut using a cryostat (Leitz).
Immunohistochemistry
Free-floating cryostat sections were processed for the visualization of BrdU, for the specific demonstration of proliferating cells. After endogenous peroxidase blocking with 2% H202 in 70% methanol for 10 minutes, DNA was denatured by first incubating in 0.082N HCl for 10 min. at 4°C and then for 30 min. in 0.82N HCl at 37°C. Sections were rinsed with TBS (0.05 M Trizma base containing 150 mM of NaCl, pH 7.4), borate buffer (pH 8.5), and 0.5% Triton X-100 in TBS (TBS-T) and then incubated in blocking buffer (BB; 10% fetal calf serum in TBS-T) for 1 hour and incubated overnight at 4°C and 1 hour at room temperature (RT) with a primary mouse anti-BrdU antibody (DAKO, Glostrup, Denmark) diluted to 1:80 in BB. After washing with TBS-T, the sections were incubated for 1 hour at RT with a secondary anti-mouse IgG biotinylated antibody (Amersham, Buckinghamshire, England, UK) in a 1:200 dilution in BB, rinsed again and incubated for 1 hour at RT in a 1:400 dilution of avidin-peroxidase (DAKO, Glostrup, Denmark) in BB. After rinsing again in TBS-T and TBS, the reaction product was detected using 50 mg of 3'3-diaminobenzidine-tetrahydrochloride (DAB, Sigma, St. Louis, MO, USA) and 33 μl of H2O2 in 100 ml of Tris buffer.
Double staining for BrdU and NeuN, GFAP and nestin was performed. Sections were first incubated with the anti-BrdU antibody overnight at 4°C and 1 hour at RT before incubation with a secondary goat anti-mouse Cy3 antibody (Amersham, Buckinghamshire, England, UK) for 1 hour at RT. Sections were then incubated overnight with either a mouse monoclonal antibody against NeuN (Chemicon, Temecula, CA, USA), a rabbit polyclonal antibody against GFAP (DAKO, Glostrup, Denmark), or a mouse monoclonal antibody against nestin (Chemicon, Temecula, CA, USA) diluted in BB at 1:500, 1:1800, and 1:1000 respectively. After rinsing, sections were incubated for 1 hour at RT with either a secondary goat anti-mouse Cy2 antibody (Amersham, Buckinghamshire, England, UK) or a secondary goat anti-rabbit Cy2 antibody (Amersham, Buckinghamshire, England, UK) diluted in BB to 1:1000. For TL histochemistry, sections processed for BrdU were incubated with a biotinylated lectin obtained from Lycopersicon esculentum (tomato; Sigma, St. Louis, MO) diluted to 6 μg/ml in TBS-T and then incubated for 1 hour at RT with a Cy2-conjugated streptavidin (Amersham, Buckinghamshire, England, UK) at a dilution of 1:1000.
Image analysis & quantification
Measurements of BrdU immunoreactive cells were performed on cryostat sections (thickness of 30 μm). Sections were digitized under a 20X objective using a digital camera (Nikon Dxm1200, Tokyo, Japan) mounted on a Leitz microscope and interfaced to a PC computer using ACT1 Imaging software. Every 8th coronal section was selected from each rat for a total of 4 sections. All BrdU immunopositive nuclei were counted, using AnalySIS software (version 3.2, Soft Image Analysis System, Münster, Germany), in digitalized sections (for the best visualization and to avoid over-sampling) of the: RMS, SVZ and the entire dentate gyrus, encompassing the 2 cell body wide SGZ along the border of the granule cell layer (GCL) and the hilus. BrdU nuclei in these areas are presented as the total number of cells/section. The total number of cells in each section was averaged to obtain a mean value for each animal.
Data were analyzed using an ANOVA one-way analysis of variance followed by a Fishers test comparison. All values are presented as Mean ± Standard Error (S.E.) Statistical significance was set at P < 0.05. Because no statistically significant difference in the amount of BrdU positive cells was observed between the saline-injected and intact control animals, data from these two groups was combined to form a collective control group, which was compared to NMDA-lesioned animals.
Double stained sections were first analyzed with an optical fluorescence microscope using 10X, 20X, and 40X dry objective lenses and a 100X oil immersion objective lens. Selected sections were then analyzed and photographed using a Leica confocal microscope (TCS SP2 AOBS) where data was acquired using a 63X oil immersion lens. For sections stained with BrdU and NeuN, GFAP, TL or nestin, red (Cy3 for BrdU) and green (Cy2 for NeuN, GFAP, TL, and nestin) fluorochromes were excited by a laser beam and emissions were sequentially acquired with 2 separate photomultiplier tubes with their respective emission filters.
Authors' contributions
MF performed the immunohistochemical study, part of the excitotoxic lesions, the quantification analysis and drafted the manuscript. LA did part of the excitotoxic lesions, helped in the analysis and interpretation of the data and the critical review of the manuscript. BC and BG coordinated the development of the study, revised the last version of the manuscript and are responsible for economic support. All authors have given final approval of the version to be published.
Acknowledgements
This work was supported by a DGES [BF2002-02079] and Fundació La Caixa [00/074-00]. We thank Miguel A. Martil for his excellent technical help.
==== Refs
Garcia-Verdugo JM Doetsch F Wichterle H Lim DA Alvarez-Buylla A Architecture and cell types of the adult subventricular zone: in search of the stem cells J Neurobiol 1998 36 234 248 9712307 10.1002/(SICI)1097-4695(199808)36:2<234::AID-NEU10>3.0.CO;2-E
Cameron HA McKay R Stem cells and neurogenesis in the adult brain Curr Opin Neurobiol 1998 8 677 680 9811628 10.1016/S0959-4388(98)80099-8
Gage FH Kempermann G Palmer TD Peterson DA Ray J Multipotent progenitor cells in the adult dentate gyrus J Neurobiol 1998 36 249 266 9712308 10.1002/(SICI)1097-4695(199808)36:2<249::AID-NEU11>3.0.CO;2-9
Palmer TD Markakis EA Willhoite AR Safar F Gage FH Fibroblast growth factor-2 activates a latent neurogenic program in neural stem cells from diverse regions of the adult CNS J Neurosci 1999 19 8487 8497 10493749
Palmer TD Ray J Gage FH FGF-2-responsive neuronal progenitors reside in proliferative and quiescent regions of the adult rodent brain Mol Cell Neurosci 1995 6 474 486 8581317 10.1006/mcne.1995.1035
Nunes MC Roy NS Keyoung HM Goodman RR McKhann G Jiang L Kang J Nedergaard M Goldman SA Identification and isolation of multipotential neural progenitor cells from the subcortical white matter of the adult human brain Nat Med 2003 9 439 447 12627226 10.1038/nm837
Belachew S Chittajallu R Aguirre AA Yuan X Kirby M Anderson S Gallo V Postnatal NG2 proteoglycan-expressing progenitor cells are intrinsically multipotent and generate functional neurons J Cell Biol 2003 161 169 186 12682089 10.1083/jcb.200210110
Kokaia Z Lindvall O Neurogenesis after ischaemic brain insults Curr Opin Neurobiol 2003 13 127 132 12593991 10.1016/S0959-4388(03)00017-5
Parent JM Injury-induced neurogenesis in the adult mammalian brain Neuroscientist 2003 9 261 272 12934709 10.1177/1073858403252680
Calza L Giardino L Pozza M Bettelli C Micera A Aloe L Proliferation and phenotype regulation in the subventricular zone during experimental allergic encephalomyelitis: in vivo evidence of a role for nerve growth factor Proc Natl Acad Sci U S A 1998 95 3209 3214 9501242 10.1073/pnas.95.6.3209
Szele FG Chesselet MF Cortical lesions induce an increase in cell number and PSA-NCAM expression in the subventricular zone of adult rats J Comp Neurol 1996 368 439 454 8725350 10.1002/(SICI)1096-9861(19960506)368:3<439::AID-CNE9>3.0.CO;2-6
Weinstein DE Burrola P Kilpatrick TJ Increased proliferation of precursor cells in the adult rat brain after targeted lesioning Brain Res 1996 743 11 16 9017224 10.1016/S0006-8993(96)00979-1
Parent JM Valentin VV Lowenstein DH Prolonged seizures increase proliferating neuroblasts in the adult rat subventricular zone-olfactory bulb pathway J Neurosci 2002 22 3174 3188 11943819
Arvidsson A Collin T Kirik D Kokaia Z Lindvall O Neuronal replacement from endogenous precursors in the adult brain after stroke Nat Med 2002 8 963 970 12161747 10.1038/nm747
Jin K Minami M Lan JQ Mao XO Batteur S Simon RP Greenberg DA Neurogenesis in dentate subgranular zone and rostral subventricular zone after focal cerebral ischemia in the rat Proc Natl Acad Sci U S A 2001 98 4710 4715 11296300 10.1073/pnas.081011098
Zhang RL Zhang ZG Zhang L Chopp M Proliferation and differentiation of progenitor cells in the cortex and the subventricular zone in the adult rat after focal cerebral ischemia Neuroscience 2001 105 33 41 11483298 10.1016/S0306-4522(01)00117-8
Parent JM Vexler ZS Gong C Derugin N Ferriero DM Rat forebrain neurogenesis and striatal neuron replacement after focal stroke Ann Neurol 2002 52 802 813 12447935 10.1002/ana.10393
Felling RJ Levison SW Enhanced neurogenesis following stroke J Neurosci Res 2003 73 277 283 12868061 10.1002/jnr.10670
Gould E Tanapat P Lesion-induced proliferation of neuronal progenitors in the dentate gyrus of the adult rat Neuroscience 1997 80 427 436 9284346 10.1016/S0306-4522(97)00127-9
Liu J Solway K Messing RO Sharp FR Increased neurogenesis in the dentate gyrus after transient global ischemia in gerbils J Neurosci 1998 18 7768 7778 9742147
Parent JM Lowenstein DH Mossy fiber reorganization in the epileptic hippocampus Curr Opin Neurol 1997 10 103 109 9146991
Nakatomi H Kuriu T Okabe S Yamamoto S Hatano O Kawahara N Tamura A Kirino T Nakafuku M Regeneration of hippocampal pyramidal neurons after ischemic brain injury by recruitment of endogenous neural progenitors Cell 2002 110 429 441 12202033 10.1016/S0092-8674(02)00862-0
Kolb B Kolb B and Tees RC Sparing and recovery of function The Cerebral Cortex of the Rat 1990 Cambridge, MA, MIT Press 537 561
Kolb B Gibb R van der Kooy D Neonatal frontal cortical lesions in rats alter cortical structure and connectivity Brain Res 1994 645 85 97 8062102 10.1016/0006-8993(94)91641-1
Kolb B Ladowski R Gibb R Gorny G Does dendritic growth underly recovery from neonatal occipital lesions in rats Behav Brain Res 1996 77 125 133 8762163 10.1016/0166-4328(95)00208-1
Kolb B Petrie B Cioe J Recovery from early cortical damage in rats, VII. Comparison of the behavioural and anatomical effects of medial prefrontal lesions at different ages of neural maturation Behav Brain Res 1996 79 1 14 8883811 10.1016/0166-4328(95)00254-5
Kolb B Gibb R Gorny G Whishaw IQ Possible regeneration of rat medial frontal cortex following neonatal frontal lesions Behav Brain Res 1998 91 127 141 9578446 10.1016/S0166-4328(97)00112-5
Levison SW Rothstein RP Romanko MJ Snyder MJ Meyers RL Vannucci SJ Hypoxia/ischemia depletes the rat perinatal subventricular zone of oligodendrocyte progenitors and neural stem cells Dev Neurosci 2001 23 234 247 11598326 10.1159/000046149
Skoff RP Bessert DA Barks JD Song D Cerghet M Silverstein FS Hypoxic-ischemic injury results in acute disruption of myelin gene expression and death of oligodendroglial precursors in neonatal mice Int J Dev Neurosci 2001 19 197 208 11255033 10.1016/S0736-5748(00)00075-7
Plane JM Liu R Wang TW Silverstein FS Parent JM Neonatal hypoxic-ischemic injury increases forebrain subventricular zone neurogenesis in the mouse Neurobiol Dis 2004 16 585 595 15262271 10.1016/j.nbd.2004.04.003
Acarin L Gonzalez B Castro AJ Castellano B Primary cortical glial reaction versus secondary thalamic glial response in the excitotoxically injured young brain: microglial/macrophage response and major histocompatibility complex class I and II expression Neuroscience 1999 89 549 565 10077335 10.1016/S0306-4522(98)00331-5
Acarin L Gonzalez B Hidalgo J Castro AJ Castellano B Primary cortical glial reaction versus secondary thalamic glial response in the excitotoxically injured young brain: astroglial response and metallothionein expression Neuroscience 1999 92 827 839 10426525 10.1016/S0306-4522(99)00022-6
Chuong CM Edelman GM Alterations in neural cell adhesion molecules during development of different regions of the nervous system J Neurosci 1984 4 2354 2368 6481452
Cramer SC Chopp M Recovery recapitulates ontogeny Trends Neurosci 2000 23 265 271 10838596 10.1016/S0166-2236(00)01562-9
Skoff RP Knapp PE Kettenmann H and Ransom BR The origins and lineages of macroglial cells Neuroglia 1995 New York, Oxford University Press 135 148
Levison SW Goldman JE Both oligodendrocytes and astrocytes develop from progenitors in the subventricular zone of postnatal rat forebrain Neuron 1993 10 201 212 8439409 10.1016/0896-6273(93)90311-E
Kee NJ Preston E Wojtowicz JM Enhanced neurogenesis after transient global ischemia in the dentate gyrus of the rat Exp Brain Res 2001 136 313 320 11243473 10.1007/s002210000591
Madsen TM Treschow A Bengzon J Bolwig TG Lindvall O Tingstrom A Increased neurogenesis in a model of electroconvulsive therapy Biol Psychiatry 2000 47 1043 1049 10862803 10.1016/S0006-3223(00)00228-6
Li Y Chopp M Temporal profile of nestin expression after focal cerebral ischemia in adult rat Brain Res 1999 838 1 10 10446310 10.1016/S0006-8993(99)01502-4
Lendahl U Zimmerman LB McKay RD CNS stem cells express a new class of intermediate filament protein Cell 1990 60 585 595 1689217 10.1016/0092-8674(90)90662-X
Lois C Alvarez-Buylla A Proliferating subventricular zone cells in the adult mammalian forebrain can differentiate into neurons and glia Proc Natl Acad Sci U S A 1993 90 2074 2077 8446631
Luskin MB Parnavelas JG Barfield JA Neurons, astrocytes, and oligodendrocytes of the rat cerebral cortex originate from separate progenitor cells: an ultrastructural analysis of clonally related cells J Neurosci 1993 13 1730 1750 8463848
Duggal N Schmidt-Kastner R Hakim AM Nestin expression in reactive astrocytes following focal cerebral ischemia in rats Brain Res 1997 768 1 9 9369294 10.1016/S0006-8993(97)00588-X
Sahin Kaya S Mahmood A Li Y Yavuz E Chopp M Expression of nestin after traumatic brain injury in rat brain Brain Res 1999 840 153 157 10517963 10.1016/S0006-8993(99)01757-6
Hardy RJ Dorsoventral patterning and oligodendroglial specification in the developing central nervous system J Neurosci Res 1997 50 139 145 9373024 10.1002/(SICI)1097-4547(19971015)50:2<139::AID-JNR3>3.0.CO;2-G
Tramontin AD Garcia-Verdugo JM Lim DA Alvarez-Buylla A Postnatal development of radial glia and the ventricular zone (VZ): a continuum of the neural stem cell compartment Cereb Cortex 2003 13 580 587 12764031 10.1093/cercor/13.6.580
Weissman T Noctor SC Clinton BK Honig LS Kriegstein AR Neurogenic radial glial cells in reptile, rodent and human: from mitosis to migration Cereb Cortex 2003 13 550 559 12764028 10.1093/cercor/13.6.550
Noctor SC Flint AC Weissman TA Dammerman RS Kriegstein AR Neurons derived from radial glial cells establish radial units in neocortex Nature 2001 409 714 720 11217860 10.1038/35055553
Alvarez-Buylla A Garcia-Verdugo JM Neurogenesis in adult subventricular zone J Neurosci 2002 22 629 634 11826091
Malatesta P Hartfuss E Gotz M Isolation of radial glial cells by fluorescent-activated cell sorting reveals a neuronal lineage Development 2000 127 5253 5263 11076748
Hartfuss E Galli R Heins N Gotz M Characterization of CNS precursor subtypes and radial glia Dev Biol 2001 229 15 30 11133151 10.1006/dbio.2000.9962
Noctor SC Flint AC Weissman TA Wong WS Clinton BK Kriegstein AR Dividing precursor cells of the embryonic cortical ventricular zone have morphological and molecular characteristics of radial glia J Neurosci 2002 22 3161 3173 11943818
Tamamaki N Nakamura K Okamoto K Kaneko T Radial glia is a progenitor of neocortical neurons in the developing cerebral cortex Neurosci Res 2001 41 51 60 11535293 10.1016/S0168-0102(01)00259-0
Suzuki SO Goldman JE Multiple cell populations in the early postnatal subventricular zone take distinct migratory pathways: a dynamic study of glial and neuronal progenitor migration J Neurosci 2003 23 4240 4250 12764112
Doetsch F Caille I Lim DA Garcia-Verdugo JM Alvarez-Buylla A Subventricular zone astrocytes are neural stem cells in the adult mammalian brain Cell 1999 97 703 716 10380923 10.1016/S0092-8674(00)80783-7
Brook GA Perez-Bouza A Noth J Nacimiento W Astrocytes re-express nestin in deafferented target territories of the adult rat hippocampus Neuroreport 1999 10 1007 1011 10321476
Johansson CB Momma S Clarke DL Risling M Lendahl U Frisen J Identification of a neural stem cell in the adult mammalian central nervous system Cell 1999 96 25 34 9989494 10.1016/S0092-8674(00)80956-3
| 15826306 | PMC1087489 | CC BY | 2021-01-04 16:39:09 | no | BMC Neurosci. 2005 Apr 12; 6:26 | utf-8 | BMC Neurosci | 2,005 | 10.1186/1471-2202-6-26 | oa_comm |
==== Front
BMC NeurosciBMC Neuroscience1471-2202BioMed Central London 1471-2202-6-271582631810.1186/1471-2202-6-27Research ArticleThe neuronal ceroid lipofuscinosis Cln8 gene expression is developmentally regulated in mouse brain and up-regulated in the hippocampal kindling model of epilepsy Lonka Liina [email protected] Antti [email protected] Outi [email protected] Mervi [email protected] Zaal [email protected] Mart [email protected] Anna-Elina [email protected] Neuroscience Center, University of Helsinki, Finland2 Folkhälsan Institute of Genetics and Department of Medical Genetics, University of Helsinki, Finland3 Institute of Biotechnology, University of Helsinki, Finland4 National Public Health Institute, Finland5 Laboratory of Neural Stem Cell Biology, Section of Restorative Neurology, Lund Strategic Research Center for Stem Cell Biology and Cell Therapy, University Hospital, Lund, Sweden2005 13 4 2005 6 27 27 29 9 2004 13 4 2005 Copyright © 2005 Lonka 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 neuronal ceroid lipofuscinoses (NCLs) are a group of inherited neurodegenerative disorders characterized by accumulation of autofluorescent material in many tissues, especially in neurons. Mutations in the CLN8 gene, encoding an endoplasmic reticulum (ER) transmembrane protein of unknown function, underlie NCL phenotypes in humans and mice. The human phenotype is characterized by epilepsy, progressive psychomotor deterioration and visual loss, while motor neuron degeneration (mnd) mice with a Cln8 mutation show progressive motor neuron dysfunction and retinal degeneration.
Results
We investigated spatial and temporal expression of Cln8 messenger ribonucleic acid (mRNA) using in situ hybridization, reverse transcriptase polymerase chain reaction (RT-PCR) and northern blotting. Cln8 is ubiquitously expressed at low levels in embryonic and adult tissues. In prenatal embryos Cln8 is most prominently expressed in the developing gastrointestinal tract, dorsal root ganglia (DRG) and brain. In postnatal brain the highest expression is in the cortex and hippocampus. Expression of Cln8 mRNA in the central nervous system (CNS) was also analyzed in the hippocampal electrical kindling model of epilepsy, in which Cln8 expression was rapidly up-regulated in hippocampal pyramidal and granular neurons.
Conclusion
Expression of Cln8 in the developing and mature brain suggests roles for Cln8 in maturation, differentiation and supporting the survival of different neuronal populations. The relevance of Cln8 up-regulation in hippocampal neurons of kindled mice should be further explored.
==== Body
Background
The neuronal ceroid lipofuscinoses (NCLs) comprise a group of human neurodegenerative disorders (CLN1-CLN8) characterized by epilepsy, visual failure, psychomotor deterioration and accumulation of autofluorescent lipopigment in many tissues, especially in neurons [1]. Six genes underlying human NCLs have been identified and the proteins initially characterized (reviewed in [2]). Naturally occurring mouse models exist for CLN6 and CLN8 [3-5], while mouse models for CLN1, CLN2 and CLN3 have been generated by gene targeting [6-10].
The ubiquitously expressed CLN8 gene encodes a transmembrane protein which localizes to the ER and the ER-Golgi intermediate compartment in non-neuronal cells and to the ER in neuronal cells [4,11,12]. Mutations in CLN8 result in two distinct NCL phenotypes in humans: Northern epilepsy (Progressive epilepsy with mental retardation, EPMR, OMIM 600143) described in Finnish patients, and variant late infantile onset NCL in a subset of Turkish patients [4,13]. EPMR is characterized by frequent drug-resistant epileptic seizures with onset at 5–10 years of age, followed by progressive mental retardation [14], while the Turkish patients show earlier onset and a more rapid progression [15]. In EPMR, intracellular storage material, including subunit c of the mitochonrdial ATP synthase as the main protein component, is most prominent in the CNS, especially in the third layer of the isocortex and hippocampal regions CA2-4 [16]. In the cerebral isocortex the pyramidal cells of the deeper parts of lamina III are severely ballooned, and hippocampal region CA2 shows neuronal loss and neuronophagy [16]. A frameshift mutation in mouse Cln8 that predicts a truncated protein underlies the phenotype of the mnd mouse, a naturally occuring NCL mouse model [4]. Contrary to human patients, epilepsy is not a prominent feature in mnd, which is characterized by progressive motor neuron dysfunction and retinal degeneration [5,17,18]. The brain appears to remain relatively intact [5,17,19]. Accumulation of subunit c of the mitochondrial ATP synthase, neurofilament redistribution in spinal motor neurons and accumulation of ubiquitin deposits are characteristic for mnd [20-22].
Here we characterized the spatial and temporal expression of Cln8 mRNA in mice. Moreover, as epilepsy is the dominant phenotype in human patients, we investigated the CNS expression of Cln8 mRNA in the hippocampal electrical kindling model of epilepsy in which repeated electrical stimulations trigger a progressive intensification of epileptiform responses, and kindled mice retain abnormal excitability thereafter [23,24].
Results
The Cln8 gene is ubiquitously expressed in mouse tissues
The expression of the Cln8 gene in mouse tissues was first analyzed by northern blot and real-time quantitative RT-PCR analyses. In northern blot analysis one ~3 kilobase (kb) Cln8 specific transcript was detected in all tissues analyzed including a 14-day embryo (E14) (Fig. 1A). In addition, one ~7 kb transcript was seen in all tissues except testis and heart (Fig. 1A). In spleen, an additional transcript of ~2 kb was detected (Fig. 1A). The ~7 kb and ~2 kb transcripts were notably weaker than the ~3 kb transcript. RT-PCR analysis covering the 867 bp open reading frame of Cln8 resulted in a single fragment of the same size in 12 different mouse tissues (data not shown), suggesting that the different transcripts detected in the northern analysis are not due to alternative splicing in the coding region.
We then estimated the level of Cln8 expression using real-time quantitative RT-PCR analysis. Cln8 showed highest expression in liver, spleen, heart and skeletal muscle, 2.5, 2.2, 1.7 and 1.6 fold higher (respectively) than in the brain (Fig. 1B). Expression was lowest in 11-day embryo and adult kidney, lung and testis, 0.6, 0.9, 0.5 and 0.03 fold lower than in the brain (Fig. 1B). This indicates that Cln8 expression levels in adult mouse tissues as well as in whole mouse embryos are not dramatically different.
Cln8 is expressed throughout mouse development
The tissue expression of Cln8 during development and brain maturation was characterized using radioactive mRNA in situ hybridization analysis. For this purpose, whole embryo sections from E13, E15.5 and E17 mice and brain sections from postnatal (P) P0, P5, P10 and adult mice were used and the expression of Cln8 was quantitated by a computer based MCID image analysis system.
In fresh-frozen E13 and E17 embryos the overall expression of Cln8 was weak. In E17 the most prominent expression was detected in the developing gastrointestinal tract (Fig. 2A, 2B). In addition, there was high expression in the DRG neurons (Fig. 2C, 2D). Some Cln8-specific signal was also detected in the developing brain (Fig. 2E). The expression of Cln8 in E13 was similar to that in E17 except in the DRG neurons, where no signal was detected (data not shown). In paraffin-embedded E15.5 embryos a Cln8-specific signal was detected in the developing gastrointestinal tract, adrenal glands and the developing brain, especially in the cortical plate (data not shown).
In postnatal mice (P0, P5, P10 and adult) Cln8 expression was quantitated in the cortex, cerebellum and different hippocampal regions. Expression at P0 was low but specific in hippocampal regions CA1, CA3 and the granular cell layer of the dentate gyrus and in the cortex (Fig. 3). At P5 and P10, the Cln8 expression increased in the hippocampal CA1 region (to 115% and 106% of P0) and especially in the CA3 region (to 139% of P0) (Fig. 3). In adult brain the expression levels of Cln8 were lower than at P0, P5 and P10 in every region analyzed (Fig. 3). When compared with adults, a statistically significant difference in Cln8 expression was detected in the cortex at P0, P5 and P10 and in the CA3 region at P10 (Fig. 3). Expression of Cln8 in adult cerebellum was 156% of expression in the cortex (data not shown). In all maturation stages, the expression of Cln8 was highest in hippocampal region CA3 (Fig. 3).
Cln8 expression is strongly up-regulated in the hippocampal electrical kindling model of epilepsy
Epileptic seizures are a characteristic feature in human patients with CLN8 mutations. We therefore analyzed, by mRNA in situ hybridization analysis, possible changes in Cln8 expression 2 h, 6 h and 24 h after kindling-induced epileptic seizures. Animals were subjected to 40 rapid kindling stimulations which lead to increased excitability in mice at 4 weeks after the stimulation [25]. Each rapid kindling stimulation induced relatively short focal (grade 0–2) or long-lasting generalized (grade 4–5) seizures. The mean duration of afterdischarges was 20 ± 3 s for focal seizures and 76 ± 8 s for generalized seizures. All animals experienced multiple (11 ± 2) grade 4–5 generalized tonic-clonic seizures. The expression of Cln8 in the control, electrode implanted but non-stimulated mice (0 h; Fig. 4) was essentially identical to that in wild-type adult mouse brain (Fig. 3). After kindling stimulations Cln8 was strongly up-regulated in the hippocampus, especially in the CA3 region and most prominently in the granular cell layer of the detate gyrus (Fig. 4). After 2 h, 6 h and 24 h expression of Cln8 in CA3 was 92%, 114% and 134% of that of controls, and in the granular cell layer of the dentate gyrus 128%, 225% and 286% of controls, respectively (Fig. 4). When compared to controls, a statistically significant increase in Cln8 expression was detected in the cortex after 6 h, and in the CA1 and CA3 regions, and the granular cell layer of the dentate gyrus after 2 h, 6 h and 24 h (Fig. 4). All changes in Cln8 mRNA expression after rapid kindling stimulations were bilateral with no differences between the sides.
Discussion
We analyzed the expression of the Cln8 gene during mouse development and in the hippocampal kindling model of epilepsy to gain understanding of the role of CLN8 in the disease mechanisms underlying both human and mouse NCLs. In northern blot analysis a Cln8 transcript of approximately 3 kb was detected in all tissues studied. Two additional transcripts, approximately 2 kb and 7 kb, were detected in several tissues. Thus, Cln8 expression resembles that of human CLN8, which is ubiquitously expressed with transcripts of 1.4 kb, 3.4 kb and 7.5 kb [4]. The open reading frame of mouse Cln8 is 82% identical with CLN8 and at the amino acid level the proteins are 85% identical [4]. In both human and mouse genes the RT-PCR analysis of the CLN8 open reading frame resulted in a single fragment, suggesting that the different size transcripts seen in northern analysis are due to alternative 3' untranslated regions as previously suggested for the human CLN8 gene [4]. It is, however, also possible that the CLN8/Cln8 genes have several transcription initiation sites. Real-time quantitative RT-PCR analysis further suggested that the expression levels of Cln8 are relatively low in both adult and embryonic tissues, and showed that expression levels in various tissues do not differ dramatically. The remarkably low expression of Cln8 in testis could reflect a high level of the control gene analyzed in this tissue. Despite ubiquitous expression of the underlying gene and accumulation of storage material in all tissues, the most severe damage in EPMR patients and mnd mice is in neuronal cells. This is also characteristic of other NCLs, suggesting that NCL proteins may have a specific function in neuronal cells distinct from the role in non-neuronal cells, or that certain neuronal populations may be more sensitive to the disturbed function of NCL proteins.
In situ hybridization analysis of developing mice indicates that Cln8 mRNA is expressed throughout development as well as in the mature brain. The expression of Cln8 in embryonic stages E13 and E17 was specific but low in many tissues including brain, but remarkably high in the developing gastrointestinal tract. This may indicate a role for Cln8 in the innervation of the developing gut. In addition, in the E17 embryo Cln8 expression was high in DRG neurons. These neurons determine the connections between interneurons and motoneurons of the spinal cord. Assuming a role for Cln8 in this process, its disturbed function might result in defective connections in DRG with consecutive dysfunction of motor neurons resulting in progressive paralysis in the mnd mouse model [5,17]. The expression of Cln8 in postnatal mouse brain was highest at developmental stages P5 and P10, especially in hippocampal region CA3. The brain expression was clearly lower in adult mice than at P0, P5 and P10, suggesting a role for Cln8 in brain maturation. This early postnatal period is characterized by a rapid increase in the number of synapses, as well as maturation of oligodendrocytes, the myelin-forming cells in the CNS [26]. Two other NCL proteins, Ppt1 and Cln3, have been localized to synaptic areas of neurons [27,28] suggesting the importance of these NCL proteins for synaptic functions. However, detailed study of the function of Cln8 in the CNS and different brain cells is currently hampered by the lack of a specific antibody detecting endogenous Cln8. Very low levels of Cln8 mRNA were also detected in developing and adult motoneurons and retina (data not shown). However, the Cln8 transcript levels in these samples were virtually indistinguishable from the respective sense controls.
The expression of several other NCL genes in mouse and/or rat has been characterized. In addition to Cln8, Cln1 and Cln5 also show developmental regulation, whereas the expression of two other NCL genes, cathepsin D underlying a naturally occuring sheep NCL model and Cln2 underlying a human late infantile-onset NCL form, is relatively constant in the rat brain during development [29-32]. Mouse Cln3 is expressed in adult mice throughout the brain, and in the hippocampus the most prominent expression is detected in the granular cell layer of the dentate gyrus and pyramidal cells [28]. In addition, human CLN1 and CLN5 genes are expressed at the beginning of neurogenesis in embryonic human brains [33]. This expression increases as brain development proceeds. Expression of NCL genes, including Cln8, in the developing and mature brain indicates that NCL proteins may have roles not only in supporting the survival of neurons but also in the maturation and differentiation of different neuronal populations during development. Defects in maturation processes could potentially lead to neuronal degeneration and loss of neurons.
In the kindling model of experimental epilepsy a rapid up-regulation of Cln8 expression was detected in the brain, especially in hippocampal regions CA3 and the granular cell layer of the dentate gyrus. The expression of Cln8 increased at every time point measured, and was highest 24 h after kindling. Whether single or focal seizures alone are sufficient to induce Cln8 expression requires further investigation. While only Cln8 expression has been investigated in the kindling model, expression of cathepsin D and the Cln1 encoded palmitoyl protein thioesterase 1 (Ppt1) protein has been studied using kainic acid-induced seizures in rats as a model. Both show increased expression in this model, cathepsin D in the hippocampus, limbic cortex and temporo-parieto-occipital neocortex [34] and the Ppt1 protein most prominently in pyramidal cells of hippocampal regions CA1 and CA3 [35]. Although different forms of NCLs are genetically heterogenous, they share a common hippocampal pathology, which is distinct from lesions caused by anoxic-ischemic events and is instead suggested to be a consequence of primary metabolic defects [36]. Increased plasma glutamate levels, decreased glutamate uptake, descreased glutamate transporters as well as changes in ionotropic glutamate receptors has been described in mnd mice [37,38]. Also, prominent loss of GABAergic interneurons in mnd as well as in the Cln1 and Cln3 knock-out mice has been reported [39-41]. Alterations in both glutamatergic and GABAergic neurotransmission may both contribute to chronic excitotoxicity, hypothesized to underlie the cellular dysfunction and brain pathology in NCL disorders [42].
Conclusion
The Cln8 gene is widely expressed in embryonic and adult mouse tissues. In line with this, and comparable with other NCL disorders, accumulation of storage deposits in CLN8-associated disorders occurs in all tissues. However, as the most striking accumulation and cellular dysfunction is limited to neurons, it is likely that CLN8/Cln8 is involved in a neuron-specific biochemical pathway or that neuronal cells are more sensitive to accumulating material or lack compensatory pathways.
The Cln8 gene is expressed throughout brain development and also in mature brain, suggesting a role for Cln8 in maturation and differentiation of neurons and in supporting the survival of neurons. The expression of Cln8 showed regional differences, suggesting a specific function for Cln8 in specific neuronal populations. The relevance of Cln8 up-regulation in hippocampal neurons of kindled mice needs to be studied further.
Methods
Northern blot analysis
FirstChoice™ Northern Blot Mouse Blot I (Ambion, Austin, TX, USA) was used to analyze the distribution of Cln8 expression in various tissues. An 872 base pair (bp) deoxyribonucleic acid (DNA) fragment including a Cln8 open reading frame of 867 bp, digested from the SvPoly-Cln8 expression vector with restriction enzyme BamHI, was used as a probe. The hybridization was performed according to the manufacturer's instructions with Ultra sensitive Hybridization buffer (Ambion), and salmon sperm DNA and Cot-1 DNA for blocking.
Real-time quantitative RT-PCR analysis of Cln8
Real-time quantitative RT-PCR analysis of Cln8 was performed in an ABI PRISM® 7000 SDS thermal cycler (Applied Biosystems, Foster City, CA, USA) using 5 μl of each tissue from Mouse Multiple Tissue cDNA (MTC™) Panel II (BD Biosciences) as template. Each 25 μl reaction contained 300 nM of reverse primer 5'-CCACTGGTTGGCCTTCCA-3' and 900 nM of forward primer 5'-GCCCTTCACCTGCATTTCC-3' as well as 200 nM of Cln8 specific probe 6FAM-TGACCACCCAGCCTTCAGGAGCAT-TAMRA in TaqMan® Universal PCR Master Mix (Applied Biosystems). The primers were optimized for analysis and the PCR fragment of 76 bp covered the region from 510 to 585 bp of the Cln8 open reading frame (GenBank sequence accession no AF125307). The Cln8 specific probe included a reporter dye 6-carboxyfluorescein (FAM) and a quencher dye 6-carboxy-tetramethyl-rhodamine (TAMRA). Endogenous control reactions were performed using Assays-on-Demand Mm00446973_m1 TATA-box binding protein amplification according to the manufacturer's instructions (Applied Biosystems). Each reaction was performed in triplicate. Standard complementary DNA (cDNA) solutions, generated from C57/BL mice cortex mRNA, and standard curve method were used (User Bulletin #2: ABI PRISM(R) 7700 Sequence Detection System. P/N 4303859B (2001) Applied Biosystems. 36 s.). Statistical analyses were performed using Microsoft Excel.
The DNA fragment covering the whole open reading frame of Cln8 was PCR amplified using primers 3'-GTGATTTCTCCGGTGCTAGG-5' and 3'-GCAACCACCATTTCTCAGGT-5'.
In situ hybridization
Paraffin and cryosections of NMRI or C57/BL mice were prepared and analyzed by mRNA in situ hybridization as previously described [43-45] with slight modifications. The isolated tissues from postnatal wild-type and kindling mice were fresh-frozen and sectioned. Embryos were either fixed in 4% paraformaldehyde and embedded in paraffin or fresh-frozen and sectioned. For all serial sections of 8–14 μm, a single 35S-labeled 370 bp Cln8 specific complementary RNA (cRNA) probe (from bp 29 to 398; GeneBank AF125307) was used. For the generation of the antisense and control sense cRNAs, the vector (pBluescript vector; Stratagene, La Jolla, CA, USA) containing the PCR-amplified Cln8 fragment was linearized and used as a template for RNA polymerases T3 and T7, respectively, using α-35S-labeled UTP (Amersham Biosciences, Uppsala, Sweden).
The paraffin sections were treated as cryosections, after first being deparaffinized in xylene and rehydrated in a descending ethanol series. All the slides were fixed in 4% paraformaldehyde, rinsed twice in phosphate-buffered saline and treated with proteinase K (20 μg/ml for paraffin sections, 1 μg/ml for cryosections). Next, the slides were postfixed in 4% paraformaldehyde, and the cryosections rinsed in 50% deionized formamide in 2 × standard sodium citrate (SSC). After being washed in distilled water, all slides were acetylated for 10 minutes with 0.25% acetic anhydride in 0.1 M triethanolamine. The paraffin sections were then dehydrated in an ascending ethanol series, while the cryosections were rinsed again with 50% formamide in 2 × SSC. Prehybridization for 1 h was subsequently performed at +52°C with hybridization buffer containing 60% formamide, 10% dextran sulphate, 1 × Denhardt's solution, 0.5 mg/ml yeast transfer RNA (tRNA), and 100 mM dithiothreitol. After prehybridization, cRNA probes were added to the hybridization buffer at a concentration of 35 000 cpm/μl, and hybridization was carried out overnight at +52°C in a chamber humidified with 60% formamide and 5 × SSC.
After hybridization the sections were first washed in 5 × SSC, 10 mM dithiothreitol at +50°C for 30 minutes, followed by 30 minutes in 50% formamide, 2 × SSC, 30 mM dithiothreitol at +65°C, and 3 × 10 minutes in NTE-buffer (0.5 M NaCl, 10 mM Tris-HCl, pH 8.0, 5 mM EDTA) at +37°C. Next, the slides were treated with RNase A (20 μg/ml) in NTE-buffer at +37°C for 30 min and washed for 15 min in NTE-buffer at +37°C. Washing with 50% formamide, 2 × SSC, 30 mM dithiothreitol was then repeated, after which the sections were washed in 2 × SSC and 0.1 × SSC, each for 15 min at +37°C. The slides were dehydrated at room temperature with ethanol containing 0.3 M ammonium acetate, air-dried, and exposed to BioMax MR Film (Kodak) together with 14C radioactive standards (Amersham Biosciences) for 5 days. Finally, the sections were dipped in NTB-2 emulsion (Kodak), developed after 3 to 4 weeks using D-19 developer (Kodak), fixed with Kodak Unifix, counterstained with hematoxylin (Shandon Inc./Thermoshandon, Pittsburgh, PA, USA), and mounted with Permount (Fisher Scientific International Inc., Hampton, NH, USA). Mice were bred in the Viikki Laboratory Animal Center, University of Helsinki. The experiments were conducted according to the "European Convention for the Protection of Vertebrate Animals used for Experimental and other Scientific purposes" with the approval of the institutional ethics committee.
Rapid kindling procedure
Adult male C57/BL mice weighing 22–25 g (n = 16) were anaesthetized with sodium pentobarbital (60 mg/kg i.p.), and bipolar stainless steel electrodes (Plastics One) were implanted in the left ventral hippocampus using a Kopf stereotaxic frame as previously described [46]. One week later, the mice were electrically stimulated with forty threshold stimulations with 5 min intervals (10 Hz frequency, 1 ms square-wave pulses for 10 s). The electroencephalogram was continuously recorded during the whole rapid kindling procedure using the MacLab/4e system (ADInstruments). To determine the threshold, current intensity was increased stepwise (10 μA increments, starting from 10 μA) every 5 min until focal epileptiform activity (afterdischarge) of at least 5 s duration was elicited. Rapid kindling stimulations were then given on the same day. Behavioral convulsions were scored according to a modification [25] of the scale of Racine [47]: grade 0, arrest, normal behavior; grade 1, facial twitches; grade 2, chewing and nodding; grade 3, forelimb clonus; grade 4, rearing, body jerks, tail upholding; and grade 5, imbalance, hind-limb clonus and vocalization. At 2, 6 and 24 h after the last stimulus-evoked seizure, animals were decapitated (four animals in each group) and brains were processed for in situ hybridization. Four non-stimulated, electrode-implanted mice served as controls and were killed together with the experimental animals from different groups. The correct localization of the electrode was controlled for each animal in the coronal sections of the brain used for in situ hybridization. All experimental procedures were approved by the Research Ethics Committee at the Medical Faculty of the University of Lund.
Quantitative and statistical analysis
Quantification of radioactive in situ hybridization films was done by digitizing the film images with a computer-based MCID image analysis system (Imaging Research). Gray levels from 14C radioactive standards were used in a third-degree polynomial calibration to obtain equivalent values of tissue radioactivity (nCi/g) for different brain regions. Measurements for each structure were carried out in several sections per animal. Since there were no significant differences between the two sides, values were pooled to obtain the mean value for each animal and brain region. Data from four control and four experimental animals for every probe and time point were used for analysis. Statistical analysis was performed using Student's t-tests. In the figures, the densities of hybridization signals of sense and antisense probes are shown separately. Sense values were not subtracted from antisense values because of different probes used. Error bars represent standard error of the mean; symbols *, ** and *** represent p < 0.05, 0.01 and 0.001, respectively.
Authors' contributions
LL carried out real-time quantitative PCR and northern blot analyses. In addition, LL was involved in the analysis of mRNA in situ results and design of the study, drafted the manuscript and made the figures. AA carried out the mRNA in situ analysis and was involved in drafting the manuscript. OK prepared several mouse tissue samples for the mRNA in situ analysis, was involved in the analysis of the mRNA in situ results, and revised the manuscript. MK established the real-time quantitative RT-PCR methodology and helped with the analysis. ZK provided all kindling samples and contributed significantly to figures 3 and 4. mRNA in situ analyses were conducted under the supervision of MS, who also revised the manuscript and was involved in the design of the study. AEL was involved in the design of the study, analysis of the results and revision of the manuscript.
Acknowledgements
We thank Teija-Tuulia Toivonen and Eila Kujamäki for technical assistance as well as Tarja Salonen and Tarja Joensuu for advice and support. Susanna Ranta is thanked for cloning and testing constructs for the mRNA in situ hybridization analysis. Pertti Panula and Oleg Anichtch are thanked for help in the quantification analysis of mRNA in situ hybridization and Matthew Phillips and Jodie Painter for critical reading of the manuscript. This work has been supported by the Academy of Finland Centre of Excellence in Disease Genetics, project 44870 Finnish Centre of Excellence programme 2000–2005, Sigrid Jusélius Foundation, Folkhälsan Research Foundation and Swedish Research Council. The Lund Stem Cell Center is supported by a Center of Excellence grant in life sciences from the Swedish Foundation for Strategic Research. Liina Lonka is a student at the Helsinki Graduate School in Biotechnology and Molecular Biology, University of Helsinki.
Figures and Tables
Figure 1 Northern blot and RT-PCR analysis of Cln8 mRNA expression in mouse tissues A: Northern blot analysis of Cln8 transcripts in mouse tissues. A β-actin probe was used as a control. Molecular weight marker in kb is shown on the left. B: Real-time quantitative RT-PCR analysis of Cln8 in mouse tissues. The expression of Cln8 in brain, given a value of 1, was used as a control. The expression of Cln8 in the other tissues is shown as a fold of the control.
Figure 2 In situ hybridization analysis of Cln8 mRNA expression in mouse embryos. In the E17 mouse embryo, developing gastrointestinal tract (A, B) and DRG (C, D), hybridized with Cln8 specific cRNA probe, are shown. Bright-field view of the cells is shown on the left (A, C) and dark-field detection of hybridization signals on the right (B,D). Optical density values of hybridization signals of Cln8 sense and antisense probes in E13 and E17 gut and brain tissues are shown in the diagram (E). Error bars represent standard error of the mean. s = sense, as = antisense
Figure 3 Differential distribution of Cln8 mRNA in brains of P0, P5, P10 and adult mouse. Dark-field emulsion autoradiographs from frontal sections, shown on the left, were analyzed by in situ hybridization analysis (A-D). Higher magnification in the hippocampal area CA3 is shown on the right. Bright-field views of the cells (a-d) and dark-field emulsion autoradiographs (a'-d') are shown. Optical density values of hybridization signals of Cln8 antisense probe are shown as bars in the diagram below (A = adult). Optical density values of hybridization signals of Cln8 sense probe are shown as a gray area behind the bars. In each brain region the optical density values at P0, P5 and P10 were compared to the optical density value of adult. Error bars represent standard error of the mean. Symbols *, ** and *** represent p < 0.05, 0.01 and 0.001, respectively. CX = cortex, DG = dentate gyrus
Figure 4 Distribution of Cln8 mRNA in brains of hippocampal kindling induced epileptic mice. Frontal mouse brain sections were analyzed 2 h, 6 h and 24 h after kindling induced epileptic seizures by in situ hybridization. Mice with electrodes implanted but without electrical stimulations were used as controls (0 h). Dark-field emulsion autoradiographs of hippocampal Cln8 expression in kindling-induced mice 0 h and 24 h after kindling are shown. Optical density values of hybridization signals of Cln8 antisense probe are shown as bars in the diagram. Optical density values of hybridization signals of Cln8 sense probe are shown as a gray area behind the bars. In each brain region the optical density values 2 h, 6 h and 24 h after kidling-induced seizures were compared to controls (0 h). Error bars represent standard error of the mean. Symbols *, ** and *** represent p < 0.05, 0.01 and 0.001, respectively. DG = dentate gyrus
==== Refs
Santavuori P Neuronal ceroid-lipofuscinoses in childhood Brain Dev 1988 10 80 83 3291628
Haltia M The neuronal ceroid-lipofuscinoses J Neuropathol Exp Neurol 2003 62 1 13 12528813
Bronson RT Donahue LR Johnson KR Tanner A Lane PW Faust JR Neuronal ceroid lipofuscinosis (nclf), a new disorder of the mouse linked to chromosome 9 Am J Med Genet 1998 77 289 297 9600738 10.1002/(SICI)1096-8628(19980526)77:4<289::AID-AJMG8>3.0.CO;2-I
Ranta S Zhang Y Ross B Lonka L Takkunen E Messer A Sharp J Wheeler R Kusumi K Mole S Liu W Soares MB Bonaldo MF Hirvasniemi A de la Chapelle A Gilliam TC Lehesjoki AE The neuronal ceroid lipofuscinoses in human EPMR and mnd mutant mice are associated with mutations in CLN8 Nat Genet 1999 23 233 236 10508524 10.1038/13868
Messer A Flaherty L Autosomal dominance in a late-onset motor neuron disease in the mouse J Neurogenet 1986 3 345 355 3783318
Gupta P Soyombo AA Atashband A Wisniewski KE Shelton JM Richardson JA Hammer RE Hofmann SL Disruption of PPT1 or PPT2 causes neuronal ceroid lipofuscinosis in knockout mice Proc Natl Acad Sci U S A 2001 98 13566 13571 11717424 10.1073/pnas.251485198
Katz ML Shibuya H Liu PC Kaur S Gao CL Johnson GS A mouse gene knockout model for juvenile ceroid-lipofuscinosis (Batten disease) J Neurosci Res 1999 57 551 556 10440905 10.1002/(SICI)1097-4547(19990815)57:4<551::AID-JNR15>3.0.CO;2-R
Mitchison HM Bernard DJ Greene ND Cooper JD Junaid MA Pullarkat RK de Vos N Breuning MH Owens JW Mobley WC Gardiner RM Lake BD Taschner PE Nussbaum RL Targeted disruption of the Cln3 gene provides a mouse model for Batten disease. The Batten Mouse Model Consortium [corrected] Neurobiol Dis 1999 6 321 334 10527801 10.1006/nbdi.1999.0267
Cotman SL Vrbanac V Lebel LA Lee RL Johnson KA Donahue LR Teed AM Antonellis K Bronson RT Lerner TJ MacDonald ME Cln3(Deltaex7/8) knock-in mice with the common JNCL mutation exhibit progressive neurologic disease that begins before birth Hum Mol Genet 2002 11 2709 2721 12374761 10.1093/hmg/11.22.2709
Koike M Nakanishi H Saftig P Ezaki J Isahara K Ohsawa Y Schulz-Schaeffer W Watanabe T Waguri S Kametaka S Shibata M Yamamoto K Kominami E Peters C von Figura K Uchiyama Y Cathepsin D deficiency induces lysosomal storage with ceroid lipofuscin in mouse CNS neurons J Neurosci 2000 20 6898 6906 10995834
Lonka L Kyttälä A Ranta S Jalanko A Lehesjoki AE The neuronal ceroid lipofuscinosis CLN8 membrane protein is a resident of the endoplasmic reticulum Hum Mol Genet 2000 9 1691 1697 10861296 10.1093/hmg/9.11.1691
Lonka L Salonen T Siintola E Kopra O Lehesjoki AE Jalanko A Localization of wild-type and mutant neuronal ceroid lipofuscinosis CLN8 proteins in non-neuronal and neuronal cells J Neurosci Res 2004 76 862 871 15160397 10.1002/jnr.20133
Ranta S Topcu M Tegelberg S Tan H Ustubutun A Saatci I Dufke A Enders H Pohl K Alembik Y Mitchell WA Mole SE Lehesjoki AE Variant late infantile neuronal ceroid lipofuscinosis in a subset of Turkish patients is allelic to Northern epilepsy Hum Mutat 2004 23 300 305 15024724 10.1002/humu.20018
Hirvasniemi A Lang H Lehesjoki AE Leisti J Northern epilepsy syndrome: an inherited childhood onset epilepsy with associated mental deterioration J Med Genet 1994 31 177 182 8014963
Topcu M Tan H Yalnizoglu D Usubutun A Saatci I Aynaci M Anlar B Topaloglu H Turanli G Kose G Aysun S Evaluation of 36 patients from Turkey with neuronal ceroid lipofuscinosis: clinical, neurophysiological, neuroradiological and histopathologic studies Turk J Pediatr 2004 46 1 10 15074367
Herva R Tyynelä J Hirvasniemi A Syrjakallio-Ylitalo M Haltia M Northern epilepsy: a novel form of neuronal ceroid-lipofuscinosis Brain Pathol 2000 10 215 222 10764041
Messer A Strominger NL Mazurkiewicz JE Histopathology of the late-onset motor neuron degeneration (Mnd) mutant in the mouse J Neurogenet 1987 4 201 213 3612378
Messer A Plummer J Wong V Lavail MM Retinal degeneration in motor neuron degeneration (mnd) mutant mice Exp Eye Res 1993 57 637 641 8282051 10.1006/exer.1993.1169
Messer A Plummer J Accumulating autofluorescent material as a marker for early changes in the spinal cord of the Mnd mouse Neuromuscul Disord 1993 3 129 134 8358238 10.1016/0960-8966(93)90004-4
Bronson RT Lake BD Cook S Taylor S Davisson MT Motor neuron degeneration of mice is a model of neuronal ceroid lipofuscinosis (Batten's disease) Ann Neurol 1993 33 381 385 7683855 10.1002/ana.410330408
Callahan LM Wylen EL Messer A Mazurkiewicz JE Neurofilament distribution is altered in the Mnd (motor neuron degeneration) mouse J Neuropathol Exp Neurol 1991 50 491 504 2061715
Mazurkiewicz JE Callahan LM Swash M Martin JE Messer A Cytoplasmic inclusions in spinal neurons of the motor neuron degeneration (Mnd) mouse. I. Light microscopic analysis J Neurol Sci 1993 116 59 66 8389815
Goddard GV McIntyre DC Leech CK A permanent change in brain function resulting from daily electrical stimulation Exp Neurol 1969 25 295 330 4981856 10.1016/0014-4886(69)90128-9
Dennison Z Teskey GC Cain DP Persistence of kindling: effect of partial kindling, retention interval, kindling site, and stimulation parameters Epilepsy Res 1995 21 171 182 8536671 10.1016/0920-1211(95)00025-6
Nanobashvili A Kokaia Z Lindvall O Generalization of rapidly recurring seizures is suppressed in mice lacking glial cell line-derived neurotrophic factor family receptor alpha2 Neuroscience 2003 118 845 852 12710991 10.1016/S0306-4522(02)00998-3
Jacobson M Developmental Neurobiology 1991 New York and London: Plenum Press
Lehtovirta M Kyttälä A Eskelinen EL Hess M Heinonen O Jalanko A Palmitoyl protein thioesterase (PPT) localizes into synaptosomes and synaptic vesicles in neurons: implications for infantile neuronal ceroid lipofuscinosis (INCL) Hum Mol Genet 2001 10 69 75 11136716 10.1093/hmg/10.1.69
Luiro K Kopra O Lehtovirta M Jalanko A CLN3 protein is targeted to neuronal synapses but excluded from synaptic vesicles: new clues to Batten disease Hum Mol Genet 2001 10 2123 2131 11590129 10.1093/hmg/10.19.2123
Suopanki J Tyynelä J Baumann M Haltia M The expression of palmitoyl-protein thioesterase is developmentally regulated in neural tissues but not in nonneural tissues Mol Genet Metab 1999 66 290 293 10191117 10.1006/mgme.1999.2839
Suopanki J Tyynela J Baumann M Haltia M Palmitoyl-protein thioesterase, an enzyme implicated in neurodegeneration, is localized in neurons and is developmentally regulated in rat brain Neurosci Lett 1999 265 53 56 10327204 10.1016/S0304-3940(99)00207-4
Isosomppi J Heinonen O Hiltunen JO Greene ND Vesa J Uusitalo A Mitchison HM Saarma M Jalanko A Peltonen L Developmental expression of palmitoyl protein thioesterase in normal mice Brain Res Dev Brain Res 1999 118 1 11 10611498 10.1016/S0165-3806(99)00115-7
Tyynelä J Sohar I Sleat DE Gin RM Donnelly RJ Baumann M Haltia M Lobel P A mutation in the ovine cathepsin D gene causes a congenital lysosomal storage disease with profound neurodegeneration Embo J 2000 19 2786 2792 10856224 10.1093/emboj/19.12.2786
Heinonen O Salonen T Jalanko A Peltonen L Copp A CLN-1 and CLN-5, genes for infantile and variant late infantile neuronal ceroid lipofuscinoses, are expressed in the embryonic human brain J Comp Neurol 2000 426 406 412 10992246 10.1002/1096-9861(20001023)426:3<406::AID-CNE5>3.0.CO;2-5
Hetman M Filipkowski RK Domagala W Kaczmarek L Elevated cathepsin D expression in kainate-evoked rat brain neurodegeneration Exp Neurol 1995 136 53 63 7589334 10.1006/exnr.1995.1083
Suopanki J Lintunen M Lahtinen H Haltia M Panula P Baumann M Tyynelä J Status epilepticus induces changes in the expression and localization of endogenous palmitoyl-protein thioesterase 1 Neurobiol Dis 2002 10 247 257 12270687 10.1006/nbdi.2002.0503
Haltia M Herva R Suopanki J Baumann M Tyynelä J Hippocampal lesions in the neuronal ceroid lipofuscinoses Eur J Paediatr Neurol 2001 5 209 211 11588999
Mennini T Bastone A Crespi D Comoletti D Manzoni C Spinal cord GLT-1 glutamate transporter and blood glutamic acid alterations in motor neuron degeneration (Mnd) mice J Neurol Sci 1998 157 31 36 9600674 10.1016/S0022-510X(98)00072-0
Mennini T Bigini P Ravizza T Vezzani A Calvaresi N Tortarolo M Bendotti C Expression of glutamate receptor subtypes in the spinal cord of control and mnd mice, a model of motor neuron disorder J Neurosci Res 2002 70 553 560 12404509 10.1002/jnr.10420
Cooper JD Messer A Feng AK Chua-Couzens J Mobley WC Apparent loss and hypertrophy of interneurons in a mouse model of neuronal ceroid lipofuscinosis: evidence for partial response to insulin-like growth factor-1 treatment J Neurosci 1999 19 2556 2567 10087069
Bible E Gupta P Hofmann SL Cooper JD Regional and cellular neuropathology in the palmitoyl protein thioesterase-1 null mutant mouse model of infantile neuronal ceroid lipofuscinosis Neurobiol Dis 2004 16 346 359 15193291 10.1016/j.nbd.2004.02.010
Cooper JD Progress towards understanding the neurobiology of Batten disease or neuronal ceroid lipofuscinosis Curr Opin Neurol 2003 16 121 128 12644737 10.1097/00019052-200304000-00001
Das AM Regulation of the mitochondrial ATP-synthase in health and disease Mol Genet Metab 2003 79 71 82 12809636 10.1016/S1096-7192(03)00069-6
Hiltunen JO Arumae U Moshnyakov M Saarma M Expression of mRNAs for neurotrophins and their receptors in developing rat heart Circ Res 1996 79 930 939 8888685
Kokaia Z Airaksinen MS Nanobashvili A Larsson E Kujamaki E Lindvall O Saarma M GDNF family ligands and receptors are differentially regulated after brain insults in the rat Eur J Neurosci 1999 11 1202 1216 10103116 10.1046/j.1460-9568.1999.00513.x
Lauren J Airaksinen MS Saarma M Timmusk T Two novel mammalian Nogo receptor homologs differentially expressed in the central and peripheral nervous systems Mol Cell Neurosci 2003 24 581 594 14664809 10.1016/S1044-7431(03)00199-4
Nanobashvili A Airaksinen MS Kokaia M Rossi J Asztely F Olofsdotter K Mohapel P Saarma M Lindvall O Kokaia Z Development and persistence of kindling epilepsy are impaired in mice lacking glial cell line-derived neurotrophic factor family receptor alpha 2 Proc Natl Acad Sci U S A 2000 97 12312 12317 11050250 10.1073/pnas.97.22.12312
Racine RJ Modification of seizure activity by electrical stimulation. II. Motor seizure Electroencephalogr Clin Neurophysiol 1972 32 281 294 4110397 10.1016/0013-4694(72)90177-0
| 15826318 | PMC1087490 | CC BY | 2021-01-04 16:03:47 | no | BMC Neurosci. 2005 Apr 13; 6:27 | utf-8 | BMC Neurosci | 2,005 | 10.1186/1471-2202-6-27 | oa_comm |
==== Front
BMC OphthalmolBMC Ophthalmology1471-2415BioMed Central London 1471-2415-5-91581712510.1186/1471-2415-5-9Research ArticleA freely accessible, evidence based, objective system of analysis of posterior capsular opacification ; Evidence for its validity and reliability Aslam Tariq M [email protected] Niall [email protected] Jim [email protected] Manchester Eye Hospital, UK2 Sir Charles Gairdner Hospital, Perth, Australia3 Dept. Imaging Sciences and Biomedical Engineering, Manchester University, UK2005 7 4 2005 5 9 9 13 11 2004 7 4 2005 Copyright © 2005 Aslam 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 this study was to develop a system of computerised analysis of digital images of posterior capsule opacification (PCO) that is evidence based, objective and freely available. The paper will present evidence for the reliability and validity of the developed system.
Methods
The system of PCO analysis was developed considering current published evidence on visual significance of PCO and additional investigative analysis of PCO images. Details of the image processing and analysis steps are discussed and a final system that measures an entropy score weighted toward proximity to central areas is described. In order to assess validity, the systems ability to measure PCO progression is assessed along with the visual significance of its final computerised scores. Reliability of the system is also assessed.
Results
The final system runs successfully and is simple to use. Analyses of PCO by the system show an ability to detect early progression of PCO as well as detection of visually significant PCO. Images with no clinical PCO produce very low scores in the analysis. Reliability of the system of analysis is shown to be satisfactory.
Conclusion
This paper presents a system of PCO analysis that is evidence based, objective and clinically useful. Substantial evidence is provided for its validity and reliability.
==== Body
Background
Despite advances in the practice of cataract surgery and intraocular lens implantation, posterior capsule opacification remains the most common post-operative cause of morbidity[1]. However, there is currently no consensus on an optimal PCO quantification method. The main objective systems of analysis, POCO[2] and AQUA[3] systems are not freely available. They do not incorporate whether PCO is peripheral or central into calculations and show limited evidence for validity. The EPCO system [4] has been assessed for evidence of construct validity [5] but is still subjective. The POCO system[6] is also subjective and is not convincing for analysis of PCO in terms of measuring progression or visual significance. There is clearly a need for a universally acceptable measure of PCO [7] that would be objective enough for scientific analysis and yet not exclude the majority of researchers by its lack of free availability. It should be based upon current evidence on the visual significance of PCO and be tested to be both valid and reliable. This paper presents the first such system of analysis.
A principal problem in any system of analysis of digital images of PCO is that of artefactual light reflections spoiling the image. Although resilient to such artefactual corruption, the presented system is designed for use on images that have had some prior mechanism of reducing unwanted reflections. [4,7] For this study all testing and development was done on images created to be free of reflections by merging of two or more original images using commercially available and common software. [5] This step has been previously demonstrated as valid and reliable [5]
The presented program is not proposed as the ultimate measurement system for PCO. We would expect to continuously update the system with future research findings as well as data on analyses. These data may be used to refine certain parameters and factors used in calculations and algorithms.
The aim of this study was to develop a system of computerised analysis of digital images of PCO that is evidence based and objective. Evidence for validity and reliability of the final system will be presented.
Methods
Development of software
This system was designed and programmed using the MatLab programming platform(MatLabTM, Matrix Laboratory,© The Mathworks, inc, MA). All programming design and writing of code was by T M Aslam and for clarity the system is hence referred to as the Aslam analysis (AA) in this paper. All images used in development of the system were of a different subset to those used in assessment of validity and reliability of the final analysis system. Informed consent was obtained from all patients involved and the study was approved by local ethics committee, in accordance with declaration of Helsinki.
The first problem encountered was one of unequal illumination in the images. Even with the large areas of aberrant light reflections removed, using merging with similar but unspoilt images, a generally variable background illumination of the entire image may cause errors in image analysis. This is solved in the AA system with an image processing step of background illumination removal. This is achieved through image manipulation techniques known as morphological opening[8] to estimate the background illumination. The resulting foreground PCO is freed of illumination variations. (fig 1b, 2b)
Next, unwanted outer segments of images, that include, for example, iris segments, must be excluded from the analysis. The desired area of interest involving the optic of the lens or area within the pupil is isolated in the AA system by the user marking out the corners of the region of interest. These primary steps lead to an isolated region of interest that is free from variations in illumination and ready for analysis. For this study, the central 3.5 mm of posterior capsule, centred using pupil borders, was used for all analyses. The dimensions of this area was noted in terms of pixel size using known diameter of intra-ocular optic in the images as a guide. This segment of PCO has been previously confirmed as being visually significant. [5]
It is established that areas of PCO distant from the centre of the visual axis have a reduced effect on vision compared to areas at the centre.[5] Although the exact mathematical relationship of this is unknown the AA system incorporates the principle of this research finding in its analysis. To do this the region to be analysed was split into multiple smaller subunits and each unit considered separately for analysis. Modifications to the analysis results of each subunit were then made depending on the distance of that subunit from the central visual axis. The further a unit was from central visual axis, the more the value of its texture would be attenuated. Finally texture scores for all subunits is summed to give a final total score. The factor used to modify PCO score according to distance needed to be large enough to exert the required effect of weighting scores towards the centre of the image, but not so large as to render the system insensitive to small amounts of peripheral PCO. Similarly, subunit size was adjusted until it was found to be large enough to incorporate significant PCO textural objects but small enough for a satisfactory total number of subunit regions to be analysed to allow for central weighting mechanism to function. Various combinations of subregion size and image score weighting were trialled before a compromise was achieved. This involved dividing each image into 121 total subunits.
Development of the PCO score itself involved considering many possible image processing and analytical techniques. Current evidence on PCO suggests that its texture, determined from retro-illuminated digital images, could be used to determine its visual significance.[2,9] Histogram analysis[10] suggested that texture analysis of the PCO might be feasible using statistical properties of the intensity histogram.(Fig 1c, 2c) Intensity levels of trial images with significant PCO showed distribution with distinct peaks at varying levels of intensity. The histograms all showed low dynamic range with narrow width compared to the entire grey scale. During development of the system, various descriptors of texture based on the intensity histogram of a region were calculated for potential use in analysis of PCO images of patients.[10] Entropy was found to provide the most useful form of texture measure. This corresponds with the fact that entropy tends to be higher in coarser textures, such as is found in pearl type opacification, that is proposed to be especially visually significant.[11] Indeed, such pearl type opacification areas showed more irregular and random peaks in histograms studied. In investigations detailed below, only the entropy measure showed consistent features of validity, and this measure was chosen as the most suitable marker for the analysis system.
Even after the above processing steps, images with no visible PCO and in which PCO would not be expected still showed on histogram analysis to have image areas with low but significant intensities of up to 15 units of brightness that contributed to elevated entropy scores. This was due to variations in intensity unrelated to significant PCO such as minimal persistent reflections, lens imperfections, photographic noise [12] and thin films of material developed over the lens. When the AA system was programmed to except this subset of values it appeared to have much better face validity for analysing small or no amounts of PCO, whilst maintaining face validity for analysis of large amounts of PCO in patients attending for capsulotomy.
Although there was apparent face validity and content validity of the final developed system, it was submitted for further testing before any confidence was held in results of its analyses. Experiments that provide evidence for such confidence in the validity and reliability of the AA system are now discussed.
Testing for validity
Measurement of progression by AA system
In order to test the system's ability to measure progression, digital images of 12 patients within a month of cataract extraction were analysed along with images taken of the same patients of 12–18 months after cataract surgery. On inspection of these images many of the patients had evidently undergone PCO progression from no or nearly no PCO to visible but mainly peripheral PCO. One would expect a model system to produce values reflecting very low levels of PCO in patients within a month of surgery, and for those values to be increased by 18 months after cataract extraction. The results for the AA system were graphically plotted and paired t-tests done to assess whether findings were significantly different in the two groups.
Visual relevance of AA analysis
Evidence for both construct validity (agreement with theoretical expectations) and convergence validity (relation to associated factors) came from a study involving thirty patients (33 eyes) that were recruited after having been referred for potential Nd: YAG capsulotomy.
On attendance, patients had CS vision tested with Pelli-Robson charts. Patients were dilated and the posterior capsules photographed with imagenet® digital photography system and Topcon® camera system at standardised settings. Images were subsequently stored onto disc. Each patient underwent Nd: YAG laser capsulotomy via a set protocol by one surgeon. Patients returned one week later and had further vision testing. They were again dilated and posterior capsules photographed at standardised settings. Full details of the experiment and image acquisition methods are described in a previous publication.[9] Once a new composite image was created, free of external reflections, it was entered into the AA software. The AA program analysed the images created and calculated a score based on an entropy calculation of texture as described above.
Assessment of visual relevance of the analysis system involved seeking evidence for construct validity and convergent validity. Evidence for construct validity was sought by assessing PCO scores of patients with visually significant PCO. These were patients referred for Nd: YAG capsulotomy. The analysis was repeated after YAG capsulotomy when visually significant PCO is in clinical practice significantly reduced. The AA system scores should also be expected to calculate significantly lower values. Paired t-test was used to compare the two groups and results plotted graphically.
For evidence of convergent validity, results of PCO analysis were correlated with vision. Specifically, the difference in PCO quantification by the AA system before and after YAG was compared to improvement in both LogMAR visual acuity and contrast sensitivity for each patient using regression analysis. Graphical representation of results was analysed as well as the indices of regression analysis.
Testing for reliability
To assess interobserver reliability of the AA system, two trained observers performed analysis on a sample of 20 images. These images included both pre- and post- Nd:;Yag capsulotomy, as well as a selection of images with no PCO. The ICC (intraclass correlation coefficient) was calculated, as well as the coefficient of repeatability and the coefficient of variation. In addition, a Bland-Altman plot to graphically examine the repeatability was plotted.
Results
Process and use of developed system
In order to use the system, at present users must have access to Matlab, which is commonplace in scientific and academic institutions. The authors are however currently working on a compiled version that would be usable without any other specialist software. There is free access to the system by contacting the authors.
Once loaded, the AA system asks the user to input an image, and through a menu driven mechanism loads the particular image ready for analysis. The image is converted to greyscale and presented on screen. A cross-hair appears on the image surface and the user clicks on the top left of the image's area of interest and then on the bottom right, thus delineating the region that is to be measured. The experimenter may choose this to be the area within a capsulorhexis, area within a pupil's borders or spreading to the extent of visible intraocular lens, depending on research needs. The program then completes the process of image processing and analysis without further input, and is thus highly objective. The processed images are displayed along with final texture measures.
Testing for validity
Progression of PCO over time
Analyses of frequency distribution of contrast sensitivity and of PCO scores show distributions that can be considered normal. PCO scores in 12 patients one month after surgery compared to 12–18 months post surgery is demonstrated in fig. 3. Paired T test showed significant differences in the two sets of values (p = 0.035). It is evident that the AA system is able to differentiate between patients with early PCO and no PCO and also to be sensitive to progression of PCO.
Visual relevance of the analysis system
Construct validity
Analyses of frequency distribution of PCO scores show distributions that can be considered normal. A paired T-test for parametric data was performed for the 21 patients and show a significant difference between AA scores before and after Nd: YAG laser treatment (P < 0.001). Validity evidence by extreme group testing of construct is therefore demonstrated.(fig. 4.)
Convergent validity
Linear regression was used in assessing the impact of the change in the AA score on change in visual function measured after Nd: YAG capsulotomy.
The first dependent variable to be studied was improvement in best-corrected distance LogMAR visual acuity (DLVA). Linear regression analysis showed that the standardised coefficient was .43 with significance of 0.013.(fig. 5a) Thus, improvements in distance vision after Nd: YAG capsulotomies are shown to be strongly correlated to the entropy score for PCO in the AA system.
The second dependent variable to be studied was improvement in contrast sensitivity (CS). Linear regression analysis showed that the standardised coefficient was .39 with significance of 0.02.(fig. 5b) Thus, improvements in contrast sensitivity after Nd:YAG capsulotomies are strongly correlated to the differences in entropy scores for PCO in the AA system. A scatter graph demonstrates this correlation of AA score with contrast sensitivity and visual acuity (fig. 6)
Testing for Reliability
A total of 22 images were analysed once by two observers. The ICC was 0.995 (95% confidence intervals 0.988 to 0.998). Coefficient of Repeatability was 0.559 (ie 95% of repeated measures would be expected to be within this margin).
A scatter plot (Bland-Altman) showed no relationship between the mean value for each image by the two observers against the standard deviation (fig 7), and we were able to calculate the coefficient of variation, as 3.3% ± 6.6%. These values represent excellent reliability.
Discussion
This paper describes a logical evidence-based development of a system of analysis of PCO. More importantly, it has presented evidence for the validity and reliability of the AA system. A mass of different textural image analysis tools exist in imaging science [13], but this study has provided evidence that the use of the statistical analysis of the image histogram to calculate entropy can provide valid information on the clinical significance of PCO and on progression of smaller amounts of PCO in clinical trials. Reliability was high, as expected in such a computerised objective system, only limited by variations in actual areas chosen to be examined by the experimenters when operating the program.
The current system of analysis has evolved through the planning, writing, testing and rewriting of many component programming procedures to overcome many practical and theoretical challenges.
On designing the experiments and system of analysis above it was initially considered that various different types of analysis might be required in order to accurately predict different types of visual function loss. However, regression analysis shows that both contrast sensitivity and visual acuity are related to the entropy of the acquired images. The authors suggest that this function calculated by the specific mechanism of the AA system should be used as the outcome measure for the assessment of PCO for clinical and experimental studies.
Evidence has been presented for this system's validity and reliability. Although it is freely available to the scientific community, users should perform their own validity studies incorporating their specific photography systems and procedures for acquiring images and removal of light reflections. At present the AA system requires a Matlab platform to operate, but a compiled version is in progress that would be usable without any prior software. Also, a system for registration based removal of light images is being tested and developed.
Conclusion
Image analysis with the AA system provides an open access, objective, valid and reliable method of PCO quantification for advancement of research into this common cause for post operative morbidity. We anticipate that with greater use, additional information will lead to even further evidence-based refinements and updates to the system.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
Tariq Aslam devised the system of analysis and did all the software programming. Niall Patton collaborated in testing of the system, in particular the reliability studies and reviewed system assessment.
Jim Graham assessed and contributed to the whole paper from an image processing viewpoint.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Figures and Tables
Figure 1 a. Original greyscale image, b. background removed, c. image histogram
Figure 2 a. Original greyscale image, b. background removed, c. image histogram
Figure 3 Progression in AA scores between patients one month and 12 months after surgery; a. Paired Samples Test, b. Differences graphically displayed
Figure 4 Comparison of AA scores of patients before and after Nd:YAG capsulotomy; a. Statistical analysis, b. Graphically displayed
Figure 5 Improvements in vision compared to improvements in AA score, after capsulotomy; a. Regression of distance vision improvement against improvement in AA score, b. Regression analysis of contrast sensitivity improvement against improvement in AA score
Figure 6 Improvement in contrast sensitivity and distance vision after capsulotomy v improvement in AA scores after capsulotomy
Figure 7 Bland-Altman Plot of interobserver repeatability of AA score between observer 1 and observer 2
==== Refs
Spalton DJ Posterior capsular opacification after cataract surgery Eye 1999 13 489 492 10627830
Barman SA Hollick EJ Boyce JF Spalton DJ Uyyanonvara B Sanguinetti G Meacock W Quantification of posterior capsular opacification in digital images after cataract surgery Invest Ophthalmol Vis Sci 2000 41 3882 3892 11053290
Findl O Buehl W Menapace R Georgopoulos M Rainer G Siegl H Kaider A Pinz A Comparison of 4 methods for quantifying posterior capsule opacification J Cataract Refract Surg 2003 29 106 111 12551676 10.1016/S0886-3350(02)01509-2
Tetz MR Auffarth GU Sperker M Blum M Volcker HE Photographic image analysis system of posterior capsule opacification Journal of Cataract & Refractive Surgery 1997 23 1515 1520 9456409
Aslam TMAPDB Posterior capsular morphology determinants of visual funtion Graefes Arch Clin Exp Ophthalmol 2003 208 212 12644945
Bender L Spalton DJ Uyanonvara B Boyce J Heatley C Jose R Khan J POCOman; New system for quantifying posterior capsule opacification J Cataract Refract Surg 2004 30 2058 2063 15474814 10.1016/j.jcrs.2004.05.010
Findl O Buehl W Siegl H Pinz A Removal of reflections in the photographic assessment of PCO by fusion of digital retroillumination images Investigative Ophthalmology & Visual Science 2003 44 275 280 12506085 10.1167/iovs.02-0619
Baldock RGJ D HB Image Processing and Analysis - A Practical Approach The Practical Approach Series 2000 Oxford, Oxford University Press
Meacock WR Spalton DJ Boyce J Marshall J The effect of posterior capsule opacification on visual function Invest Ophthalmol Vis Sci 2003 44 4665 4669 14578383 10.1167/iovs.02-0634
Gonzalez RWREESL Digital image processing using Matlab 2004 New Jersey, Pearson Prentice Hall
Cheng CY Yen MY Chen SJ Kao SC Hsu WM Liu JH Visual acuity and contrast sensitivity in different types of posterior capsule opacification Journal of Cataract & Refractive Surgery 2001 27 1055 1060 11489575 10.1016/S0886-3350(00)00867-1
McAndrew A Introduction to Digital Image Processing with Matlab 2004 Boston, MA, Thomson Course Technology
Sonka M HVBR Image Processing, Analysis and Machine Vision 1999 Pacific Grove, CA, Brooks/Cole publishing company
| 15817125 | PMC1087491 | CC BY | 2021-01-04 16:03:50 | no | BMC Ophthalmol. 2005 Apr 7; 5:9 | utf-8 | BMC Ophthalmol | 2,005 | 10.1186/1471-2415-5-9 | oa_comm |
==== Front
BMC Public HealthBMC Public Health1471-2458BioMed Central London 1471-2458-5-341581118410.1186/1471-2458-5-34Research ArticleEffect of a chemical manufacturing plant on community cancer rates Mannes Trish [email protected] Katy [email protected] Alan [email protected] Tim [email protected] Vicky [email protected] Jill [email protected] New South Wales Public Health Officer Training Program, Centre for Epidemiology and Research, NSW Department of Health, LMB 961, North Sydney NSW, 2059 Australia2 New South Wales Biostatistical Officer Training Program, Centre for Epidemiology and Research, NSW Department of Health, Sydney, Australia3 Centre for Epidemiology and Research, NSW Department of Health, Sydney, Australia4 Environmental Health Branch, NSW Department of Health, Sydney, Australia2005 5 4 2005 5 34 34 13 10 2004 5 4 2005 Copyright © 2005 Mannes et al; licensee BioMed Central Ltd.2005Mannes 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 conducted a retrospective study to determine if potential past exposure to dioxin had resulted in increased incidence of cancer in people living near a former manufacturing plant in New South Wales, Australia. During operation, from 1928 to 1970, by-products of the manufacturing process, including dioxin and other chemical waste, were dumped into wetlands and mangroves, discharged into a nearby bay and used to reclaim land along the foreshore, leaving a legacy of significant dioxin contamination.
Methods
We selected 20 Census Collector Districts within 1.5 kilometres of the former manufacturing plant as the study area. We obtained data on all cases of cancer and deaths from cancer in New South Wales from 1972 to 2001. We also compared rates for some cancer types that have been associated with dioxin exposure. Based on a person's residential address at time of cancer diagnosis, or at time of death due to cancer, various geo-coding software and processes were used to determine which collector district the case or death should be attributed to. Age and sex specific population data were used to calculate standardised incidence ratios and standardised mortality ratios, to compare the study area to two comparison areas, using indirect standardisation.
Results
During the 30-year study period 1,106 cases of cancer and 524 deaths due to cancer were identified in the study area. This corresponds to an age-sex standardised rate of 3.2 cases per 1,000 person-years exposed and 1.6 deaths per 1,000 person-years exposed. The study area had a lower rate of cancer and deaths from cancer than the comparison areas. The case incidence and mortality due to lung and bronchus carcinomas and haematopoietic cancers did not differ significantly from the comparison areas for the study period. There was no obvious geographical trend in ratios when comparing individual collector districts to New South Wales according to distance from the potential source of dioxin exposure.
Conclusion
This investigation found no evidence that dioxin contamination from this site resulted in increased cancer rates in the potentially exposed population living around the former manufacturing plant.
==== Body
Background
The former Union Carbide site at Rhodes was used for manufacturing chemicals from 1928 until 1986, including the timber preservative creosote, xanthates, the pesticide DDT and herbicides (2,4-dichlorophenoxyacetic acid and 2,4,5-trichlorophenoxy acetic acid). Dioxin and other chemical wastes, which were by-products of the manufacturing processes for some of the above, were dumped into the estuarine wetlands and mangroves along the Rhodes peninsula foreshore until 1970 when it was discovered that these by-products were highly toxic. Until this time dioxin effluent was also discharged into Homebush Bay and dioxin contaminated solid waste was used to reclaim land along the Rhodes foreshore.
Dioxin is a generic name that refers to a group of persistent chlorinated contaminants (polychlorinated dibenzo-p-dioxins, PCDD's, and polychlorinated dibenzofurans, PCDF's), which are unintended by-products of many industrial activities, such as combustion processes, including power generation, metal works and waste incineration, as well as certain types of chemical manufacture. Dioxin is also produced as a result of some natural processes such as bushfires (forest fires) and volcanic activity. Dioxin is well known for its association with a number of adverse health effects, notably cancer. The most toxic of this group of chemicals is 2,3,7,8-tetrachlorodibenzo-p-dioxin (2,3,7,8-TCDD), which has been classified as a Class 1 human carcinogen by the International Agency for Research on Cancer (IARC)[1].
There is concern amongst residents of the Rhodes Peninsula that potential past exposure to dioxin could have resulted in increased incidence of cancer in people living around the Rhodes peninsula. The purpose of this study, therefore, is to determine if potential exposure to dioxins and other pollutants released by the former Union Carbide plant resulted in any difference in the historical incidence of, and mortality due to, cancer in people living on or around the Rhodes peninsula, compared to other people in NSW.
Methods
Selection of study area
The suburb of Rhodes is located on a peninsula in the Parramatta River approximately 12 km west of Sydney Harbour and the central business district, in the state of New South Wales (NSW), Australia. Based on the presumed distribution of past dioxin contamination, the study area was defined as the immediate surrounds of the source site as well as areas directly across the river. Australian Bureau of Statistics (ABS)[2] Census Collector Districts were used to define the study area, as these are the finest level of geographical aggregation, containing approximately 200 households, for which age- and sex-specific population estimates are available. Census Collector Districts (CDs) are defined for each quinquennial Census and are current only at Census time[2].
Collector districts were selected for inclusion in the study area if the majority of the geographic extent of the Collector District fell predominantly within a 1.5 km radius from the Union Carbide site. Where inclusion of a Collector District was marginal based on the distance from the site it was included if the outer boundaries were consistent with boundaries for CDs across censuses from 1971 to 2001. Twenty collector districts were chosen from the 2001 Census. The boundaries for the 1976–1996 censuses are consistent, but there are small changes for the 1971 and 2001 Censuses, covering either end of the study period. All selected CDs fell within the Statistical Local Areas (SLA) of Ryde and Concord. SLA is an aggregate of CDs based on local government boundaries.
Selection of cancer cases
We obtained data on all cases of cancer and deaths from cancer in NSW from 1972 to 2001 from the NSW Central Cancer Registry, which has recorded all cases of malignant neoplasm diagnosed in NSW residents since 1972. Data collected include identifying and demographic information as well as the anatomical and histological characteristics of the disease in each case[3]. Based on a person's residential address at time of cancer diagnosis, or at time of death due to cancer, various geo-coding software and processes were used to ascertain the geographic location of each case or death expressed as latitude and longitude coordinates, using the GDA94 projection. These coordinates were then used to determine to which CD the case or death should be assigned.
Geographical mapping of cancer cases
The first method used to determine geographic location was MapInfo's GeoLoc product[4]. This is a commercial geo-coding product that attempts to match street addresses with a spatial reference database of streets and approximate street number distributions. Any addresses with postal codes of interest (2112, 2114 and 2138) that were unable to be matched by GeoLoc were then investigated individually using MapInfo's MapMarker software[5], which is a newer geo-coding product of similar design from the same vendor. GeoLoc was used first because some of the authors had found on past projects that the address matching algorithms were superior to those found in MapMarker, although the latter uses a more recent reference database. Addresses that were only approximately matched by either GeoLoc or MapMarker were verified manually. Approximately matched addresses are those in which it was necessary to either ignore one or more of street number, street name, street type or suburb name, or to use phonetic encoding of street details in order to obtain a match against the spatial database.
At the conclusion of this process any addresses still unmatched were then looked up in a file of land parcel information (a cadastral database) supplied by the NSW Department of Lands[6]. Finally, any remaining unmatched addresses were investigated, using a street directory and a motor vehicle, to determine whether they fell within the area of interest. Addresses that were not located (21 addresses) were excluded.
Population data
A population breakdown by sex and five-year age group was obtained for each quinquennial Census between 1971 and 2001 for all of NSW, Ryde and Concord SLAs and each of the Collector Districts in the study area. Population counts were used to calculate age-sex standardised incidence ratios to compare the study area to the two standard populations: all of the state of NSW (approximately 6 million people) and the statistical local areas (SLAs) of Ryde and Concord that were not part of the study area. Estimated Resident Populations (ERPs) for 30 June each year for NSW and Ryde and Concord SLAs, were obtained from the Australian Bureau of Statistics (ABS).
The Estimated Resident Population (ERP) is an estimate of the Australian population based on results of the quinquennial Census and is compiled as at 30 June of each year. Inter-Censal ERPs are updated using demographic statistics (births, deaths, overseas and interstate migration) and estimates of housing growth or decline[2]. ERPs are not available at the collector district level, thus unadjusted Census counts were obtained for each Census between 1971 and 2001.
An adjustment to the population figures was required for the collector district that includes Concord Repatriation General Hospital (2001 Collector District 1410103). The NSW Cancer Registry, which provided the numerator data for this study, records residential addresses at time of cancer diagnosis, or death. On the other hand the Census, which provided the denominator data for this study, records all people present on Census night at a particular location, including people in Concord Hospital. We thus attempted to estimate the population of the collector district minus the hospital population. We obtained the average number of occupied beds for each year from hospital and health service annual reports. The age-sex structure of this population was then estimated using data from the NSW Inpatient Statistics Collection[7], which is a complete enumeration of hospital admissions in NSW. Data for Concord hospital is only available for 1993 onwards; the year it became integrated into the State Health System, thus 1993 proportions were used for each census between 1971 and 1991.
Data analysis
Indirect standardisation was used to create various age-sex standardised incidence and mortality ratios (SIRs and SMRs)[8]. Ninety-five per cent confidence limits were constructed around these ratios assuming a Poisson distribution of errors[9]. SIRs and SMRs were calculated for all cancers and for lung and bronchus carcinoma and haematopoietic cancers (such as lymphomas and leukaemias). Dioxin is known to act primarily as a promoter of all neoplasms, based on observations in cell and animal models[1]. However, end-points of carcinoma of the lung and bronchus, and haematopoietic neoplasms were chosen for closer investigation as evidence for increased risk of these cancers were found in occupational cohort studies used by the IARC in the evaluation of the carcinogenicity of 2,3,7,8-TCDD[1]. Cases of carcinoma of the lung and bronchus and haematopoetic cancers were identified using the International Classification of Diseases – Tenth revision (ICD-10) codes.
The area that the 30-year SIRs and SMRs were based on was not completely consistent over the whole study period as the collector district boundaries changed over time. Changes to collector district geographic boundaries occurred between the 1971 and 1976 Censuses and between the 1996 and 2001 Censuses. Thus SMRs and SIRs based on these boundaries included a number of cases (or deaths) that were not enumerated on the same basis as the population for all time periods. To check that these discrepancies did not bias the results, a 25-year count covering the period 1974–1998 was also calculated for comparison. This used 1996 boundaries, which were consistent for each census going back to 1976.
All analysis and data manipulation was carried out using the SAS System® version 8.02 and Microsoft® Excel 2000.
Results
The study area surrounding the Union Carbide site at Rhodes contains 20 Census Collector Districts, representing a resident population of 11,800 people as at the 2001 Census. The population fluctuated between approximately 10,800 and 11,800 during the 30 year study period. The median age in the study area was 37 years, whereas the median age in Concord SLA was 36, Ryde SLA was 34 and the median age in NSW was 35 years. 1,106 cases of cancer and 524 deaths due to cancer were identified in the study area during the study period, corresponding to an age-sex standardised rate of 3.2 cases per 1,000 person-years exposed and 1.6 deaths per 1,000 person-years exposed. Table 1 provides details of the most common types of cancers observed in the study area during the study period.
Table 1 Number (and percentage) of cases of selected types of cancers occurring in the Rhodes study area compared with New South Wales (based on 30 year counts).
Cancer type Rhodes (number of new cases) Percentage
Lung and Bronchus Carcinomas 143 12.93%
Haematopoietic 85 7.69%
Leukaemia 31
Non-Hodgkin's Lymphoma 40
Multiple Myeloma 9
Prostate 116 10.49%
Breast cancer 142 12.84%
Colorectal 153 13.83%
Gynaecological 61 5.52%
Urothelial 35 3.16%
Melanoma of the skin 94 8.50%
Site unknown 57 5.15%
Definitions:
Colorectal: Cancers of colon, rectum and anus (ICD-10 C18-C21)
Haematopoietic: Lymphoma, leukaemia, multiple myeloma (ICD-10 C81-C96)
Gynaecological: Uterus, ovary, other female genital organs (ICD-10 C51-C58)
Urothelial: Urinary tract excluding body of kidney (ICD-10 C65-C68)
Table 2 presents the results of SIR and SMR analyses for 30 year and 25 year group counts for all cancers. Aggregated over the entire study period, the study area had a significantly lower rate of cancer than the whole of NSW (SIR = 88%, 95%CI 83–93). The study area also had a significantly lower rate of death due to cancer than the whole of NSW during the entire study period (SMR = 89%, 95%CI 81–97).
Table 2 SIR and SMR for Rhodes study area compared with New South Wales and compared with Ryde and Concord SLAs for all cancers.
Time Frame Number of cases/ deaths Comparison Area Case SIR (Indirectly Standardised) 95% Confidence Interval Death SMR (Indirectly Standardised) 95% Confidence Interval
30 years 1106/524 New South Wales 87.
6# (82.5, 92. 9) 88.7# (81.3, 96.7)
30 years 1106/524 Concord and Ryde SLAs 90.5# (85.2, 96.0) 92.9 (85.1, 101.2)
25 years 922/530 New South Wales 91.2# (85.4, 97.2) 93.5 (85.2, 102.3)
Notes:
SLA = Statistical Local Area
SIR = Standardised Incidence Ratio
SMR = Standardised Mortality Ratio
# Statistically significantly lower than the comparison area
* Statistically significantly higher than the comparison area
A comparison was also made to the numbers expected if the rate observed in the balance of the SLAs of Ryde and Concord was experienced in the study area. The case incidence in the study area was statistically significantly lower than the rate in the surrounding local area (SIR = 91%, 95%CI 85–96), while the death rate was marginally lower (SMR = 93%, 95%CI 85–101).
The 25 year results (Table 2) were similar to the results obtained using 30 year counts, suggesting that the slight variation in the definition of the study area used for the 30 year period did not bias the results.
The yearly SIRs for the study area compared to NSW varied between 51% (95%CI 50.0–80.0) in 1976 and 130% (95%CI 97.2–107.6) in 1984, while the yearly SMRs were in the range 43% (95%CI 19.8–82.3) in 1991 to 124% (95%CI 82.0–181.0) in 1984. There was no obvious trend in ratios with apparently random, symmetrical fluctuation around the overall 30-yr averages of 87.6% (95%CI 82.5–92.9) and 88.7% (95%CI 81.3–96.7), for SIR and SMR respectively. The count of cases and deaths that determined these ratios varied between 20–50 cases and 10–30 deaths per year.
Similarly, when groups of years were aggregated around census years for year-block analysis there was no obvious pattern in the SIR and SMR for the study area compared to NSW. The SIR ranged from 77% (95%CI 64.5–91.2) in the period 1974–1978 to 99% (95%CI 85.3–113.0) in the period 1984–1988, while the SMR varied between 76% (95%CI 60.4–94.7) in the period 1989–1993 to 108% (95%CI 89.0–129.4) in the period 1984–1988.
Table 3 shows SIR and SMR for the individual CDs compared to NSW. The SIR for the individual CDs compared to NSW varied between 49% (95%CI 25–84) and 184% (128–256), while the SMR ranged from 22% (95%CI 3–78) to 137% (95%CI 102–181). Again, there was no obvious trend in ratios with symmetrical fluctuation around the overall total study area 25 year averages of 91% (95%CI 85–97) and 94% (95%CI 85–102), for SIR and SMR respectively. There was no apparent relationship between SIR, SMR and distance of the CD relative to the source of dioxin contamination.
Table 3 SIR and SMR for each census collector district in the Rhodes study area compared with New South Wales for all cancers (based on 25 year counts).
Geographical Region Number of cases Distance from site+ (metres) Case SIR (Indirectly Standardised) 95% Confidence Interval Death SMR (Indirectly Standardised) 95% Confidence Interval
1410101 76 200 113.6 90.3–141.0 119.0 85.4–161.4
1410102 34 985 149.0* 108.7–199.3 113.8 66.3–182.3
1380508 14 1014 49.0# 26.8–82.3 52.3 19.2–113.9
1380507 13 1190 48.5# 25.0–84.6 21.7# 2.6–78.2
1380503 35 1241 96.3 67.8–132.7 111.2 68.0–171.8
1380509 58 1248 184.1* 128.3–256.1 128.2 61.5–235.8
1380501 76 1250 76.3# 57.8–98.9 66.8# 42.8–99.3
1410103 78 1302 91.3 74.0–111.5 81.9 58.8–111.1
1381207 61 1415 80.1 61.1–103.1 79.5 54.0–112.9
1381209 76 1428 102.8 81.9–127.4 137.4* 102.3–180.6
1380505 25 1527 94.3 63.2–135.5 107.7 58.9–180.6
1380504 58 1530 83.0 63.5–106.6 93.1 66.2–127.2
1380510 51 1561 71.8# 51.3–97.8 79.9 52.2–117.1
1381211 38 1576 101.3 72.4–138.0 83.5 46.7–137.7
1380506 22 1633 100.9 63.2–152.8 116.7 58.3–208.8
1381210 11 1638 56.3 28.1–100.6 88.3 38.1–174.0
1410104 8 1682 84.5 56.2–122.2 93.4 52.3–154.1
1380502 34 1717 105.8 80.2–137.1 109.1 72.5–157.6
1410105 71 1874 89.6 68.8–114.6 99.0 68.6–138.3
1410301 83 1935 81.1 63.6–102.0 82.6 58.2–113.9
Notes:
SLA = Statistical Local Area
SIR = Standardised Incidence Ratio
SMR = Standardised Mortality Ratio
+ Distance from site = distance between the centroid of the collector district from the centre of the former Union Carbide site
# Statistically significantly lower than the comparison area
* Statistically significantly higher than the comparison area
Table 4 shows the SIR and SMR for carcinoma of the lung and bronchus and haematopoietic carcinoma. Over the 30 year study period there were 143 cases and 121 deaths due to carcinomas of the lung and bronchus identified in the Rhodes study area. The case incidence and mortality ratios due to lung and bronchus carcinoma were not significantly different to NSW or to the statistical local areas of Ryde and Concord for the study period.
Table 4 SIR and SMR for Rhodes study area compared with New South Wales and compared with Ryde and Concord SLAs for lung and bronchus carcinoma and haematopoietic cancers (based on 30 year counts)
Cancer type Number of cases/ number of deaths in Rhodes study area Comparison Area Case SIR (Indirectly Standardised) 95% Confidence Interval Death SMR (Indirectly Standardised) 95% Confidence Interval
Lung and Bronchus Carcinomas 143/121 NSW 105.5 88.9–124.2 107.1 88.9–128.0
Haematopoietic neoplasms 85/46 NSW 80.6# 64.3–99.6 80.8 59.2–107.8
Lung and Bronchus Carcinomas 143/121 Concord & Ryde SLAs 115.7 97.5–136.3 116.2 96.4–138.8
Haematopoietic neoplasms 85/46 Concord & Ryde SLAs 75.0# 59.9–92.7 78.3 57.3–104.4
Notes:
SLA = Statistical Local Area
SIR = Standardised Incidence Ratio
SMR = Standardised Mortality Ratio
# Statistically significantly lower than the comparison area
* Statistically significantly higher than the comparison area
Over the 30 year study period there were 85 cases and 46 deaths due to the haematopoietic neoplasm identified in the Rhodes study area. The case incidence of the haematopoietic class of cancers in the Rhodes study area was marginally statistically significantly lower than NSW and Ryde and Concord over the study period. There is no significant difference in the mortality due to this class of cancers.
Discussion
We examined age-sex standardised incidence and mortality ratios due to cancer in the 20 Collector Districts within 1.5 km from the former Union Carbide site (the study area). We used incidence and mortality for all of NSW and for the balance of the Ryde and Concord Statistical Local Areas, which enclose the study area, as the comparison. We calculated standardised incidence and mortality ratios for all cancer as well as carcinoma of the lung and bronchus and haematopoietic neoplasm. Annual, twenty-five year and thirty year comparison periods were used. We found that the incidence and mortality ratios for all types of cancer for all comparison periods of study in the study area did not differ significantly from expectation, based on cancer incidence and mortality rates in all of NSW and in the areas immediately surrounding the study area.
There are a number of strengths to this study. We were able to obtain information on all cases of and deaths from cancer in the study and control areas during the study period as the NSW Central Cancer Registry is supported by compulsory notification of all cancers in NSW. We selected those collector districts whose population fell within 1.5 km of the former Union Carbide site. The study area included 20 Collector Districts with a population of approximately 11,500 people. Since there was data available for a number of years we were able to examine cancer incidence and mortality in individual Collector Districts, demonstrating that even the residential areas closest to the site did not have increased cancer rates.
The study did present several challenges, however. As with any small area study, one of the biggest difficulties was the definition of study boundaries, which are consistent over a sufficiently long period to accumulate a reasonable number of cases or deaths. This was a challenge for a number of reasons. Firstly, the exposed community was not easily defined, because the potential exposure route was unclear. Historically there was no known exposure via air. Potential exposure routes may have been contact with Homebush Bay, eating fish caught in the bay, soil transferred from the site or occupational exposure from working at the facility. It is also possible that no exposure to dioxins occurred in residents surrounding the Union Carbide plant.
Secondly, the choice of possible boundaries was limited to those for which denominator (population) counts were available, as these were needed to calculate expected counts of cases and deaths due to cancer. The smallest units for which population data are available are the ABS Census Collector Districts. Unfortunately, these boundaries were still too coarse to enable definition of a study area with a circumference that was a uniform distance from the putative exposure source. Also, the ABS may change the boundaries of Collector Districts at each Census in response to significant changes in population. Fortunately for 25 of the 30 years of this study the boundaries remained consistent allowing direct comparisons across time periods.
The geo-coding process used to determine if a case belonged to the exposed study area presented some difficulties. Although the address information for each case was quite "clean" and well formatted, only eighty one percent (4592 out of 5644) of the cases falling into the postcodes containing the study area could be matched exactly using an automatic process. On further investigation, approximately seven percent of cases originally attributed to the study area, but not matched exactly, were re-allocated to another Collector District. However, overall the total number of cases falling into the study area changed by only 2 (out of 1106). The number of cases where the address was unable to be found, and therefore the case excluded from the study area, was a very small proportion of the total number of cases recorded in these postcodes (21 cases out of 5644, or 0.4 percent).
An additional challenge in this study was assigning population estimates to areas for the purpose of calculating expected numbers of cases and deaths. The methods used to assign population estimates varied between the study and control areas. Since yearly population estimates (in the form of ERPs) were not available at the Census Collector District level, Census year enumerated population counts were used and extrapolated to inter-Censal years. ERPs (from the Australian Bureau of Statistics) were used for NSW and the SLAs of Ryde and Concord and more closely reflect actual population. The use of enumerated usual resident populations in the study area instead of ERPs may result in small errors in the SIRs and SMRs. However, the magnitude of these errors is unlikely to be large enough to affect the results of the study. Additionally, population figures for the Collector Districts, which included Concord Hospital, overestimate the residential population due to counting of hospital inpatients on Census night. The population figures were adjusted by subtracting the estimated number of people at Concord Hospital on Census night, thus providing a better estimated residential population in this Collector District.
Another limitation of this study is our inability to adjust for potential confounders in the relationship between cancer incidence and mortality and place of residence. Potential confounders include smoking rates and socio-demographic variables. The comparison of the study area to the balance of the SLAs of Ryde and Concord may compensate in part for this deficiency, as this area, which is contiguous with the study area, is likely to have demographic characteristics which are very similar to the study area.
One of the major weaknesses in small area studies is that the numbers of events being examined are often very small and therefore variation in these numbers can more often be attributed to chance than any exposure effect. This was particularly the case when we looked at annual incidence and mortality. Due to the small numbers of events each year the confidence intervals around the ratios are wide and consequently there are only three significant results, with the SIR for 1974 and 1976, and the SMR for 1991 being significantly lower than 100%. When testing the significance of this many ratios (sixty in total) at the 95 percent confidence level, three significant results are expected purely as a result of chance. Given the negative result no attempt was made to adjust for the effect of multiple comparisons. Based on this and the fact that there is no obvious trend in the yearly SIRs and SMRs we conclude that there is no time effect on the incidence of and mortality due to cancer in the Rhodes area compared to NSW.
It is possible that many workers at the Union Carbide plant resided in the study area. Had occupational exposure to dioxin on the site been high, workers may have experienced greater rates of cancer than the general population and may have raised the observed risk of cancer in our study. The cancer registry does not collect occupational data, thus it was not possible to adjust for such potential confounding.
There are a number of additional issues to be considered when interpreting studies of cancer incidence around a point source of carcinogenic exposure. Cancer risk is likely to be significantly greater for those working at the plant than those residing near to the plant, because potential exposure to carcinogens is likely to be much greater in workers. Accordingly, in residents the same cancers would be expected as those observed in workers, but at lower incidence. Thus, many studies of cancers in residents are negative and may be underrepresented in the literature. In some cases where an increased risk of cancer has been demonstrated in residents, a causal association has not been assumed, as the cancers observed are different to those seen in occupational studies[10], or there is an absence of measured exposure to the contaminant of concern[11].
Human exposure to TCDD, studied in the context of industrial accidents and occupational exposure, is associated with increased risk of cancer (all sites), and in some groups, increased risk of lung cancer, soft tissue sarcoma and non-Hodgkin lymphoma have been observed[1,12]. TCDD, therefore, is considered to be an overall promoter of cancer. Cancer excesses have not been observed in TCDD-exposed cohorts with lower exposures[12]. Increased risk of cancer reasonably attributed to site contamination by TCDD (as distinct from exposure to TCDD consequent upon an industrial disaster or in workers) has never been demonstrated[1]. In the present study we found no evidence that dioxin from this site has resulted in increased cancer rates in the potentially exposed population on the Rhodes peninsula and in the surrounding area. The findings of this study are consistent with the understanding that the risk of cancer attributable to residing near a site contaminated by TCDD is low.
Conclusion
We did not find a significant increase in the historical incidence of, or mortality due to, cancer for people living around the Rhodes peninsula compared to other people in NSW. The findings of this study are consistent with the understanding that the risk of cancer attributable to residing near a site contaminated by TCDD is low.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
TM managed the project in the later stages and prepared the manuscript. KE managed the project in the early stages, including compiling data, geo-coding cases and deaths and calculating summary statistics. KE also prepared an early draft of the manuscript. AW assisted with the geo-coding of cases and deaths using Geoloc and MapMarker software, and the LPI database. TC and VS were involved in planning and advising on the project throughout. JK provided statistical advice throughout the project. All authors provided comment on early drafts of the manuscript and all authors read and approved the final manuscript.
Figure 1 Map of study area with Census collector district boundaries. A: Approximate 1.5 km distance from the former Union Carbide plant. B: The site of the former Union Carbide plant. Study Area: Includes all Collector Districts marked on the aerial photograph. Ryde Statistical Local Area: Is located to the north of the study area. Concord Statistical Local Area: Is located to the south of the study area. The area to the west of the former Union Carbide plant is former industrial land. No residential occupation occurred here until approximately the year 2000.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We acknowledge the contribution of residents of the Rhodes Peninsula community, (including Paul Hanly, Rhodes Peninsula Group and Carol Kendall and Jenny Nicholls, Rhodes Resident's Group) who were involved in planning the investigation. We also acknowledge the guidance of the New South Wales Health Rhodes Peninsula Steering Committee (including Alison Rutherford and Central Sydney, Northern Sydney and Western Sydney Public Health Units).
==== Refs
IARC Monographs on the Evaluation of Carcinogenic Risks to Humans and their Supplements Polychlorinated Dibenzo-para-Dioxins and Polychlorinated Dibenzofurans 1997 69 Lyon: International Agency for Research on Cancer 1 631
Australian Bureau of Statistics Census Dictionary: Census Concepts and Definitions (Catalogue no 29010) 2001 Canberra: ABS Accessed June 2003.
Tracey EA Supramaniam R Chen W Cancer in New South Wales: Incidence and Mortality 2001 2003 Sydney: The Cancer Council NSW
MapInfo Corporation MapInfo Professional Geoloc module New York: MapInfo Corporation
MapInfo Corporation MapMarker software New York: MapInfo Corporation
New South Wales Department of Lands Land and Property Information database 2003 Sydney: New South Wales Department of Lands Accessed June 2003.
NSW Department of Health NSW Inpatient Statistics Collection (HOIST) 2003 Centre for Epidemiology and Research, NSW Department of Health Accessed June 2003.
Armitage P Berry G Matthews JNS Statistical Methods in Medical Research 2002 4 Blackwell Scientific Publications, Oxford
Washington State Department of Health Guidelines for Using Confidence Intervals for Public Health Assessment 2002 Accessed June 2003.
Bithel JF Dutton SJ Draper GJ Neary NM Distribution of childhood leukaemias and non-Hodgkin's lymphomas near nuclear installations in England and Wales BMJ 1994 309 501 505 8086902
Westley-Wise VJ Stewart BW Kries I Ricci PF Hogan A Darling C Corbett S Kaldor J Stacey NH Warburton P Investigation of a cluster of leukaemia in the Illawarra region of New South Wales, 1989–1996 MJA 1999 171 178 183 10494232
Crump KS Canady R Kogevinas M Meta-analysis of dioxin cancer dose response for three occupational cohorts Environmental Health Perspectives 2003 111 681 85 12727594
| 15811184 | PMC1087492 | CC BY | 2021-01-04 16:28:56 | no | BMC Public Health. 2005 Apr 5; 5:34 | utf-8 | BMC Public Health | 2,005 | 10.1186/1471-2458-5-34 | oa_comm |
==== Front
BMC PsychiatryBMC Psychiatry1471-244XBioMed Central London 1471-244X-5-181580435510.1186/1471-244X-5-18Research ArticlePharmacokinetics, safety, and tolerability of a depot formulation of naltrexone in alcoholics: an open-label trial Galloway Gantt P [email protected] Monika [email protected] Ryan [email protected] David E [email protected] Haight Ashbury Free Clinics, San Francisco, USA2 The Permanente Medical Group, Chemical Dependency and Recovery Program, Vallejo, USA3 Department of Clinical Pharmacy, University of California San Francisco, San Francisco, USA2005 1 4 2005 5 18 18 5 10 2004 1 4 2005 Copyright © 2005 Galloway et al; licensee BioMed Central Ltd.2005Galloway 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
Naltrexone is an effective medication for treatment of alcohol dependence, but its efficacy is limited by lack of adherence to the oral dosage form. A long-acting depot formulation of naltrexone may increase adherence.
Methods
A single site, 6-week open label study was conducted with 16 alcohol dependent subjects each receiving 300 mg of Naltrexone Depot by intramuscular injection. The main outcomes were safety and tolerability of the Naltrexone Depot formulation, blood levels of naltrexone and its main metabolite 6-beta naltrexol, and self-reported alcohol use. All subjects received weekly individual counseling sessions.
Results
The medication was well tolerated with 88% of subjects completing the 6-week trial. The most common side effect experienced was injection site complications. There were no serious adverse events. Subjects had naltrexone and 6-beta-naltrexol concentrations throughout the trial with mean values ranging from 0.58 ng/mL to 2.04 ng/mL and 1.51 ng/mL to 5.52 ng/mL, respectively, at each sampling time following administration. Compared to baseline, subjects had significantly reduced number of drinks per day, heavy drinking days and proportion of drinking days.
Conclusion
Naltrexone Depot is safe and well tolerated in alcoholics and these findings support the further investigation of its utility in larger double-blind placebo controlled trials.
==== Body
Background
Naltrexone is an opiate receptor antagonist that was approved in 1984 for the treatment of opiate dependence. The safety and efficacy of naltrexone in reducing alcohol consumption were established in controlled clinical trials [1,2] which led, in 1994, to United States Food and Drug Administration (FDA) approval of a 50 mg oral tablet for treatment of alcohol dependence. Since receiving FDA approval, several studies on naltrexone's effectiveness in reducing alcohol consumption have been conducted. Naltrexone has demonstrated effects on drinking behavior in alcoholics through reducing alcohol use among subjects who sample alcohol as well as promotion of alcohol abstinence [3-5]. One study that used cognitive behavior therapy and random assignment to oral naltrexone or placebo in recently abstinent alcoholics found that 62% of the naltrexone group did not relapse into heavy drinking in comparison with 40% of the placebo treated subjects [4]. In the laboratory setting, naltrexone has been shown to increase the latency to drink alcohol in social drinkers [6], to reduce drinking in heavy drinkers [7] and increase certain discriminant and sedative effects of ethanol while reducing the positive reinforcing effects of ethanol in non-alcoholics [8].
While many single site studies and 2 meta-analyses [9,10] indicate that naltrexone is more effective than placebo, there have also been 2 multisite studies and 1 single site study in which naltrexone was not found to be effective in decreasing alcohol consumption [11-13].
One factor that appears to be important for naltrexone's effectiveness is adherence. In 2 studies, oral naltrexone was found to have an effect only in the population of patients who were highly compliant with their medication [14,15]. Alcoholics have been shown to have particularly low rates of medication adherence [16,17]. With treatments intended to reduce or prevent relapse to alcohol or drug abuse, non-adherence has an additional aspect. If a patient on naltrexone wants to resume drinking or using drugs, they can discontinue their medication and experience the full effect of the drug.
In alcoholics, an injectable sustained release formulation (SRF) would be highly advantageous. A SRF of naltrexone would reduce the number of opportunities to impulsively discontinue their medication and ensure that discontinuation came to the attention of the health care provider scheduled to administer the injection. A SRF would minimize the number of doses required and guarantee exposure to the medication for at least the duration of the first injection. It would have the added benefit of producing a more consistent and predictable drug blood level since a depot injection bypasses first pass metabolism [18].
Kranzler et al. [5] conducted a 12-week study of a different depot formulation of naltrexone in 20 alcohol dependent subjects. Subjects received either a subcutaneous injection of 206 mg NTX (N = 15) or placebo (N = 5) along with eight weekly coping skills sessions. Compared to placebo, subjects who received the SRF of naltrexone had fewer heavy drinking days. These initial results support the continuation of research into the use of a SRF of naltrexone in patients being treated for alcohol dependence.
Comer et al. [19], evaluated the effectiveness, time course and safety of the same naltrexone depot formulation used in Kranzler's study [5], in 12 heroin dependent subjects over an 8 week time span. The results from this study showed that blood plasma levels remained above 1 ng/mL for 3–4 weeks after receiving either 192 mg or 384 mg of naltrexone depot. In this study there were few adverse events reported other than the discomfort associated with the actual injection of the depot formulation. It was shown that the depot formulation of naltrexone at both doses provided safe, effective and long lasting antagonism of the effects of heroin.
The current study, a single site open-label study, was designed to examine the safety, pharmacokinetics and tolerability of the DrugAbuse Sciences (DAS; Hayward, CA) Naltrexone Depot formulation for treatment of alcohol dependence. During the study, subjects received 1 depot injection of naltrexone and 6 weeks of individual counseling.
Methods
Study design
Sixteen subjects participated in this 6-week, single site, open-label study to investigate the safety and tolerability of Naltrexone Depot along with naltrexone and 6-beta-naltrexol blood levels. The study was conducted at Friends Research Associates in Berkeley, CA. All procedures were approved by the Institutional Review Board of Friends Research Institute.
Following initial screening, subjects were required to have 3 days of abstinence from alcohol. Subjects who tolerated four days of oral naltrexone returned to the clinic the day following their last dose of oral naltrexone and received intramuscular injections of Naltrexone Depot containing 300 mg of naltrexone. See figure 1 for an outline of study procedures.
Figure 1 Study design of naltrexone depot in alcohol dependence.
Subjects
Subjects were recruited through advertising. Subjects were males and non-pregnant, non-nursing females, age 18 to 65, with DSM-IV diagnoses of alcohol dependence who expressed a desire to stop drinking. To meet inclusion criteria, subjects had to have had at least 1 day of heavy drinking (≥ 5 drinks/day for males or ≥ 4 drinks/day for females) within the preceding 14 days. As noted above, they had to be able to achieve at least three consecutive days of sobriety, without detoxification medications. Female subjects were required to practice effective birth control for the duration of the trial and all subjects were required to provide a urine sample negative for amphetamines, barbiturates, benzodiazepines, cocaine and opiates.
Subjects were excluded from the study if they were currently taking disulfiram, naltrexone or a neuroleptic medication; if they needed medical detoxification from alcohol; or if they had a DSM IV diagnosis of dependence on any drug of abuse other than nicotine or alcohol. Subjects were also excluded if they had an ALT or AST elevation more than 3 times the upper limit of normal or any psychiatric condition (e.g., depression with suicidal ideation) or medical condition that would preclude safe participation in the protocol. A complete list of inclusion and exclusion criteria is presented in Table 1.
Table 1 Inclusion and Exclusion Criteria
Inclusion Criteria
• Males or non-pregnant, non-nursing females, age 18 to 65, with a DSM-IV diagnosis of alcohol dependence who express a desire to stop drinking.
• Heavy drinking (5 drinks/day for males or 4 drinks/day for females) within 14 days prior to randomization
• Able to achieve at least three continuous days of sobriety, without detoxification medications, immediately before beginning the oral naltrexone run-in dosing
• Willing and able to give informed consent
• Willing to practice effective birth control for duration of trial (female patients only)
• Available to participate in the study for 7 weeks
• Willing to provide names and permission to contact someone (e.g., spouse, parent, friend) who would likely know their whereabouts for follow-up tracking
• Naltrexone tolerance as demonstrated during run-in dosing period.
Exclusion Criteria
• Currently taking disulfiram (Antabuse), naltrexone (Revia or generic) or neuroleptic medication
• In need of medical detoxification from alcohol
• DSM-IV diagnosis of dependence on any drug of abuse other than nicotine or alcohol or drug screen showing benzodiazepines, marijuana, cocaine, methamphetamine, barbituates or heroin
• Clinical evidence of cardiac ischemia (by EKG or medical history) or history of myocardial infarction within the previous 2 years
• History of pancreatitis
• Planned surgery within 7 weeks of screening
• Any chronic or episodic painful conditions that would reasonably require opiate medications for pain control
• ALT or AST elevations more than 3 times the upper limit of normal
• Subjects with any psychiatric (e.g., depression with suicidal ideation) or medical condition that would preclude safe participation in the protocol.
• History of allergic or adverse response to naltrexone
• Participation in a trial of an investigational medication within 30 days prior to study enrollment
•Subjects mandated by court for alcohol or drug abuse treatment or having pending legal proceedings that could result in incarceration within 7 weeks of screening
Study procedures
Before participation in study procedures began, prospective subjects had the study explained to them by a member of the research team. The purpose of the study was reviewed, and the potential risks and discomforts of study participation were explained. Subjects were given a copy of an informed consent form to read. Just prior to signing the consent form a Breathalyzer test was performed; subjects were not able to sign the consent unless their BAL was less than 0.04 mg/dL. Following signing of the consent form, a medical history and physical examination were conducted to assess eligibility for study participation. In addition, subjects had an EKG performed along with screening hematology, clinical chemistry, serum pregnancy test (for women of childbearing age), urine toxicological screen and a routine urinalysis.
Following screening, subjects were administered one 50 mg naltrexone tablet at the clinic. Three additional tablets were dispensed for the subjects to take on their own (1 each day for 3 days). Subjects then returned to the clinic site the day following the final naltrexone dose. Subjects who tolerated the oral naltrexone were eligible to participate in the study. The rationale for an oral naltrexone run-in was to avoid exposing naltrexone intolerant individuals to Naltrexone Depot, which once it is given cannot be removed. Eligible subjects received 2 intramuscular injections of 150 mg of Naltrexone Depot, one in each buttock, for a total dose of 300 mg.
Subjects then received brief counseling following their injection and at each weekly study visit throughout the 6-week trial. A therapist provided a manualized form of counseling called BRENDA. BRENDA is an acronym for a six step approach to counseling: Biopsychosocial evaluation, Report to patient on assessment, Empathetic understanding of the patients problems, Needs expressed by the patient which should be addressed, Direct advice on how to meet those needs, and Assess response/behaviors of patients to advise and adjust treatment recommendations. BRENDA was chosen because, with treatment of alcohol and drug abuse increasingly occurring in the context of mainstream medical care, it is important to use brief counseling therapies that are feasible within these settings [20,21]. Subjects were permitted but not required to attend Alcoholics Anonymous (AA) or other self-help recovery programs.
Outcome measures
Each week, subjects were provided with a drinking diary. In the diary, subjects were asked to record all of the alcoholic drinks that they consumed and to return to clinic with the completed diary each week. At each weekly visit, the research assistant reviewed the diary with the subject and recorded adverse events, service utilization and drinking history. Timeline follow-back procedures were used [22-25]. Using timeline follow-back procedures, the number of standard drinks per day was recorded for each day since the last visit. If a subject missed a data collection visit, the timeline follow-back was extended to the last visit attended. Subjects rated their maximum craving for alcohol during the past week on a visual analog scale. The research assistant also queried subjects about alcohol craving and concurrent medications. Subjects had their injection sites examined, a Breathalyzer test performed, a blood draw taken and a urinalysis collected during each weekly visit. The blood was analyzed for naltrexone and 6-beta-naltrexol plasma levels along with GGT levels.
An LC-MS/MS method was used to determine plasma naltrexone and 6-beta-naltrexol levels simultaneously. A liquid/liquid extraction under basic conditions was done before injection into the LC-MS/MS. A PE Sciex API III+, using an electrospray interface, was employed in this study. Four analytical runs were required to process the samples from this study. Naltrexone and 6-beta-naltrexol analyses were performed by MDS Pharma Services, Lincoln, NE.
The linear range for naltrexone was 0.100 to 50.000 ng/mL with a limit of quantitation of 0.100 ng/mL. Naltrexone quality control samples analyzed with each analytical run had coefficients of variation less than or equal to 6.14% and absolute relative errors less than or equal to 7.67%. The linear range for 6-beta-naltrexol was 0.400 to 100.000 ng/mL with a limit of quantitation of 0.400 ng/mL. 6-beta-naltrexol quality control samples analyzed with each analytical run had coefficients of variation less than or equal to 7.40% and absolute relative errors of less than or equal to 7.70%.
Drinking measures were: drinking days (days on which any drinking occurred), heavy drinking days (≥ 5 drinks/day for men; ≥ 4 drinks/day for women) and drinks per drinking day. These measures were cumulated for the 14-day baseline period and for the 6 weeks following the naltrexone injection.
Alternate measures were generated for days on which drinking data were not obtained after naltrexone was administered. These missing days were counted as both drinking days and heavy drinking days and the number of drinks per drinking day were set equal to the mean number of drinks per drinking day during the baseline period. Analyses using these alternate measures did not yield significantly different results and are not presented.
Study medication
Naltrexone is a competitive antagonist at the mu opiate receptor. Following oral administration naltrexone is almost completely absorbed, but subject to first pass metabolism; the bio-availability ranges from 5% to 60% [26,27]). Naltrexone is primarily eliminated by the liver with only 1% of an oral naltrexone dose excreted in urine [28]. Conjugated and non-conjugated 6-beta-naltrexol are the major metabolites found in plasma, urine and feces. The half-life of elimination of parent naltrexone is 2.9 hours and 8.8 hours for 6-beta-naltrexol, which is a weak antagonist at the mu opiate receptor [27,29].
Naltrexone Depot consists of microspheres of poly (D, L-lactide) and naltrexone, administered by intra-muscular injection. The microspheres are mixed with a diluent containing water, mannitol, carboxymethylcellulose and polysorbate 80 to form a suspension for injection. Once injected, the suspension forms a solid gel pellet with the rate of release proportional to the surface area of the gel pellet, the loading of naltrexone in the microspheres and the porosity of the microspheres. The polylactide polymers are broken down to monomers by hydrolysis, releasing naltrexone and ultimately being metabolized to carbon dioxide and water.
Statistical analysis
Concentration-time data were analyzed by noncompartmental techniques using WinNonlinProfessional 3.1. The pharmacokinetic parameters calculated were Cmax, tmax, AUC0-t, AUC0-last, and the metabolite to parent AUC0-t ratio.
Drinking data were analyzed using the univariate procedure of SAS, version 8.2. Changes in liver function tests were evaluated using a one-way analysis of variance comparing baseline values to end of study values.
Results
Subjects
Of the 17 subjects screened for study participation, 16 subjects (94.9%) met the inclusion/exclusion criteria and received study medication. Of the 16 participating subjects, 14 (87.5%) completed the study through week 6. Of the 2 subjects (12.5%) who did not complete the study, one subject was lost to follow-up while the other was unwilling to complete the study due to the time required. The mean age of the participants in the study was 49 years with a range of 27–60 years. Thirteen (81%) of the subjects were male and 10 (62.5%) of the subjects were Caucasian. Demographic and baseline clinical characteristics are presented in Table 2.
Table 2 Baseline Demographic and Clinical Characteristics
Characteristic
Age in yr, mean (SD) 49.2 (13.4)
Gender, N (%)
Male 13 (81.3)
Female 3 (18.8)
Ethnicity, N (%)
Caucasian 10 (62.5)
Black 5 (31.3)
Hispanic 11 (6.3)
DSM-IV Alcohol Dependence Criteria Met, mean (SD) 5.4 (0.7)
Baseline Heavy Drinking Days, mean (SD) 10.9 (3.5)
Baseline Drinks Per Drinking Day, mean (SD) 8.4 (3.5)
Baseline Heavy Drinking Days, mean (SD) 8.1 (3.8)
Safety measures
As shown in Table 3, there was no significant difference between patients' vital signs and liver function tests at the beginning of the study and at the completion of the study.
Table 3 Vital Signs and Liver Function Tests
Parameter Pre-Treatment (= day -7) mean ± sd Post-Treatment (last visit) mean ± sd
Number of Patients 16 14
Systolic Blood Pressure, mm Hg 135 ± 13 132 ± 10
Diastolic Blood Pressure, mm Hg 85 ± 9 83 ± 6
Pulse, beats/min 80 ± 5 79 ± 9
AST, IU/L 31.9 ± 14.6 27.6 ± 22.7
Total Bilirubin, mg/dL 0.6 ± 0.3 0.6 ± 0.3
Over the course of the study all 16 subjects had 1 or more adverse events. A total of 15 subjects had injection site adverse reactions. Table 4 shows the injection site adverse events experienced by subjects in the study. Three subjects experienced drainage from their injection site. The drainage fluid of 2 of the subjects was available for culture and in both cases did not grow out any organisms. Of the 198 adverse events reported, 17 were rated severe: nausea, flatulence, gastrointestinal pain, fatigue, lethargy, somnolence (2 reports), depression, irritability, headache (4 reports from 3 subjects), back pain, injection site mass, injection site pain and elevated GGT. There were no serious averse events. As shown in Table 5, there were 13 reported changes in biochemical markers that were outside of normal ranges.
Table 4 Adverse Injection Site Events
Adverse Event* Mild Moderate Severe
Bruising 2 0 0
Inflammation 1 1 0
Mass 3 3 1
Pain 3 6 1
Pigmentation Change 2 1 0
Site Reaction NOS 2 1 0
*All values, N (%)
Table 5 Laboratory Value Abnormalities
Adverse Event % Reporting Adverse Event
Blood Lactate Dehydrogenase Increased 31.3
Gamma-Glutamyl Transferase Increased 18.8
Red Blood Cells Present In Urine 12.5
Proteinuria Present 12.5
Alanine Aminotransferase Increased 12.5
Blood Creatine Increased 12.5
Hemoglobin Decreased 12.5
Hematocrit Decreased 6.3
Lymphocyte Count Decreased 6.3
Neutrophil Count Increased 6.3
Protein Total Increased 6.3
Aspartate Aminotransferase Increased 6.3
Uric Acid Increased 6.3
Urine Analysis Abnormal NOS 6.3
White Blood Cell Count Increased 6.3
White Blood Cells In Urine 6.3
Drinking measures
All measures of drinking declined from the baseline period to the 6-week treatment period: proportion of drinking days from 0.77 (s.d. 0.25) to 0.46 (0.33) (signed rank; S = -47.5, p = 0.0044), drinks per drinking day from 8.4 (s.d. 3.5) to 4.9 (s.d. 3.4) (S = -64, p = 0.0002) and proportion of heavy drinking days from 0.58 (s.d. 0.27) to 0.25 (s.d. 0.27) (S = -56, p = 0.0004).
Pharmacokinetics
The pharmacokinetic results are based on weekly plasma sampling for six weeks on day 0, 7, 14, 21, 28, 35 and 42 following the single dose administration of 300 mg (2 × 150 mg) Naltrexone Depot. On Day 0 blood samples were drawn prior to dosing of Naltrexone Depot. Pre-dose concentrations reflect naltrexone and 6-beta-naltrexol plasma concentrations following multiple administration of 50 mg naltrexone oral tablets from study days -4 through -1, which were not expected to influence levels in subsequent samples. Pre-dose plasma concentrations on day 0 were set to O for the calculation of AUCs and were excluded from the reporting of tmax. For naltrexone, mean (range) Cmax value was 2.640 (0.57–11.02) ng/mL. Values for tmax were variable and ranged from 168 hr to 1,008 hr (median: 540 hr) post-dose. In terms of exposure, mean (range) values for AUC0–28 and AUC 0–42 were 829 (149–3,332) ng/hr/mL and 1089 (341–3899) ng/hr /mL, respectively. For the major metabolite, 6-beta-naltrexol, the mean (range) Cmax value was 8.71 (0.99–41.13) ng/mL. Values for tmax ranged from 168 hr to 1,008 hr (median: 336 hr). The mean 6-beta-naltrexol to naltrexone ratios were 3.5 for AUC0–28 and 3.4 for AUC0–42. The mean (range) values for AUC0–28 and AUC0–42 were 2528 (606–8872) ng/hr/mL and 3286 (988–8872) ng/hr/mL, respectively. Plasma concentrations of 6-beta-naltrexol were approximately 3-fold higher than naltrexone at all sampling times.
Fourteen (88%) subjects had detectable plasma concentrations of naltrexone at 42-days post-dose, 1 (6%) subject had a detectable concentration at their last visit 35-days post-dose and 1 (6%) subject had a detectable concentration at their last visit 14-days post-dose. The mean concentrations of naltrexone ranged from 0.58 ng/mL to 2.04 ng/mL at each sampling time after administration; see Figure 3
Figure 3 Serum Naltrexone Levels.
Thirteen (81%) subjects had detectable 6-beta-naltrexol concentrations at 42-days post-dose, 1 (6%) subject had a detectable concentration at their last visit 35-days post-dose, 1 (6%) subject had a detectable concentration at their last visit 14-days post-dose and 1 subject had a detectable concentration through 28 days post dose, after which 6-beta-naltrexol was undetectable. The mean concentrations of 6-beta-naltrexol ranged from 1.51 ng/mL to 5.52 ng/mL at each sampling time after administration; see Figure 4.
Figure 4 6-beta-Naltrexol Levels.
Discussion
The results from this 6-week, single site, open-label study to investigate the safety and pharmacokinetics of Naltrexone Depot showed it to be safe, well tolerated and led to sustained plasma levels of naltrexone. The overall completion rate for study participants was 88%. The 2 subjects who did not complete the study were not lost because of adverse effects.
Injection site complications were by far the most common side effect experienced by subjects participating in this study. Of the side effects experienced the most severe were drainage from the injection site. In the absence of any evidence of infection and given the history of inflammatory reactions from naltrexone injections [30], these cases of drainage were determined to have been caused by a local inflammatory reaction. Although this rate of drainage was concerning, it is important to note that none of the subjects required hospitalization, the drainage was self limiting, and the subjects were remarkable unconcerned about it. Other common side effects were gastrointestinal, nervous system, and changes in metabolic markers, but these occurred at a similar rate as in other naltrexone trials in alcohol and heroin dependent subjects [5,11,31].
The mean plasma concentration across all sampling times ranged from 0.12 ng/mL to 2.04 ng/mL. On the last sampling day (day 42) all subjects available for testing had detectable plasma concentrations of naltrexone. These results are encouraging and show that the depot formulation of naltrexone can provide plasma concentrations of active drug for over 30 days. For a depot formulation of naltrexone to be accepted in clinical practice, a duration of action of at least 1 month is probably necessary [32]. This is the case because it limits the number of visits a patient has to make to their health care provider and limits the number of injections they must receive. Although detectability for the study duration is important, further information is needed on what constitutes an optimal level.
Plasma levels of 6-beta-naltrexol were 3 times higher than naltrexone levels at all sampling times. When naltrexone is administered through the IM route it bypasses 1st pass metabolism, which leads to a lower ratio of 6-beta-naltrexol to naltrexone [33]. It is still unclear how much 6-beta-naltrexol contributes to the efficacy and adverse effects of Naltrexone Depot, but the lower ratio of 6-beta naltrexol to naltrexone may reduce adverse effects which are due to the 6-beta-naltrexol. In a preliminary study looking at the relationship between 6-beta naltrexol levels and the incidence of subjective side effects following oral administration of 50 mg of naltrexone, subjects who experienced more side effects had significantly higher urinary levels of 6-beta-naltrexol [33,34]. These side effects included headache, nausea, anxiety and spontaneous erection. Overall, subjects had a 10:1 urinary ratio of 6-beta-naltrexol to naltrexone. It has been suggested that 6-beta-naltrexol contributes to the continuing antagonism of heroin's effects due to the fact that its blood levels are consistently higher than naltrexone's after depot administration [19]. The pharmacological properties and activity of 6-beta-naltrexol need to be investigated further before a definitive answer can be given on the role of 6-beta-naltrexol in its ability to modulate the response to alcohol and heroin.
Subjects participating in this study experienced a decrease in drinking days, drinks per drinking day, and heavy drinking days. These results are consistent with the study by Kranzler [5](1998) using a different depot formulation of naltrexone. In that study there was a clinically significant effect on the frequency of heavy drinking days. The results from these 2 studies provide the foundation for large multi-site studies on the efficacy and safety of Naltrexone Depot.
Limitations of this study include the fact that this it was small, single site, open-label, and the sample is primarily older and male. The lack of a placebo group makes it difficult to assess the contribution of Naltrexone Depot to the reduction in drinking. Nonetheless, the results from this open-label study are encouraging and warrant further evaluation of Naltrexone Depot as a treatment option for alcohol dependent patients. Approaches such as this that optimize adherence have the potential to maximize the efficacy of naltrexone in the treatment of alcohol dependence.
Competing interests
Funding for this study was provided to Gantt P. Galloway under a contract from DrugAbuse Sciences.
Authors' contributions
GG carried out study procedures, analyzed data, and participated in design of the study. MK carried out study procedures. RC analyzed data. DS carried out study procedures. All authors read and approved the final manuscript.
Figure 2 Adverse Events by Body System.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The guidance of Donald R. Wesson, M.D. in the design of this study and the assistance of Linda Moore in preparing this manuscript are gratefully acknowledged.
==== Refs
O'Malley SS Jaffe AJ Chang G Schottenfeld RS Meyer RE Rounsville B Naltrexone and coping skills therapy for alcohol dependence: a controlled study Arch Gen Psychiatry 1992 49 881 887 1444726
Volpicelli JR Alterman AI Hayashida M O'Brien CP Naltrexone in the treatment of alcohol dependence Arch Gen Psychiatry 1992 49 876 880 1345133
Balldin J Berglund M Borg S Mansson M Bendtsen P Franck J Gustafsson L Halldin J Nilsson LH Stolt G Willander A A 6-month controlled naltrexone study: combined effect with cognitive behavioral therapy in outpatient treatment of alcohol dependence Alcohol Clin Exp Res 2003 27 1142 1149 12878920 10.1097/01.ALC.0000075548.83053.A9
Anton RF Moak DH Waid LR Latham PK Malcolm RJ Dias JK Naltrexone and cognitive behavioral therapy for the treatment of outpatient alcoholics: results of a placebo-controlled trial Am J Psychiatry 1999 156 1758 1764 10553740
Kranzler HR Modesto-Lowe V Nuwayser ES Sustained-release naltrexone for alcoholism treatment: a preliminary study Alcohol Clin Exp Res 1998 22 1074 1079 9726277
Davidson D Swift R Fitz E Naltrexone increases the latency to drink alcohol in social drinkers Alcohol Clin Exp Res 1996 20 732 739 8800392
Davidson D Palfai T Bird C Swift R Effects of naltrexone on alcohol self-administration in heavy drinkers Alcohol Clin Exp Res 1999 23 195 203 10069545
Swift RM Whelihan W Kuznetsov O Buongiorno G Hsuing H Naltrexone-induced alterations in human ethanol intoxication Am J Psychiatry 1994 151 1463 1467 8092339
Kranzler HR Van Kirk J Efficacy of naltrexone and acamprosate for alcoholism treatment: a meta-analysis Alcohol Clin Exp Res 2001 25 1335 1341 11584154 10.1097/00000374-200109000-00014
Streeton C Whelan G Naltrexone, a relapse prevention maintenance treatment of alcohol dependence: a meta-analysis of randomized controlled trials Alcohol Alcohol 2001 36 544 552 11704620
Krystal JH Cramer JA Krol WF Kirk GF Roseheck RA Naltrexone in the treatment of alcohol dependence: a meta-analysis of randemized controlled trials Alcohol Alcohol 2001 36 544 552 11704620
Gastpar M Bonnet U Boning J Mann K Schmidt LG Soyka M Wetterling T Kielstein V Labriola D Croop R Lack of efficacy of naltrexone in the prevention of alcohol relapse: results from a German multicenter study J Clin Psychopharmacol 2002 22 592 598 12454559 10.1097/00004714-200212000-00009
Kranzler HR Modesto-Lowe V Van Kirk J Naltrexone vs. nefazodone for treatment of alcohol dependence. A placebo-controlled trial Neuropsychopharmacology 2000 22 493 503 10731624 10.1016/S0893-133X(99)00135-9
Chick J The efficacy of treatments in reducing alcohol consumption: a meta-analysis Subst Use Misuse 1996 31 1081 8853230
Volpicelli JR Rhines KC Rhines JS Volpicelli LA Alterman AI O'Brien CP Naltrexone and alcohol dependence. Role of subject compliance Arch Gen Psychiatry 1997 54 737 742 9283509
Fuller RK Branchey L Brightwell DR Derman RM Emrick CD Iber FL James KE Lacoursiere RB Lee KK Lowenstam I Disulfiram treatment of alcoholism. A Veterans Administration cooperative study Jama 1986 256 1449 1455 3528541 10.1001/jama.256.11.1449
Cox WM Blount JP Crowe PA Singh SP Diabetic patients' alcohol use and quality of life: relationships with prescribed treatment compliance among older males Alcohol Clin Exp Res 1996 20 327 331 8730225
Johnson DA Observations on the use of long-acting depot neuroleptic injections in the maintenance therapy of schizophrenia J Clin Psychiatry 1984 45 13 21 6143743
Comer SD Collins ED Kleber HD Nuwayser ES Kerrigan JH Fischman MW Depot naltrexone: long-lasting antagonism of the effects of heroin in humans Psychopharmacology (Berl) 2002 159 351 360 11823887 10.1007/s002130100909
Kaempf G O'Donnell C Oslin DW The BRENDA model: a psychosocial addiction model to identify and treat alcohol disorders in elders Geriatr Nurs 1999 20 302 304 10601893 10.1053/gn.1999.v20.103922001
Pettinati HM Volpicelli JR Pierce JDJ O'Brien CP Improving naltrexone response: an intervention for medical practitioners to enhance medication compliance in alcohol dependent patients J Addict Dis 2000 19 71 83 10772604 10.1300/J069v19n01_06
Sobell LC Sobell MB Litten R and Allen J Timeline follow-back: a technique for assessing self-reported alcohol consumption Measuring Alcohol Consumption 1992 Clifton, N.J., The Humana Press, Inc. 41 72
Sobell MB Sobell LC Klajner F Pavan D Basian E The reliability of a timeline method for assessing normal drinker college students' recent drinking history: utility for alcohol research Addict Behav 1986 11 149 161 3739800 10.1016/0306-4603(86)90040-7
Sobell LC Sobell MB Leo GI Cancilla A Reliability of a timeline method: assessing normal drinkers' reports of recent drinking and a comparative evaluation across several populations Br J Addict 1988 83 393 402 3395719
Sobell LC Brown J Leo GI Sobell MB The reliability of the Alcohol Timeline Followback when administered by telephone and by computer Drug Alcohol Depend 1996 42 49 54 8889403 10.1016/0376-8716(96)01263-X
Meyer MC Straughn AB Lo MW Schary WL Whitney CC Bioequivalence, dose-proportionality, and pharmacokinetics of naltrexone after oral administration J Clin Psychiatry 1984 45 15 19 6469932
Wall ME Brine DR Perez-Reyes M Metabolism and disposition of naltrexone in man after oral and intravenous administration Drug Metab Dispos 1981 9 369 375 6114837
Cone EJ Gorodetzky CW Yeh SY The urinary excretion profile of naltrexone and metabolites in man Drug Metab Dispos 1974 2 506 512 4156313
Goodman LS Gilman A Hardman JG Gilman AG Limbird LE Molinoff PB Ruddon RW Goodman & Gilman's the pharmacological basis of therapeutics 1996 9th New York, McGraw-Hill, Health Professions Division xxi, 1905 p.
Sidman KR Schwope AD Steber WD Rudolph SE Use of synthetic polypeptides in the preparation of biodegradable delivery systems for narcotic antagonists NIDA Res Monogr 1981 28 214 231 6791006
Guardia J Caso C Arias F Gual A Sanahuja J Ramirez M Mengual I Gonzalvo B Segura L Trujols J Casas M A double-blind, placebo-controlled study of naltrexone in the treatment of alcohol-dependence disorder: results from a multicenter clinical trial Alcohol Clin Exp Res 2002 26 1381 1387 12351933 10.1097/00000374-200209000-00011
Olsen JL Kincl FA A review of parenteral sustained-release naltrexone systems NIDA Res Monogr 1981 28 187 193 6791004
King AC Volpicelli JR Gunduz M O'Brien CP Kreek MJ Naltrexone biotransformation and incidence of subjective side effects: a preliminary study Alcohol Clin Exp Res 1997 21 906 909 9267542
Verebey K Mule SJ Naltrexone pharmacology, pharmacokinetics, and metabolism: current status Am J Drug Alcohol Abuse 1975 2 357 363 1227297
| 15804355 | PMC1087493 | CC BY | 2021-01-04 16:33:03 | no | BMC Psychiatry. 2005 Apr 1; 5:18 | utf-8 | BMC Psychiatry | 2,005 | 10.1186/1471-244X-5-18 | oa_comm |
==== Front
BMC PsychiatryBMC Psychiatry1471-244XBioMed Central London 1471-244X-5-191581713010.1186/1471-244X-5-19Research ArticleAdd-on topiramate reduces weight in overweight patients with affective disorders: a clinical case series Kirov George [email protected] John [email protected] Department of Psychological Medicine, School of Medicine, Cardiff University, Heath Park, Cardiff, CF14 4XN, UK2005 7 4 2005 5 19 19 30 11 2004 7 4 2005 Copyright © 2005 Kirov and Tredget; licensee BioMed Central Ltd.2005Kirov and Tredget; 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 weight-gain caused by many psychotropic drugs is a major cause for poor compliance with such medications and could also increase cardio-vascular morbidity among psychiatric patients. Recent reports have shown that the anticonvulsant topiramate causes weight loss in various patient groups. The drug has also shown effectiveness in open trials as a mood stabilizer in patients with affective disorders, but not in controlled trials in the acute treatment of mania. We used topiramate to treat 12 patients with affective disorders who had a body-mass index >30 kg/m2.
Methods
Topiramate was prescribed as part of our routine clinical practice, as an add-on medication, or as a replacement of a mood stabilizer. Patients' weight was recorded in 1 to 2 monthly intervals. Patients were followed up for between 6 and 12 months. The final dose of topiramate varied from 200 to 600 mg/day.
Results
Topiramate was effective in reducing the weight in 10 out of the 12 patients. At six months the 12 patients had lost a mean of 7.75 kg (SD = 6.9 kg, p < 0.001) and at 12 months 9 patients had lost a mean of 9.61 kg (SD = 6.7 kg, p = 0.003). Three patients stopped the treatment: one due to side effects, one due to possible side effects, and one suffered a manic relapse and showed no sustained weight loss. There were no other clear changes in the course of illness of the patients.
Conclusion
The evidence of a strong weight-reducing potential of topiramate is indisputable and clinically significant. Topiramate could be considered in the treatment of bipolar patients who are overweight, or whose concerns about weight gain compromise their compliance with long-term prophylactic medication. So far there is no evidence that topiramate has anti-manic effect and it should not be used as monotherapy.
==== Body
Background
Psychiatric patients who receive long-term mood-stabilizing agents can gain excess body weight [1-3]. Weight gain is induced by the majority of antipsychotics [1,4], lithium [2,5,6], antidepressants [7] and valproate [5,8,9]. This could be a major problem with compliance and is one of the main reasons for discontinuation of treatment [1,10]. Increased body weight is also a major health hazard. Mortality rates for both men and women show a strong association with the body mass index (BMI) [11]. The risk starts to increase when the BMI is >25 kg/m2 and doubles for BMI values of >40 kg/m2, where the correlation curve becomes progressively steeper [11]. It has been estimated that obesity causes 300,000 deaths per year in the USA alone [12].
Topiramate is a relatively new antiepileptic drug. Its effects include sodium-channel-blocking activity, enhancement of cerebral GABA concentrations and antagonism of AMPA/kainate receptors, which leads to a decreased glutamate-mediated excitation [13]. It has been shown to cause weight loss in a variety of treatment populations. In an add-on uncontrolled study on 34 epileptic patients there was a 5.9 kg weight loss at 1 year, while among the obese patients in that trial (BMI> = 30 kg/m2) the weight loss reached 10.9 kg [14]. The weight reducing potential of topiramate has been used to treat obesity in subjects with no concomitant psychiatric disorder or epilepsy. Bray et al, [15] treated 385 subjects in a randomised, double-blind, placebo-controlled study and reported weight losses of 4.8%–6.3% of body weight for increasing doses of topiramate treatment. Topiramate was also effective in the treatment of binge-eating disorder. In an open-label study of 13 patients topiramate resulted in 11.8 kg mean weight loss over a period of 3–30 months [16]. McElroy et al, [17] treated 61 outpatients affected with binge-eating associated with obesity for 14 weeks in a placebo-controlled trial. Mean weight loss was 5.9 kg and the binge-eating frequency was also reduced. The beneficial effects on weight and binge-eating were sustained during a 42-weeks open continuation of that trial [18], but a number of patients dropped out from the study. Another double-blind study on 69 patients with bulimia nervosa reported improvements in self-esteem and anxiety, in addition to the behavioural dimensions of bulimia [19].
Several studies have investigated the use of topiramate in bipolar disorders. Marcotte [20] evaluated 58 psychiatric patients, mostly with bipolar affective disorder. They were treated with add-on topiramate for a mean of 16 weeks. 62% were rated improved (weight changes were not reported). McElroy et al, [21] gave topiramate to 56 outpatients, either manic, depressed, or euthymic. The manic patients were reported improved, while the depressed (N = 11) did not change. Significant decreases in weight and BMI were recorded at every follow-up point, reaching 6.0 kg after 1 year. Gupta et al [22] performed a retrospective chart review of 5 patients with bipolar or schizoaffective disorder and reported a mean weight loss of 10 kg. Vieta et al, [23] treated 25 patients with treatment-resistant bipolar spectrum disorders in an open study and reported significant improvements in rating scales for depression and mania. Over 50% of all patients were considered to be responding to topiramate. Ten patients in that study experienced weight loss. Guille and Sachs, [24] treated 14 patients with refractory bipolar illness for an average of 22.4 weeks. Nine of them (64%) improved. Four patients with BMI>28 kg/m2 experienced mean weight loss of 13.5 kg. In an open study topiramate was given together with Risperidone to 58 bipolar patients during a manic episode, of whom 41 completed the 12 months follow-up [25]. A significant improvement in rating scales for mania was recorded and relapse rates were lower compared to the preceding year. The mean weight of patients decreased by 1.1 kg by the end of the study. The same team examined the effect of the co-administration of topiramate and olanzapine over a 12 months interval [26]. Thirteen patients completed 12 months of treatment and their mania and depression ratings improved significantly. At the endpoint they showed 0.5 kg of weight loss, an impressive result given the known problems of weight gain associated with olanzapine [4]. A study using retrospective chart review on patients given topiramate showed a mean loss of 1.2 kg [27]. Topiramate was also used to treat acute manic episodes. In one study 18 patients with manic, mixed, or rapid-cycling episodes, resistant to current mood-stabilizers were given topiramate [28]. There were 12 patients who were judged responders after 5 weeks, and they had a mean weight loss of 4.3 kg. In a study conducted by Grunze et al, [29] 11 patients were given add-on topiramate, of whom 7 showed good improvement. In a study reported by Bozikas et al, [30], 14 manic patients received topiramate either as monotherapy, or in combination with antipsychotics for 4 weeks. Response rate was 61.5% and all patients tolerated topiramate well. Both weight loss and weight gain were reported in that study. Calabrese et al, [31] treated 10 patients hospitalised with mania with topiramate monotherapy for 4 weeks and 8 patients were described as responding. In a retrospective chart review of 76 patients, 13% showed a moderate to marked improvement on the Clinical Global Impression scale (CGI) [32]. There was a weight loss in 50% of patients in that sample, with a mean weight loss of 6.5 kg. A better response and higher weight loss correlated with higher doses of medication. Topiramate has been used in the treatment of the depressive phase of bipolar disorder. McIntyre et al, [33] compared bupropion and topiramate in 36 depressed BP outpatients in an open trial lasting 8 weeks. There was a similarly good improvement in both groups, with a 56% response rate on the Hamilton Rating Scale for Depression in the topiramate group (N = 18). The weight loss in this study was 5.8 kg in the topiramate group and 6 patients withdrew due to side effects. Lykouras and Hatzimanolis, [34] gave topiramate to 56 patients who had suffered relapses of bipolar disorder in the previous 12 months as an add-on treatment. There was a significant reduction of new manic and depressive episodes in the 50 patients who completed 12 months treatment.
All these open studies should be considered in the context of four large unpublished placebo-controlled trials of topiramate for the treatment of acute mania, which produced negative results (cited in Yatham [35]). They indicate clearly that monotherapy with topiramate has no acute anti-manic property. In contrast, almost all of the above open trials used topiramate as an add-on therapy.
Many patients in our Affective Disorders Clinic were concerned with excessive weight gain which they blamed at least partly on their treatments. As weight loss was a listed side effect of topiramate and several small studies had already reported on its use in bipolar affective disorders, we decided to use it as an add-on mood stabilizer in patients who had weight problems. We reasoned that topiramate could also have mood-stabilizing properties, like several other anticonvulsants, a finding already reported in open trials. We also felt that this off-licence use of topiramate was justified, as this intervention enabled patients to comply better with the rest of their medication, thus reducing potential relapses of illness, and that the weight loss could improve their risk for cardio-vascular disorders. We were not aware of the results from the controlled trials until all the patients described here were already taking topiramate.
Patients
Patients in our mood disorders clinic are seen in regular intervals for long-term maintenance of their illness (some have attended for over 7 years now). As part of the service that we provide, patients have regular blood tests, reviews of medication and checks of weight, usually at intervals of 3 or 4 months. Topiramate was given to some obese patients as part of our normal clinical practice, rather than as an experimental intervention (therefore we sought no Ethics committee approval). We decided to publish our results only after the patients had taken topiramate for a considerable time, because we felt that we should share our experience with other health professionals, as the results looked so clear. The only difference to treatment as usual was that we checked the weight of these patients on a more regular basis than we would have normally done. All 12 patients included in this report had a pre-topiramate BMI of >30 kg/m2, i.e. they were all obese according to the National Heart, Lung and Blood Institute guidelines [36]. According to the British National Formulary guidelines [37], an anti-obesity drug should be considered only for those with a BMI of 30 or greater (page 207), and we felt that this recommendation should be followed in the case of Topiramate, even if this drug is not licenced for this use. The mean BMI of patients at the start was 37.95 kg/m2, SD = 5.66 and the mean weight was 104.38 kg, SD = 21.46 kg. The patients had the following diagnoses: Bipolar Affective Disorder type I (BPI) = 5, BPII = 4 and Unipolar Depression (UD) = 3. Four of them were rapid cyclers. Two of the three unipolar patients had co-morbid epilepsy and the third one reached a BMI of nearly 45 (extreme obesity). We felt that it was justified to give her this medication on general health grounds, although benefits from topiramate on unipolar depression had not been demonstrated before. Our clinic treats chronic and treatment resistant patients, and 9 of the 12 patients could be described as treatment resistant or incomplete responders, who had either frequent relapses of mood disorder (more than 2 per year), or remained chronically depressed (i.e. had not reached criteria for remission for a period of at least several weeks over the previous two years) despite adequate treatment. One more overweight patient was prescribed topiramate. He did not like its effect and stopped treatment after one month. He is therefore not included in any analysis.
Concomitant medication
The nine treatment resistant/incomplete responder patients took combinations of medications: all of them took antidepressants, seven took antipsychotics, five took lithium and five took anticonvulsants (Table 1). One other patient was in a full remission on lithium monotherapy and one had recently been discharged from hospital after a manic relapse, having been relatively stable for several years prior to that. In four cases an anticonvulsant was exchanged with topiramate, in the remaining cases topiramate was given as an add-on treatment. No one received monotherapy of topiramate.
Table 1 Characteristics of the patients and changes of their weight during the study. Both individual and mean weight changes are presented. p-values are based on paired samples tests for the patients who have completed the corresponding follow-up interval. The dose of topiramate is the highest dose reached. Under "Concomitant medication" the drugs that are undelined were exchanged with topiramate.
Patient N Gender Age DSM-IV Concomitant medication Topiramate dose (mg) Start Weight change from start (months)
BMI (kg/m2) weight (kg) 3 months 6 months 9 months 12 months
1 f 49 UD Nortryptiline, carbamazepine 600 42.4 110 -8 -10 -11 -12.5
2 f 48 BPI Risperidone, lamotrigine, lofepramine 400 37.2 87 -9.5 -14 -12 -8
3 f 49 BPI Quetiapine, mirtazapine, lithium, venlafaxine 200 42.3 99 -1 -5 -11 -14
4 m 50 BPII Lithium, paroxetine 300 47.7 146 -10.5 -21 -26 -12.5
5 f 28 BPI Lithium 200 31.5 79 -6 -10.5
6 m 31 BPII Lithium, reboxetine, trazodone, valproate 300 40.5 120 -6 -7.5 -6.5 -8
7 m 40 BPI Olanzapine, fluoxetine 400 30.3 102.5 -1 +5 -1.5 -5.5
8 f 41 BPII Fluphenazine, carbamazepine, paroxetine 250 34.7 90 -7 -14 -18 -22.5
9 f 70 UD Mirtazapine, lithium, amisulpride, venlafaxine 200 44.4 116.5 -3 -4 -4 -4
10 f 54 BPII Lithium, clomipramine, flupenthixol, trazodone 250 34.7 80 -1.5 -2.5 -2 +0.5
11 f 48 UD Phenelzine, amisulpride, valproate 250 31.0 88.5 -3.5 -8.5
12 m 33 BPI Lithium, carbamazepine 200 38.7 134 -2.5 -1
Mean reduction (kg) 4.96 7.75 10.2 9.61
SD 3.33 6.9 7.98 6.69
p-value 0.001 0.003 0.005 0.003
Topiramate dosage
Topiramate dose was increased in steps of 25–50 mg every one to two weeks. After some of the first patients complained of paraesthesia and headaches, we routinely started making the increases only at 2-weekly intervals. The final dose of topiramate ranged between 200 and 600 mg, with a mean of 296 mg. The dose was increased until we could see a clear effect on the weight, or the patient reported some side effects. The patient whose topiramate was increased to 600 mg/day suffers with poorly controlled epilepsy and we increased the dose in order to achieve a better seizures control, rather than weight loss.
Statistical analysis
The level of statistical significance for weight loss at each time-point during the follow-up, compared to baseline, was determined with a paired-samples t-test. This test assumes that comparisons at the four time points are independent, whreas the observations within individuals are, of course correlated. Therefore the test for statistical significance is slightly anti-conservative.
Results
Weight changes
Patients have received topiramate for between 6 months and one year by the time of writing this report. Six of them are still continuing the treatment. The weight changes are shown in Figure 1 and Table 1. All post-treatment levels and the pre-treatment levels from the previous two years, where available, are shown in Figure 1. At 3 and 6 months, 12 patients lost a mean of 4.96 and 7.75 kg respectively. The nine patients who completed 12 months treatment showed a weight loss of 10.2 and 9.61 kg at the 9th and 12th month respectively. All these changes are significant at p < = 0.005.
Figure 1 Weight changes in the patients described in the study. The day of treatment is on the x-axis. The day when topiramate was started is day 0 and weight measurements before that point are supplied where available. Each patient's number as referred to in the text is given on the right of the figure.
Side effects
One patient (N 5) noticed thinning of her hair and stopped the topiramate after six months. This side effect is known with other mood-stabilizers such as valproate. If true, it is likely to be quite rare, as it has not been mentioned in previous papers, or listed as a side effect. We submitted a report for a suspected adverse drug reaction to the Committee on Safety of Medicines in the UK. One patient (N 10) complained of paraesthesia, sore tongue and bad taste in her mouth and stopped topiramate during an earlier treatment (on Figure 1 this coincides with an earlier short period of weight loss). She restarted it 7 months later (as her weight had increased again by 5.5 kg). This time the medication was titrated more slowly by 25 mg/day every two weeks and she did not experience these side effects. This second start of treatment with topiramate is taken as day 0 for the purposes of this report, as it was of a longer duration and is still on-going. One patient (N 2) complained of paraesthesia, later developed unsteadiness of gait and finally severe Parkinsonism, therefore most of her medications, including topiramate were stopped. This patient was taking a number of psychotropic and cardiovascular drugs (a total of nine!), so we do not know whether topiramate was to blame for her neurological signs, especially as she had been taking this drug for over one year already. She recovered completely around 4 months later and no cause for her condition was identified, despite numerous investigations and consultations. We cannot exclude the possibility that topiramate was at least partially responsible, alone, or in combination with several other drugs, therefore we report this event here. Two patients complained of memory and concentration problems but preferred to continue with the medication despite this. These side effects are also well recognised. One patient did not like the appetite suppressing effect of topiramate and stopped it after only one month, so his data is not analysed in the study.
Clinical changes
We do not perform regular formal ratings of the mood of our patients. However we regularly record their condition and changes since the previous appointment. Clear improvement of the course of illness was noticed only in one patient (N 3), while eight patients continued to display a similar pattern of illness, as they had done before the treatment. None of them was judged to have deteriorated. The patients in remission remained well. One patient (N 12) whose carbamazepine was changed to topiramate developed a manic episode, therefore we stopped his topiramate after 6 months and asked him to resume carbamazepine.
Discussion
This is a naturalistic case-series and the absence of a placebo control group could have influenced the results. On the other hand, these patients had already made unsuccessful efforts to lose weight and, as can be seen on Figure 1, their pre-topiramate weight had been either stable, or in most cases had increased steadily during the period of observation. The weight-reducing potential of topiramate was impressive in 10 of the 12 patients and the effect was maintained for the duration of the observation, in most cases for more than one year. These results are in keeping with a large body of evidence supporting the weight-losing potential of topiramate, as presented in the Background. In fact, we could not find a single study that reported a mean weight gain with topiramate. The weight loss in our patients was present already at the third month (mean = 4.96 kg, SD = 3.33, p < 0.001) and continued to increase at least until the 9th month. Only one patient (N 10) had a weight gain or 0.5 kg at the end of the observation which still represented an improvement on her tendency to gain weight in the previous years.
Obesity is a very common problem in developed countries. Its prevalence in the USA is rising and has been estimated at 22.5% among adults aged 20 and over in the year 2001 [38]. Weight gain is even more prevalent in bipolar disorder patients [2]. The increased weight in this population is due to a number of factors, including life-style and diet, but a major cause is the side effect profile of the medications that are prescribed long-term for their illness. If deaths due to suicide and accidents are excluded, there is still a substantially increased mortality in patients with mood disorders, which is due mostly to circulatory disorders [39,40]. Increased weight, high levels of smoking and reduced exercise are likely to be the main factors leading to such an increased mortality. Health professionals cannot just give advice on diet and exercise to overweight patients, because the weight gain is to a large extent caused by the more sedentary life imposed by the illness, combined with the side effects of the majority of drugs they are prescribed. Psychiatrists should have a responsibility in managing obesity in their patients by choosing more appropriate drugs in patients prone to weight gain. This will bring several benefits: 1) It will improve the quality of life of their patients, as weight gain is among the most distressing side effects of psychotropic drugs [41]. 2) A reduction in weight could increase their life-expectancy, as weight loss can reduce mortality [1,11]. 3) An improved weight can increase the adherence to long-term treatment with mood stabilizers.
Several studies, including this one, show that topiramate has a strong potential to induce weight loss which is sustained for at least one year. Despite the side effects reported in this and other studies, topiramate is generally well tolerated. Most of the patients in our series who complained of side effects, still preferred to continue the treatment, as they liked its overall effect. In our experience it is very unusual to have such a high rate of patients who continue to take a new medication. The effect on memory and concentration should however receive closer examination in future studies.
The main unanswered question is whether topiramate has any mood-stabilizing properties, like other anticonvulsants such as carbamazepine, valproate and lamotrigine. Open studies for maintenance treatment in bipolar disorder and in the acute treatment of mania have so far been encouraging (as reviewed in the Background). However, four large unpublished placebo-controlled monotherapy trials failed to confirm the efficacy of topiramate in the treatment of acute mania (cited in [35]), indicating a poor antimanic effect, at least in monotherapy. Our work was not designed to examine any mood-stabilizing properties of topiramate, as we did not use regular mood charts, the period of 12 months is too short to evaluate long-term benefits on the pattern of episodes, and there was no comparison group. In addition, many of our patients were already talking a combination of mood stabilizers without complete treatment effect, having failed to achieve a long-term stability on a variety of treatments that had been tried over the years. In view of that, we did not expect any clear improvements in the mood or course of illness of this patient population. All we can conclude so far is that topiramate does not appear to be superior to other mood stabilizers in the long-term treatment of patients with treatment resistant affective disorders. One of our patients had a manic relapse after we changed his carbamazepine with topiramate (he was also receiving regular lithium prophylaxis). In addition, we would like to report that topiramate was given to two more patients in our clinic who had a BMI of <30 kg/m2 and are therefore not presented as part of this case-series. These two patients had refused to take other mood-stabilizing drugs for fear of weight gain and only agreed to take topiramate as monotherapy, having heard that it will not cause weight gain. Both of them suffered relapses of mania, which necessitated the addition of atypical antipsychotics and admission to hospital. Although these three observations are anecdotal, they strengthen the impression that topiramate monotherapy does not have antimanic effect. Whether topiramate has any effect on the depressive side of bipolar illness is not yet know.
Conclusion
Topiramate appears a very useful drug for weight reduction in patients with bipolar disorders. The mood stabilizing effect of topiramate is however questionable. The very different results obtained from the controlled monotherapy trials of topiramate in mania and the open add-on trials summarised in the Background indicate that topiramate should not be prescribed as monotherapy in bipolar disorders as it has no acute antimanic effect. In our opinion the current place of topiramate in the treatment of affective disorders is as an add-on treatment for patients who experience clinically significant weight gain, which could either compromise their physical health, or influence them to stop taking established mood stabilizers.
Recommendations
Topiramate should not be used in monotherapy and unless new research shows otherwise, psychiatrists should assume that it has no mood-stabilizing properties.
Dose titration
Topiramate dose should be increased gradually and in two-weekly intervals.
Cognitive function
Topiramate is known to cause neurocognitive side effects. Psychiatrists should monitor their patients carefully for the emergence of such side effects.
Pregnancy
As with most other mood-stabilizers (anticonvulsants and lithium), women who may become pregnant should be warned of the possible consequences of taking such drugs during pregnancy [37].
Adequate hydration
Patients should be advised to ensure adequate hydration, especially if they are predisposed to nephrolithiasis [37].
Competing interests
G. Kirov has received a grant from the Janssen Research Foundation for the collection of families with psychiatric disorders in Bulgaria for genetic studies. The current work is completely independent of that study, we have not informed the company of our work with topiramate, and have not shown them the manuscript. The Mood Disorders Clinic in Cardiff does not receive financial or any other support from Janssen-Cilag (Janssen-Cilag is the producer of topiramate). G. Kirov has received honoraria for lecturing from Eli Lilly and AstraZeneca and grant support from the Wellcome Trust, MRC, London, and Sanofi Synthelabo. J. Tredget's post is funded by AstraZeneca. None of the financial support received by the authors has been in any way related with the work described in this paper.
Authors' contributions
GK was the psychiatrist in charge for the patients reported in this study. He performed most of the analyses and wrote the draft of the paper.
JT is a Research Nurse who was closely involved in the day-to-day monitoring and supervision of the patients, performed some of the assessments, and organised the running of the clinic, thus contributing significantly towards data acquisition. He revised critically the paper and contributed towards the analysis and discussion.
Pre-publication history
The pre-publication history for this paper can be accessed here:
==== Refs
Aronne LJ Segal KR Weight gain in the treatment of mood disorders J Clin Psychiatry 2003 64 22 29 12892538
Keck PE JrMcElroy SL Bipolar disorder, obesity and pharmacotherapy-associated weight gain J Clin Psychiatry 2003 64 1426 1435 14728103
Nemeroff CB Safety of available agents used to treat bipolar disorder: Focus on weight gain J Clin Psychiatry 2003 64 532 539 12755655
Allison DB Mentore JL Heo M Chandler LP Cappelleri JC Infante MC Weiden PJ Antipsychotic-induced weight gain: a comprehensive research synthesis Am J Psychiatry 1999 156 1686 1696 10553730
Baptista T Teneud L Contreras Q Alastre T Burguera JL de Burguera M de Baptista E Weiss S Hernandez L Lithium and body weight gain Pharmacopsychiatry 1995 28 35 44 7624385
Silverstone T Romans S Long term treatment of bipolar disorder Drugs 1996 51 367 382 8882376
Fava M Weight gain and antidepressants J Clin Psychiatry 2000 61 37 41 10926053
Silberstein SD Collins SD for the Long-Term Safety of Depakote in Headache Prophylaxis Study Group Safety of divalproex sodium in migraine prophylaxis: an open-label, long-term study Headache 1999 39 633 643 11284461 10.1046/j.1526-4610.1999.3909633.x
Jallon P Picard F Body weight gain and anticonvulsants: a comparative review Drug Safety 2001 24 969 978 11735653
Sachs GS Guille C Weight gain associated with use of psychotropic medications J Clin Psychiatry 1999 60 16 19 10548137
Bray G Bouchard C Obesity King RA, Rotter JI, Motylsky AG The Genetic Basis of Common Diseases 2002 Oxford University Press, Oxford
Allison DB Fontaine KR Manson JE Stevens J VanItallie TB Annual deaths attributable to obesity in the United States JAMA 1999 282 1530 1538 10546692 10.1001/jama.282.16.1530
Suppes T Review of the use of topiramate for treatment of bipolar disorders J Clin Psychopharmacol 2002 22 599 609 12454560 10.1097/00004714-200212000-00010
Ben-Menachem E Axelsen M Johanson EH Stagge A Smith U Predictors of weight loss in adults with topiramate-treated epilepsy Obes Res 2003 11 556 562 12690085
Bray GA Hollander P Klein S Kushner R Levy B Fitchet M Perry BH A 6-month randomised, placebo-controlled, dose-ranging trial of topiramate for weight loss in obesity Obes Res 2003 11 722 733 12805393
Shapira NA Goldsmith TD MsElroy SL Treatment of binge-eating disorder with topiramate: a clinical case series J Clin Psychiatry 2000 61 368 372 10847312
McElroy SL Arnold LM Shapira NA Keck PE JrRosenthal NR Karim MR Kamin M Hudson JI Topiramate in the treatment of binge eating disorder associated with obesity: a randomsied, placebo-controlled trial Am J Psychiatry 2003 160 255 261 12562571 10.1176/appi.ajp.160.2.255
McElroy SL Shapira NA Arnold LM Keck PE Rosenthal NR Wu SC Capece JA Fazzio L Hudson JI Topiramate in the long-term treatment of binge-eating disorder associated with obesity J Clin Psychiatry 2004 65 1463 9 15554757
Hedges DW Reimherr FW Hoopes SP Rosenthal NR Kamin M Karim R Capece JA Treatment of bulimia nervosa with topiramate in a randomized, double-blind, placebo-controlled trial, part 2: improvement in psychiatric measures J Clin Psychiatry 2003 64 1449 1454 14728106
Marcotte D Use of topiramate, a new anti-epileptic as a mood stabilizer J Affect Disord 1998 50 245 251 9858083 10.1016/S0165-0327(98)00110-4
McElroy SL Suppes T Keck PE Frye MA Denicoff KD Altshuler LL Brown ES Nolen WA Kupka RW Rochussen J Leverich GS Post RM Open-lable adjunctive topiramtae in the treatment of bipolar disorders Biol Psychiatry 2000 47 1025 1033 10862801 10.1016/S0006-3223(99)00316-9
Gupta S Masand PS Frank BL Lockwood KL Keller PL Topiramate in Bipolar and Schizoaffective Disorders: Weight Loss and Efficacy Prim Care Companion J Clin Psychiatry 2000 2 96 100 15014655
Vieta E Torrent C Garcia-Ribas G Gilabert A Garcia-Pares G Rodriguez A Cadevall J Garcia-Castrillon J Lusilla P Arrufat F Use of topiramate in treatment-resistant bipolar spectrum disorders J Clin Psychopharmacol 2002 22 431 435 12172346 10.1097/00004714-200208000-00017
Guille C Sachs G Clinical outcome of adjunctive topiramate treatment in a sample of refractory bipolar patinets with comorbid conditions Prog Neuropsychopharmacol Biol Psychiatry 2002 26 1035 1039 12452523 10.1016/S0278-5846(01)00278-0
Vieta E Goikolea JM Olivares JM Gonzalez-Pinto A Rodriguez A Colom F Comes M Torrent C Sanchez-Moreno J 1-year follow-up of patients treated with risperidone and topiramate for a manic episode J Clin Psychiatry 2003 64 834 839 12934986
Vieta E Sanchez-Moreno J Goikolea JM Colom F Martinez-Aran A Benabarre A Corbella B Torrent C Comes M Reinares M Brugue E Effects on weight and outcome of long-term olanzapine-topiramate combination treatment in bipolar disorder J Clin Psychopharmacol 2004 24 374 8 15232327 10.1097/01.jcp.0000130556.01373.45
Chengappa KN Chalasani L Braar JS Parepally H Houck P Levine J Changes in body weight and body mass index among psychiatric patients receiving lithium, valproate, or topiramate: an open-label nonrandomised chart review Clin Ther 2002 24 1576 1584 12462287 10.1016/S0149-2918(02)80061-3
Chengappa KN Rathore D Levine J Atzert R Solai L Parepally H Levin H Moffa N Delaney J Brar JS Topiramate as add-on treatment for patients with bipolar mania Bipolar Disord 1999 1 42 53 11256656 10.1034/j.1399-5618.1999.10111.x
Grunze HC Normann C Langosch J Schaefer M Amann B Sterr A Schloesser S Kleindiest N Walden J Antimanic efficacy of topiramate in 11 patients in an open trial with an on-off-on design J Clin Psychiatry 2001 62 464 468 11465524
Bozikas VP Petrikis P Kourtis A Youlis P Karavatos A Treatment of acute mania with topiramate in hospitalised patients Prog Neuropsychopharmacol Biol Psychaitry 2002 26 1203 1206 10.1016/S0278-5846(01)00323-2
Calabrese JR Keck PE JrMcElroy SL Shelton MD A pilot study of topiramate in the treatment of acute mania J Clin Psychopharmacol 2001 21 340 342 11386499 10.1097/00004714-200106000-00015
Ghaemi SN Manwani SG Katzow JJ Ko JY Goodwin FK Topiramate treatment of bipolar disorders: a retrospective chart review Ann ClinPsychiatry 2001 13 185 189 10.1023/A:1014627001201
McIntyre RS Mancini DA McCann S Srinivasan J Sagman D Kennedy SH Topiramate versus bupropion SR when added to mood stabiliser therapy for the depressive phase of bipolar disorder: a preliminary single-blind study Bipolar Disord 2002 4 207 213 12180276 10.1034/j.1399-5618.2002.01189.x
Lykouras L Hatzimanolis J Adjunctive topiramate in the maintenance treatment of bipolar disorders: an open-label study Curr Med Res Opin 2004 20 843 7 15200741
Yatham LN Newer Anticonvulsants in the treatment of bipolar disorder J Clin Psychiatry 2004 65 28 35 15242330
National Heart Lung and Blood Institute (NHLBI) Obesity Task Force Clinical guidelines on the identification, evaluation, and treatment of overweight and obesity in adults: the evidence report Obes Res 1998 6 51 209
British Medical Association & Royal Pharmaceutical Society of Great Britain British National Formulary (BNF 48) BMA & RPSGB, London 2004
National Center for Health statistics National Health Interview Survey (NHIS)
Aagaard J Vestergaard P Predictors of outcome in prophylactic lithium treatment: a 2-year prospective study J Affect Disorders 1990 18 259 266 2140378 10.1016/0165-0327(90)90077-L
Angst F Stassen HH Clayton PJ Angst J Mortality of patients with mood disorders: follow-up over 34–38 years J Affect Disord 2002 68 167 181 12063145 10.1016/S0165-0327(01)00377-9
Weiden PL Miller AL Which side effects really matter? Screening for common and distressing side effects of antipsychotic medications J Psychiatr Pract 2001 7 41 47 15990500 10.1097/00131746-200101000-00005
| 15817130 | PMC1087494 | CC BY | 2021-01-04 16:33:03 | no | BMC Psychiatry. 2005 Apr 7; 5:19 | utf-8 | BMC Psychiatry | 2,005 | 10.1186/1471-244X-5-19 | oa_comm |
==== Front
BMC SurgBMC Surgery1471-2482BioMed Central London 1471-2482-5-71580435210.1186/1471-2482-5-7Research ArticleAdequate symptom relief justifies hepatic resection for benign disease Fioole Bram [email protected] Marike [email protected] Hillegersberg Richard [email protected] Rinkes Inne HM [email protected] Department of Surgery (G.04.228), University Medical Centre Utrecht, PO Box 85500, 3508 GA Utrecht, The Netherlands2005 1 4 2005 5 7 7 11 1 2005 1 4 2005 Copyright © 2005 Fioole et al; licensee BioMed Central Ltd.2005Fioole 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 was to evaluate the long-term results of partial liver resection for benign liver lesions.
Methods
All patients operated on for benign liver lesions from 1991 to 2002 were included. Information was retrieved from medical records, the hospital registration system and by a telephonic questionnaire.
Results
Twenty-eight patients with a median age of 41 years (17–71) were operated on (M/F ratio 5/23). The diagnosis was haemangioma in 8 patients, FNH in 6, HCA in 13 and angiomyolipoma in 1. Eight patients were known to have relevant co-morbidity. Median operating time was 207 minutes (45–360). The morbidity rate was 25% and no postoperative mortality was observed. Twenty-two patients (79%) had symptoms (mainly abdominal pain) prior to surgery. Twenty-five patients were reached for a questionnaire. The median follow up was 55 months (4–150). In 89% of patients preoperative symptoms had decreased or disappeared after surgery. Four patients developed late complications.
Conclusion
Long-term follow up after liver surgery for benign liver lesions shows considerable symptom relief and patient satisfaction. In addition to a correct indication these results justify major surgery with associated morbidity and mortality.
==== Body
Background
Partial liver resection is an accepted treatment for primary and secondary malignancies of the liver. In experienced hands this operation is associated with mortality rates of less than 5% and morbidity of approximately 30% [1-4].
Unlike malignant liver tumours, the indication for resection of benign hepatic lesions, including haemangiomas, focal nodular hyperplasia (FNH) and hepatocellular adenomas (HCA) remains controversial [5-7]. Indications for resection of benign liver masses include: 1) severe or progressive symptoms, 2) uncertain diagnosis with a suspicion for malignancy, and 3) risk of haemorrhage or rupture. If possible, it is important to discern whether the presenting symptoms are due to the liver lesion detected, before proceeding with surgical intervention. Several studies have reported about the indication for surgery in various benign hepatic tumours. However, less is known about the long-term results of surgical treatment, particularly regarding symptom relief.
The present study was undertaken to evaluate the long-term results of partial liver resections for benign liver lesions, with emphasis on the course of symptomatology, long-term complication rate, and patient satisfaction.
Methods
All patients treated by partial liver resection for benign lesions in the University Medical Centre Utrecht between January 1991 and December 2002 were included. Information about these patients was retrieved from medical records and the hospital registration system.
Preoperative parameters consisted of age, sex, diagnosis, co-morbidity, presenting symptoms and indication for resection. In the preoperative work up we have routinely performed physical diagnostic investigation, ultrasound and computed tomography (CT) to exclude other pathology causing the symptoms. On indication additional gastroscopy or endoscopic retrograde cholangiopancreatography was performed. The indications for resection of a haemangioma were persisting symptoms and rapid growth. After exclusion of other pain aetiology a period of 3 months observation was allowed to asses the persistence of the symptoms. In case of FNH the indications for resection were persisting symptoms and to exclude malignancy. HCA's were resected if symptomatic or larger than 5 cm.
Date of resection, extent of resection, number of perioperative blood transfusions and duration of resection were considered perioperative parameters. A major resection was defined as a resection of three segments or more. Perioperative blood transfusion was defined as at least one unit of packed cells infused within 24 hours after surgery.
Documented postoperative parameters consisted of clinically relevant complications, postoperative mortality, duration of admission and follow up. Postoperative mortality was defined as in-hospital mortality. Long-term follow up was obtained by a telephonic questionnaire. Information was collected about the presenting symptoms, the relief of these symptoms after surgery and the impact on physical and social activities. The limitations on physical and social activities as a result of the presenting symptoms were divided in severe, moderate and none. Moderate limitations were defined as limitations for activities at least once a week, while severely limited patients experienced daily restrictions. Symptom relief was defined as a decrease or absence of the presenting symptoms after surgery.
Results
A total of 28 patients were operated on for benign liver lesions. The diagnosis was a haemangioma in 8 patients (29%), FNH in 6 (21%), HCA in 13 (46%) and angiomyolipoma in 1 (4%). The group consisted of 23 female and 5 male patients with a median age of 41 years (range 17–71). Eight patients (29%) were known to have relevant co-morbidity, including severe cardiac or pulmonary diseases, diabetes mellitus, hepatitis, liver fibrosis, adipositas and multiple sclerosis.
Twenty-two patients were known to have symptoms prior to surgery (79%). The most frequent presenting symptom was upper abdominal pain (64%). Other presenting symptoms are shown in table 1. The most important indications for resection were symptoms and excluding malignancy (table 2). The symptoms mentioned in this table consisted of abdominal pain in all patients. Three patients with HCA presented with haemorrhage as a result of spontaneous rupture. One patient was immediately operated on. The other 2 patients were stabilized and resection was performed after the haematoma was resolved.
Table 1 Predominant presenting symptoms in 28 patients
Symptom Haemangioma (n = 8) FNH (n = 6) HCA (n = 13) Angiomyolipoma (n = 1) Total (n = 28)
Abdominal pain 7 4 7 0 18
Swelling 1 1 0 1 3
Nausea 0 0 1 0 1
No symptoms 0 1 5 0 6
Table 2 Indication for resection in 28 patients
Indication Haemangioma (n = 8) FNH (n = 6) HCA (n = 13) Angiomyolipoma (n = 1) Total (n = 28)
Symptoms 7 1 3 0 11
Excluding malignancy 0 5 5 1 11
Haemorrhage 0 0 3 0 3
Risk of malignant degeneration 0 0 2 0 2
Size 1 0 0 0 1
All operations concerned partial liver resections, of which 11 resections were major (39%) (Table 3). The median operating time was 207 minutes (range 45–360). Eighteen procedures necessitated perioperative blood transfusions (median 2, range 0–16).
Table 3 28 Partial liver resections
Extent of resection
Diagnosis Patients L Hemi R Hemi Ext R Segm
Haemangioma 8 1 1 1 5
FNH 6 0 2 0 4
HCA 13 0 4 0 9*
Angiomyolipoma 1 1 0 0 0
All 28 2 7 1 18
L Hemi = Left hemihepatectomy
R Hemi = Right hemihepatectomy
R Ext = Extended right hemihepatectomy
Segm = Segmentectomy
* One patient treated with a central resection segment 4, 5 and 8
One patient underwent extended left hemihepatectomy for an asymptomatic, but rapidly growing giant haemangioma (25 cm). This major resection was accompanied by massive intraoperative haemorrhage from direct venous hepatocaval branches (7500 ml). After the resection the transected surface kept oozing. Therefore haemostasis was obtained by packing with gauzes. About 36 hours later the gauzes were removed. Further postoperative recovery was uneventful. A second patient underwent reoperation for wound dehiscence. The median hospital stay was 9.5 days (range 5–39). Seven patients developed one or more postoperative complications resulting in a morbidity rate of 25% (table 4). No postoperative mortality was observed.
Table 4 Seven patients with 8 postoperative complications
Complication Patients
Wound infection 2
Urinary tract infection 2
Bile leakage 1
Severe ascites 1
Pneumonia 1
Deep venous thrombosis 1
Total 8
During follow up one patient died as a result of a cause other than liver surgery. Three months after a left hemihepatectomy for a large tumour which turned out to be a angiomyolipoma this patient died of cerebral stroke. Of the remaining 27 patients we were able to contact 25 for a questionnaire (figure 1). The median follow up of the interviewed patients was 55 months (range 4–150). Six of the 25 interviewed patients had no symptoms prior to surgery and underwent resection because of an uncertain diagnosis of an incidentally discovered liver lesion. Before surgery 7 patients were severely limited in their physical activities as a result of the liver lesion, 5 were moderately limited and 13 were not limited. Considering social activities 4 patients were severely limited prior to surgery, 4 were moderately limited and 17 were not limited. After surgery the symptoms had decreased or disappeared in 17 of the 19 interviewed patients with preoperative symptoms (89%). In two patients the symptoms were unchanged. One patient underwent partial liver resection for an adenoma. After the resection the preoperative abdominal pain never decreased. In a second patient preoperative abdominal pain did not decrease after resection of a large haemangioma. In these two cases the preoperative symptoms were probably not related to the liver lesion. Four of the 25 interviewed patients had developed late complications as a result of the operation. These complications consisted of a hypertrophic scar (n = 2) and incisional hernia (n = 2). Twenty-three of 25 interviewed patients were satisfied with the result of the resection (92%).
Figure 1 Patients summary
Discussion
Symptom relief was observed in 89% of the patients (17/19), while in two patients the preoperative symptoms had not decreased. Only a few studies concerning long-term follow up of resections for benign liver tumours have been published. Terkivatan et al. described 80% symptom relief in surgically treated patients [5]. In this study surgery was only indicated in case of suspicion of malignancy, severe or increasing symptoms or a HCA larger than 5 cm. They compared this group retrospectively with patients treated conservatively, of whom 34% presented with symptoms, e.g. non-specific complaints of fatigue and mild abdominal pain considered unrelated to the tumour. During long-term follow up 87% symptom relief was observed in the group who was treated conservatively. Charny et al. registered 93% symptom regression in patients who underwent partial liver resection for benign liver tumours and 86% symptom regression in patients who were observed for symptoms considered unrelated to the tumour or treated for unrelated conditions [8]. Therefore partial liver resection for a symptomatic liver lesion should only be performed when symptoms are most likely related to the lesion.
Because of the nature of benign liver tumours, clear indications are needed for partial liver resection, an operation associated with substantial postoperative morbidity and mortality. Indications for resection of a cavernous haemangioma of the liver are the development of complications, rapid growth, the presence of persisting symptoms or the need to establish a confident diagnosis. The potential for complications of a liver haemangioma (mainly rupture) is not an indication for resection of all liver haemangiomas. Spontaneous rupture of a haemangioma is infrequent and could be controlled with transcatheter hepatic artery embolization prior to resection [9]. As for rapid growth, we have operated on 1 patient with an asymptomatic, but rapidly growing haemangioma in our series. Little is known about the natural history of these large haemangiomas and resection can be very challenging, since they are at risk for massive intraoperative haemorrhage. In case of abdominal pain the main challenge is to determine whether these symptoms are due to the haemangioma or an associated condition [10-12]. Farges et al. described other pain aetiology in 54% of patients with symptomatic haemangiomas [7]. Gandolfi et al. observed only 7% symptomatic giant haemangiomas in their series [13]. Haemangiomas are not likely to cause diagnostic uncertainty. In our series we have not performed diagnostic resections for haemangiomas.
Unlike FNH, HCA is often symptomatic and is noted for its spontaneous rupture and malignant transformation. Looking at the potential complications, HCA's with a diameter of more than 5 cm should be resected, while for smaller HCA's and FNH observation is justifiable [6,14-16]. Increasing size on radiographic imaging during observation is an indication for resection. FNH and HCA occur predominantly in females and are associated with long-term contraceptive steroid use [17]. This medication should be stopped, when FNH or a small HCA is not treated surgically. In case of FNH and HCA the most frequent indication for resection is the uncertain diagnosis of a hepatic mass with suspicion for malignancy. In addition to ultrasound, CT and MRI, positron emission tomography has proved to be a helpful modality distinguishing between benign and malignant liver lesions [18,19]. On the other hand, the use of needle biopsy should be reserved for atypical cases, since the limited material is rarely sufficient to exclude malignancy. In our series diagnostic uncertainty accounted for operation in 11 of 20 resected patients and was, together with symptoms, the main indication for resection. The final diagnosis after pathological examination of resected specimens was FNH in 5 patients, HCA in 5 and angiomyolipoma in 1. All HCA's were larger than 5 cm, but no malignancies were observed.
Conclusion
We have shown in this consecutive series that partial liver resection for benign disease is a very effective procedure to relief invalidating abdominal symptoms. Benign liver lesions should only be resected when symptoms are most likely related to the lesion. In experienced centres the resection can be performed with acceptable morbidity and low mortality.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
BF designed the study, helped with the data acquisition and drafted the manuscript. MK carried out the data acquisition and participated in the study design. RH participated in the design of the study and in drafting and revising the manuscript. IBR conceived of the study and coordinated the draft and revision of the manuscript. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
==== Refs
Jarnagin WR Gonen M Fong Y DeMatteo RP Ben-Porat L Little S Corvera C Weber S Blumgart LH Improvement in perioperative outcome after hepatic resection: Analysis of 1,803 consecutive cases over the past decade Ann Surg 2002 236 397 407 12368667 10.1097/00000658-200210000-00001
Belghiti J Hiramatsu K Benoist S Massault PP Sauvanet A Farges O Seven hundred forty-seven hepatectomies in the 1990s: an update to evaluate the actual risk of liver resection J Am Coll Surg 2000 191 38 46 10898182 10.1016/S1072-7515(00)00261-1
Scheele J Altendorf A Resection of colorectal liver metastases Langenbeck's Arch Surg 1999 384 313 327 10473851 10.1007/s004230050209
Fioole B Liem MS Hennipman A Borel Rinkes IH Partial liver resections: mortality, morbidity and risk factors for postoperative complications in 133 patients/137 operations; Utrecht University Medical Center 1991/2000 Ned Tijdschr Geneeskd 2002 146 210 213 11851082
Terkivatan T de Wilt JH de Man RA van Rijn RR Zondervan PE Tilanus HW IJzermans JN Indications and long-term outcome of treatment of benign hepatic tumors. A critical appraisal Arch Surg 2001 136 1033 1038 11529826 10.1001/archsurg.136.9.1033
Nagorney DM Benign hepatic tumors: Focal nodular hyperplasia and hepatocellular adenoma World J Surg 1995 19 13 18 7740799 10.1007/BF00316973
Farges O Daradkeh S Bismuth H Cavernous hemangiomas of the liver: Are there any indications for resection? World J Surg 1995 19 19 24 7740805 10.1007/BF00316974
Charny CK Jarnagin WR Schwartz LH Frommeyer HS DeMatteo RP Fong Y Blumgart LH Management of 155 patients with benign liver tumors Br J Surg 2001 88 808 813 11412249 10.1046/j.0007-1323.2001.01771.x
Yamamoto T Kawarada Y Yano T Noguchi T Mizumoto R Spontaneous rupture of hemangioma of the liver: treatment with transcatheter hepatic arterial embolization Am J Gastroenterol 1991 86 1645 1951244
Pietrabissa A Giulianotti P Campatelli A Di Candio G Farina F Signori S Mosca F Management and follow-up of 78 giant hemangiomas of the liver Br J Surg 1996 83 915 918 8813773
Brouwers MA Peeters PM De Jong KP Haagsma EB Klompmaker IJ Bijleveld CM Zwaveling JH Slooff MJ Surgical treatment of giant hemangioma of the liver Br J Surg 1997 84 314 6 9117293 10.1046/j.1365-2168.1997.02534.x
Gedaly R Pomposelli JJ Pomfret EA Lewis WD Jenkins RL Cavernous hemangioma of the liver. Anatomic resection vs enucleation Arch Surg 1999 134 407 411 10199314 10.1001/archsurg.134.4.407
Gandolfi L Leo P Solmi L Vitelli E Verros G Colecchia A Natural history of hepatic hemangiomas: clinical aand ultrasound study Gut 1991 32 677 680 2060877
De Wilt JH de Man RA Lameris JS Zondervan PE Tilanus HW IJzermans JN Hepatocellular adenoma in 20 patients; recommendations for treatment Ned Tijdschr Geneeskd 1998 142 2459 63 10028326
Croes EA van Gulik TM Bosma A de Wit LT Gouma DJ Treatment of 8 patients with an acute hemorrhage in a hepatocellular adenoma at the Academic Medical Center, Amsterdam Ned Tijdschr Geneeskd 1998 142 2463 2468 10028327
De Wilt JH de Man RA Lameris JS IJzermans JN Focal nodular hyperplasia of the liver: assessment of diagnosis and treatment in 31 patients in 15 years Ned Tijdschr Geneeskd 1996 140 18 22 8569904
Marks WH Thompson N Appleman H Failure of hepatic adenomas (HCA) to regress after discontinuance of oral contraceptives. An association with focal nodular hyperplasia (FNH) and uterine leiomyoma Ann Surg 1988 208 190 5 2840865
Delbeke D Martin WH Sandler MP Chapman WC Wright JK Pinson CW Evaluation of benign vs malignant hepatic lesions with positron emission tomography Arch Surg 1998 133 510 515 9605913 10.1001/archsurg.133.5.510
Ho CL Yu SC Yeung DW 11C-acetate PET imaging in hepatocellular carcinoma and other liver masses J Nucl Med 2003 44 213 221 12571212
| 15804352 | PMC1087495 | CC BY | 2021-01-04 16:28:04 | no | BMC Surg. 2005 Apr 1; 5:7 | utf-8 | BMC Surg | 2,005 | 10.1186/1471-2482-5-7 | oa_comm |
==== Front
Cancer Cell IntCancer Cell International1475-2867BioMed Central London 1475-2867-5-81581396710.1186/1475-2867-5-8Primary ResearchAndrogen receptor signaling is required for androgen-sensitive human prostate cancer cell proliferation and survival Yang Qing [email protected] Kar-Ming [email protected] Wanda V [email protected] Bradley P [email protected] Hsueh-Kung [email protected] Department of Urology, University of Oklahoma Health Sciences Center, Oklahoma City, OK, USA2 Department of Pathology, University of Oklahoma Health Sciences Center, Oklahoma City, OK, USA3 Department of Veterans Affairs Medical Center, Oklahoma City, OK, USA2005 6 4 2005 5 8 8 15 12 2004 6 4 2005 Copyright © 2005 Yang et al; licensee BioMed Central Ltd.2005Yang 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
Androgens and androgen receptors (AR) regulate normal prostate development and growth. They also are involved in pathological development of prostatic diseases, including benign prostatic hyperplasia (BPH) and prostate cancer (PCa). Antiandrogen therapy for PCa, in conjunction with chemical or surgical castration, offers initial positive responses and leads to massive prostate cell death. However, cancer cells later appear as androgen-independent PCa. To investigate the role of AR in prostate cell proliferation and survival, we introduced a vector-based small interfering RNA (siRNA). This siRNA targeted 5'-untranslated region of AR mRNA for extended suppression of AR expression in androgen-sensitive human prostate LNCaP cells.
Results
The siRNA design successfully suppressed endogenous AR expression, as revealed by western blotting and immunofluorescence staining in LNCaP cells. LNCaP cells did not proliferate in the absence of AR and underwent apoptosis, based on elevated phospho-Histone H2B expression and higher number of apoptotic body as compared to control cells.
Conclusion
We demonstrated that AR is vital for prostate cell proliferation and survival in this androgen-sensitive prostate cell line. These results further strengthen the hypothesis that AR can be a therapeutic target for treating androgen-sensitive stages of PCa. Unlike antiandorgens, however, siRNA targeting AR provides a direct inactivation of AR function through the suppression of AR protein expression.
androgen receptorRNA interferenceprostate cancer
==== Body
Background
Androgens are critical for the development and growth of normal prostate. They also are responsible for the development of prostate diseases, including benign prostatic hyperplasia (BPH) and prostate cancer (PCa). Androgen receptors (AR) transduce androgen signals in prostate cells to regulate the physiological and pathological development of the gland [1]. It is classically characterized that after ligand binding [mainly 5α-dihydrotestosterone (5α-DHT)] the ligand-AR receptor complex with associated proteins translocates into the nucleus, binds to the consensus sequence of androgen response elements (AREs) [2], and regulates androgen responsive genes (ARGs) expressions [3]. Conditions that activate abnormal AR trans-activation through AR mutations [4,5], amplification of AR [6], or androgen-independent signaling pathways [7,8] can lead to or be a result of the development of prostatic diseases or androgen refractory PCa.
Clinically, androgen ablation therapy is used to reduce AR ligand production or to block AR-mediated signaling. Finasteride, a 5α-reductase type 2 inhibitor, has been used for treating patients with BPH. Finasteride slows the progression of BPH by suppressing 5α-DHT synthesis [9]. Androgen ablation therapy for PCa has been achieved using compounds called antiandrogens. Antiandrogens compete with 5α-DHT for AR binding and intend to block AR-mediated signaling. Antiandrogens are classified, according to their chemical structure, as either steroidal antiandrogens (ie, cyproterone acetate or medroxyprogesterone acetate) or non-steroidal anti-androgens [ie, nilutamide (Anandron™), flutamide (Eulexin™), and bicalutamide (Casodex™)] [10]. Unfortunately, essentially all of the patients who show initially favorable responses to the androgen blockade eventually become refractory to this treatment [11].
Abruption of AR signaling has been examined in experimental models to investigate AR-mediated prostate cell physiology and pathophysiology. An antisense oligonucleotide [12], a hammerhead robozyme [13], and microinjection of an AR neutralizing antibody [14] have been designed to suppress AR expression or block AR-mediated signaling in in vitro models. The advance of RNA interference (RNAi) technology provides a tool to suppress gene expression through post-translational gene silencing (PTGS) [15]. This technology utilizes duplex small interfering RNAs (siRNAs) of about 19–23 nucleotides to suppress the expression of the homologous gene [16,17]. Wright et al. reported a successful use of a synthetic double-stranded (ds) siRNA to suppress endogenous AR expression in LNCaP cells [18]. In this study, we established a vector-based siRNA plasmid construct targeting the 5'-untranslated region (5'-UTR) of the AR in the same prostate cell line. We successfully achieved AR silencing in LNCaP cells, and further demonstrated that these cells could neither proliferate nor survive in the absence of the AR. We concluded that AR signaling is required for both prostate cell proliferation and survival, at least in androgen-sensitive states. Specific blocking of AR expression and AR-mediated signaling is required for a successful androgen blockade in patients with prostate diseases. In addition, the vector-based AR silencing design provides a tool to study AR-associated genomic and non-genomic effects and to identify potential pathways that cross-talk with AR.
Results and discussion
To identify the role of AR in AR-positive, androgen-sensitive LNCaP cells, we designed a vector-based siRNA plasmid construct to achieve "long-term" AR silencing. This construct contained the U6 promoter-driven sense and antisense strands of AR siRNA that have a termination signal consisting of 6 thymidines. This construct also included a constitutively expressed CMV promoter-driven green fluorescence protein (GFP), designated pSiAR-EGFP (Fig. 1). Because there is an adverse effect on androgen-sensitive prostate cell proliferation in the absence of the AR [12,14,18], the inclusion of the CMV promoter-driven GFP expression could be used to select and enrich AR-negative LNCaP cells following transfection. A construct, named pSiCon-EGFP, was established with a scrambled AR target sequence and served as a control. At 16 hours following transfection, GFP was expressed and detectible under a fluorescence microscope (data not shown).
Figure 1 A vector-based plasmid construct for suppression of AR expression in LNCaP cells. (A) Two single-stranded oligonucleotides were synthesized consisting of a 19 nucleotide of -25 to -7 bp 5'-UTR of the AR mRNA separated by a short loop sequence from the reverse complement of the same sequence. The sense strand of synthesized oligonucleatides was ended with five thymidines as termination signal. The annealed ds DNA contained Bgl II and Sfi I restriction at its 5' and 3' end of the oligonucleotides, respectively. (B) Schematic drawing of the pSiAR-EGFP vector. The expression construct was designed to expression AR siRNA driven by U6 promoter for suppression of endogenous AR expression in mammalian cells. The expression construct also contained EGFP driven by the CMV promoter constitutive expression of GFP in target cells.
Plasmid siRNA construct targeting 5'-UTR of AR mRNA suppressed AR expression in LNCaP cells
To determine whether the pSiAR-EGFP construct was capable of expressing siRNA, thereby suppressing AR expression in LNCaP cells, cells were transfected with pSiAR-EGFP or with pSiCon-EGFP construct. AR expression was monitored by staining these cells with mouse-anti-human AR monoclonal antibody. Immunocytochemical staining of LNCaP cells showed that portions of cells transfected with pSiAR-EGFP plasmid had no detectible AR expression, whereas cells transfected with pSiCon-EGFP continued to show AR expression in all cells (Figure 2A). However, at 2 weeks following transfection, there was no detectible GFP-positive or AR-negative LNCaP cells in the pSiAR-EGFP transfected group (data not shown). This suggested that either the cells lost their transfected pSilencer plasmids, or that the cells receiving the pSiAR-EGFP could, in fact, not survive in the absence of the AR.
Figure 2 Suppression of endogenous AR expression in LNCaP cells using the pSiAR-EGFP expression construct. LNCaP cells were transfected with either the pSiAR-EGFP or the pSiAR-EGFP construct using the Lipofectemine 2000 protocol. (A) Immunocytochemical detection of AR expression in transfected cells. To determine AR expression in LNCaP cells, following transfection, cells were stained with mouse anti-human AR monoclonal antibody (1:100 dilution) followed by AlexaR FluorR 594 conjugated goat anti-mouse IgG secondary antibody (2 mg/ml) incubation. The AR staining was detected by fluorescent microscopy (BX51, Olympus). The images were composite between the red with the positive in AR staining with the phase-contract from the same field. Suppression of endogenous AR expression was demonstrated by the absence of red fluorescence staining as indicated by arrows. DAPI staining showed the number of cells in the same field. (B) Western blot analysis of AR expression in pSiAR-EGFP and control transfected cells. At 7 days following transfection, both GFP-positive and GFP-negative cells were collected through cell sorter; and cells were lysed to prepare cellular protein extracts. Aliquots of 20 μg total cellular protein were loaded into Tris-HCl gels and transferred to PDVF membranes. The AR expression was determined by incubating with mouse anti-human AR monoclonal antibody (1:500) followed by the HRP-conjugated anti-mouse IgG (1:125,000) secondary antibody incubation. Immunoreactive signals were detected using ECL. Levels of β-actin expression was also determined in each sample and used as protein loading control.
To demonstrate that GFP can be used as a marker for selecting AR-negative LNCaP cells, pSiAR-EGFP or pSiCon-EGFP transfected cells were trypsinized 24 hours following transfection. GFP-positive cells were collected through a cell sorter to determine AR protein expression. Western blot analysis showed no detectible AR protein expression in GFP-positive cells transfected with pSiAR-EGFP. However, GFP-positive cells receiving the control pSiCon-EGFP construct continued to express AR protein (Figure 2B). This result demonstrated that GFP positive cells are AR negative cells.
The use of 5'-UTR as a target for designing siRNA has been discouraged, because UTR binding proteins and/or translation initiation complexes may interfere with siRNA endonuclease complex binding [19]. There are multiple mutations in cases of PCa being identified throughout the AR coding regions [20-22]. This makes the selection of the AR coding region for siRNA design possibly inappropriate and insufficient to block the expression of mutated AR in PCa patients. We selected three 5'-UTR regions of the AR for our siRNA design; only the region between -25 to -7 bp could successfully suppress AR expression in LNCaP cells (data not shown). Our results further demonstrated that siRNA's efficacy of suppressing gene expression varies from one region to another, due to target sequence structures [23].
LNCaP cells required AR to support cell proliferation
To determine LNCaP cell proliferative capability in the absence of AR, cell proliferation with and without cell sorting was performed following the silencer construct transfection. When cell proliferation was performed without cell sorting, during a 9 day period the pSiAR-EGFP transfected LNCaP cells consistently proliferated slower than the parental cells and the pSiCon-EGFP transfected cells (Figure 3A). After 9 days in culture, both parental and control cells reached confluence. The reduced cell proliferation in pSiAR-EGFP transfected cells might reflect that cells receiving the AR siRNA construct failed to proliferate in the absence of AR.
Figure 3 Suppression of LNCaP cell proliferation in the absence of endogenous AR. LNCaP cells were seeded in tissue culture plates and transfected with a mixture of either pSiAR-EGFP or pSiCon-EGFP plasmid construct with Lipofecamine 2000 in OPTI-MEM. (A) Cell proliferation without separation of GFP-positive and negative cells. At 24 hours following transfection, cells were trypsinized and distributed into each well (1,000 cells/well) of 96-well tissue culture plates in the presence the complete medium. Cell proliferation was determined using the XTT assay kit for a period of 9 days; data from days 11 and 14 were not included since parental and pSiCon-EGFP transfected LNCaP cells reached confluence after day9. (B) LNCaP cell proliferation following enrichment of GFP-positive cells. At 24 hours after transfection, cells were trypsinized and EGFP-positive cells were collected through the MoStar cell sorting system. GFP-positive cells were seeded into each well (1,000 cells/well) of 96-well plates for cell viability assay. Cell proliferation was determined for a period of 14 days. Data were calculated as absorbance at days following transfection normalized to the absorbance at the day of cell sorting, and presented as fold induction in absorbance. LNCaP cells transfected with pSiCon-EGFP plasmid construct were used as the AR-positive control. * represents significant statistical difference between LNCaP cells with and without AR (P < 0.001). Each time point represents the mean ± SEM from 3 independent experiments.
To confirm that AR signaling is required for LNCaP cell proliferation, GFP enriched LNCaP cells were collected through cell sorting at 24 hours following transfection. LNCaP cells transfected with the pSiCon-EGFP construct continued to grow over the entire experimental period (Figure 3B). This indicates that the silencer control construct did not affect cell proliferation. However, pSiAR-EGFP transfected LNCaP cells (GFP-positive, AR-negative) failed to proliferate (Figure 3B), indicating that the androgen-sensitive prostate cells require AR for proliferation.
Suppression of AR expression and inactivation of AR function in prostate cells has been achieved through the use of an antisense oligonucleotide [12,24], a hammerhead robozyme [13], and a synthesized ds siRNA duplex [18] to target AR mRNA. Microinjection of an AR neutralizing antibody has also been reported to block AR-mediated signaling in LNCaP cells [14]. Recently, the use of vector-based siRNA targeting AR has been reported to study the involvement of AR signaling in vitamin D-suppressed prostate cell growth [25]. These published approaches were intended for transient gene silencing in target cells, and provided short term elimination of AR expression. All of the currently published results are consistent with our observations that disruption of AR signaling adversely affects androgen-sensitive LNCaP cell proliferation.
LNCaP cells required AR to maintain survival
To determine whether LNCaP cells could survive without AR, we determined the levels of phospho-histone H2B expression and the number of apoptotic body following AR silencing. Histon H2B phosphorylation has been shown to be uniquely associated with apoptosis, specifically in apoptosis-induced chromatin condensation in mammalian cells [26]. Its expression significantly increases in cells undergoing apoptosis [27]. Using Western blot analysis, phospho-histone H2B expression was quantitated in LNCaP cells transfected with either pSiAR-EGFP or pSiCon-EGFP. There was no detectible phospho-histone H2B signal in pSiAR-EGFP or control transfected cells at 4 days after transfection (data not shown). Six days following pSiAR-EGFP transfection, however, elevated levels of phospho-histone H2B expression were detected in AR negative LNCaP cells. At that time, there was no detectible phospho-histone H2B in the pSiCon-EGFP transfected LNCaP cells (Figure 4).
Figure 4 Elevated expression of phospho-histone H2B S(14) in AR-knockdown LNCaP cells. LNCaP cells were either transfected with pSiAR-EGFP or pSiCon-EGFP plasmid construct. On day 6 after transfection, cells were harvested. Cellular proteins were extracted using acid extraction method, electrophoresized through gradient Tris-HCl gels, and electroblotted onto PVDF membranes. Levels of phospho-histone H2B expression was detected by an immunoassay procedure.
To prove that AR signaling is required for LNCaP cell survival, the number of cells undergoing apoptosis (demonstrated by the presence of apoptotic bodies) was determined in cells transfected with pSiAR-EGFP and control constructs (Figure 5). The average number of apoptotic bodies in pSiAR-EGFP transfected cells was 52.89 per 1,000 cells, versus 6.53 in 1,000 pSiCon-EGFP transfected cells (Table 1). The transfection efficiencies for each vector were not statistically significant – pSiAR-EGFP was 17.0% and pSiCon-EGFP was 22.9% (Table 1). When the number of apoptotic bodies was corrected for transfection efficiency, ABI was 296.2 for pSiAR-EGFP and 29.2 for pSiCon-EGFP transfected cells (Table 1). The ABI was statistically higher in pSiAR-EGFP transfected cells versus pSiCon-EGFP transfected cells. These results demonstrated that LNCaP cells lacking AR would go through the process of apoptosis.
Figure 5 Elevated number of apoptotic LNCaP cells transfected pSiAR-EGFP construct. Apoptotic bodies in LNCaP cells transfected with control vector pSiCon-EGFP (A) and pSiAR-EGFP (B). Following transfection with these plasmid constructs, LNCaP cells were collected through centrifugation, fixed in neutralized formalin, and encased in agrose blocks. Cell blocks were paraffin-embedded, sectioned, deparaffinized, rehydrated, and stained with hematoxylin. Apoptotic bodies which showed condensed and/or cleaved nucleus were counted from random fields and numerical data are shown in Table 1.
Table 1 Apoptotic body indices in pSiAR-EGFP and pSiCon-EGFP transfected LNCaP cells
# of apoptotic bodies (per 1,000)a Transfection efficiency (%) Apoptotic body indexc
Mean + S.D. Mean + S.D. Mean + S.D.
pSiAR-EGFP 52.89 ± 22.12b 17.0 ± 3.45 296.2 ± 78.2b
pSiCon-EGFP 6.53 ± 2.85 22.9 ± 2.84 29.2 ± 15.4
a The number of apoptotic bodies were was determined by counting 1,000 cells from each of 3 independent hematoxylin stained slides.
b Significant difference between pSiAR-EGFP and pSiCon-EGFP transfected cells (p < 0.05).
c Apoptotic body indices were calculated after normalizing to transfection efficiency.
To demonstrate that LNCaP cells could survive with transient suppression of AR, we also performed AR protein expression knockdown using a synthesized, pooled ds siRNA, siRNA SMARTpool AR® (Upstate). This siRNA pool suppressed a majority (greater than 95%) of AR protein expression in LNCaP cells at 16 hours after transfection, as compared to cells transfected with a non-specific control pool. The cells maintained suppressed levels of AR protein expression for at least 72 hours following transfection. They resumed AR expression, to 20–30% of normal levels, after 72–96 hours post-transfection (data not shown). LNCaP cells could survive in the absence of AR for at least 3–4 days, as indicated by the lack of detectable phospho-histone H2B signals at 96 hours following the pSiAR-EGFP transfection (data not shown). Due to the transient nature of the use of antisense oligonucleotides, hammerhead robozymes, ds siRNA duplexes, and microinjected neutralizing antibodies against AR, cell survival in the extended absence of AR could not be assessed. Using the combination of the vector-based siRNA delivery and enrichment of AR-negative cells through the selection of GFP-positive cells enabled us to study cell behavior in the absence of AR over an extended time-period.
Antiandrogens are used to prevent the acquisition of a transcriptionally active conformation of the AR. The induction of apoptosis in prostate cells treated with bicalutamide has been observed [28]. However, several reports also indicate that in the presence of bicalutamide, prostate cells survive in culture [29], regain cell growth after an extended period of exposure to the antiandrogen [30,31], and become a more invasive phenotype [32]. Bicalutamide can also act as an AR agonist, resulting in the stimulation of AR trans-activation [33]. Furthermore, it has been indicated that bicalutamide may support prostate cell survival and progression through selection of cells with AR mutations [34] or cells with elevated expression of growth factors [35]. These mutations generate receptors that respond to other steroids and antiandrogens by increased trans-activation [4]. These results suggest that antiandrogens may be insufficient to block AR signaling through multiple candidate pathways.
The potential use of RNAi technology has been investigated in the field of gene therapy [36,37]. Targeting AR suppression using RNAi might be more efficient and specific than using antiandrogens to inactivate AR trans-activation and turn off ARGs expression. LNCaP cells potentially consist of multiple cell lineages, due to the development of multiple sublines from the original LNCaP cells [38-40]. Although all LNCaP cells transfected by the AR siRNA construct could neither proliferate nor survive in the absence of the AR in this study, we do not have information regarding specific sub-populations that may be more susceptible to transfection and sensitive to the absence of the AR. The development of a more efficient delivery system, such as the lentiviral-based gene delivery system [41], may allow for further study of AR signaling in multiple prostate cell lines.
Conclusion
We concluded that all PCa cells, at least in the androgen-sensitive and AR-positive stages, require AR for continued proliferation and survival. The identification of AR pathway and ARG(s) that transduce androgen signaling might provide a specific target to block androgen-activated, AR-mediated prostate cell growth. Specifically targeting AR or its downstream signaling molecules potentially will be effective in achieving total androgen blockade. With the development of "long-term" siRNA and efficient delivery systems in vivo, we may achieve a total AR blockade in the prostate, thereby improving treatment for patients with prostate diseases.
Methods
Establishment of AR silencer plasmid constructs
To establish a siRNA plasmid construct with fluoresce selection marker, a multiple cloning site (MCS) was excised from the pSE380 (Invitrogen) using Bcl I and Hind III. The MCS containing fragment was subcloned into Bam HI and Hind III linearized pSilencer 2.1-U6 hygro vector (Ambion) to obtain pSilencer 2.1-U6 MCS hygro. The IRES2-EGFP fragment coding for GFP was excised from pIRES2-EGFP (BD Biosciences Clontech) using Bgl II and Afl II. This excised fragment was then subcloned into the pSilencer 2.1U6 MCS hygro linearized with Bcl I and Sal I in the presence of a short adaptor containing Afl I site on its 5'ends and Sal I site on its 3'ends. This established plasmid was named pSilencer 2.1-U6-IRES2-EGFP. To create the silencer construct targeting AR, two complementary strands of oligonucleotides targeting -25 to -7 bp upstream from the ATG transcription start codon of the AR (GenBank accession number M20132; Figure 1A) with Bgl II and Sfi I sites on its 5' and 3' ends, respectively, was synthesized and annealed. The ds oligonucleotide was cloned into Bam HI and Hind III linearized pSilencer 2.1-U6-IRES2-EGFP. The 19-nucleotide AR target sequence is indicated in the sense strand of the synthesized ds DNA. The U6 promoter-driven siRNA expresses the sense and antisense strands of AR siRNA [42] that have a termination signal consisting of 6 thymidines [43].
Since the SV40 promoter could not provide sufficient levels of GFP expression in LNCaP cells for monitoring successful transfection (data not shown), the SV40 promoter was replaced by cymagalovirus (CMV) promoter. The CMV promoter was excised from pIRES2-EGFP with Nsi I and Xho I, and ligated into Aat II and Xho I linearized pSilencer 2.1-U6-IRES2-EGFP. The final construct was designated pSiAR-EGFP (Figure 1B). Control construct also was established by cloning an annealed ds DNA fragment with a scrambled 19-nucleotide AR target sequence, named pSiCon-EGFP. The scrambled sequence was subjected to Blast search to ensure no match to any known transcript.
LNCaP cell culture, transfection, and selection of cells transfected with the AR silencer expression construct
LNCaP cells were purchased from ATCC (CRL-1740). They were cultured and maintained in complete growth medium consisting of RPMI1640 (Invitrogen) supplemented with 10% fetal bovine serum (FBS), 100 unit/ml penicillin, and 100 μg/ml streptomycin (Invitrogen) in a humidified atmosphere containing 5% CO2 at 37°C. Transfection of cells with the plasmid silencer constructs was performed in either 60 or 100 mm tissue culture plates using Lipofecamine 2000 (Invitrogen) when cells reached 80–90% confluence. Briefly, for transfection in 100 mm plates, 24 μg of plasmid DNA and 60 μl of Lipofecamine 2000 transfection reagent were diluted separately in 1.5 ml of OPTI-MEM (Invitrogen). They were then combined, and incubated at room temperature for 20 minutes. The mixture was added to LNCaP cells in the presence of 10 ml OPTI-MEM (Invitrogen). The serum free medium was replaced by the complete growth medium 4–6 hours following transfection. The day of transfection was defined as day 0. To select cells that received the silencer constructs, LNCaP cells were trypsinized at 24 hours after transfection (day 1). GFP-positive LNCaP cells were collected using MoStar cell sorting system. For determining the apoptotic index, transfection efficiency was calculated at 24 hours after transfection. A total of 1,000 cells were randomly selected under a microscope equipped with fluorescence filters from 2–3 fields. The percentages of cells expressing GFP were termed "transfection efficiency". The transfection efficiency average was calculated from three independent experiments for each plasmid transfection.
Immunocytochemical staining of AR in LNCaP cells
To demonstrate that the AR silencer plasmid construct could suppress AR expression in LNCaP cells, at 24 hours post-transfection the cells were re-seeded onto cover slips in 24-well tissue culture plates at the density of 2,000 cells/well. Cells were incubated overnight for adherence. They were sequentially fixed in methanol for 10 minutes and acetone for 1 minute. Cells were then permeabilized with 0.2% triton-X-100 (Simga) in 1x phosphate buffered saline (PBS) at room temperature for 20 minutes. Following incubation with 10% goat serum in 1x PBS to block non-specific binding, cells were incubated for another 1 hour with mouse anti-human AR monoclonal antibody (1:100; NOVOCastra) in 1x PBS containing 10% goat serum. Following three washes with PBS supplemented with 0.5 mM CaCl2 and MgCl2, cells were incubated with Alexa Fluor® 594-conjugated goat anti-mouse IgG (1:200; Molecular Probes) secondary antibody. For nuclei staining, cells were incubated with 0.1 μg/ml 4',6'-diamidino-2-phenylindole hydrochloride (DAPI) at room temperature for 20 minutes. Images were captured by fluorescent microscopy (BX51, Olympus) equipped with the SPOT software (Diagnostic Instrument).
Protein extraction and Western blot analysis
Western blot analysis was performed to determine AR levels and phospho-histone H2B expression in LNCaP cells transfected with the silencer constructs. To determine levels of AR expression, pSiAR-EGFP and control transfected GFP-positive LNCaP cells were collected through cell sorting. They were lysed with cell lysis buffer consisting of 5 mM EDTA, 0.5% Triton-X-100, and 0.1 mM phenylmethylsulphonylfluoride (PMSF) in 1x PBS at 1 μl/104 cells. Total cellular extracts were collected following centrifugation. Protein concentrations were determined by bicinchoninic acid (BCA) protein assay kit (Pierce). Aliquots of 20 μg of the protein extracts were separated on 10% Tris-HCl gel (Bio-Rad). Proteins were then transferred to PVDF membranes (Bio-Rad). AR protein was detected by incubating the membranes with the mouse anti-human AR monoclonal antibody (1: 500) at room temperature for 2 hours. This was followed by the horseradish peroxidase (HRP)-conjugated anti-mouse IgG (1:125,000; KPL) secondary antibody incubation at room temperature for another hour. Immunoreactive protein was detected using the enhanced chemiluminescent (ECL) reagent (Pierce) according to the manufacturer's protocol.
To determine histone H2B phosphorylation levels in LNCaP cells in the absence of AR protein, cells were transfected with either pSiAR-EGFP or control construct as described. Cellular proteins were extracted with acid extraction buffer consisting of 10 mM Hepes (pH 7.9), 1.5 mM MgCl2, 10 mM KCl, 0.5 mM DTT, and 1.5 mM PMSF in the presence of 0.2 M hydrochloric acid. This was completed according to procedures provided by the antibody supplier (Upstate). Soluble proteins were collected from supernatants after centrifugation at 11,000 × g for 10 minutes. Proteins were then dialyzed overnight against distilled water. Aliquots of 30 μg proteins were prepared from each sample, separated on 4–20% gradient Tris-HCl gels (Bio-Rad), and transferred to PVDF membranes. After blocking non-specific binding, the membranes were incubated overnight with rabbit anti-phospho-histone H2B S(14) antibody (1:500; Upstate) at 4 °C. This was followed by a HRP-conjugated anti-rabbit IgG (1:10,000; Cell Signaling) secondary antibody incubation followed by ECL detection.
Cell proliferation assay
Cell proliferation assay was determined in pSiAR-EGFP and pSiCon-EGFP transfected LNCaP cells. At 24 hours following transfection, they were either directly seeded into tissue culture plates or subjected to cell sorting to collect GFP-positive cells prior to cell seeding. Cells were seeded into 96-well plates at 1,000 cells/well. Cell proliferation was determined using XTT cell proliferation assay kit (Roche) following the manufacturer's protocol. Proliferation was evaluated for a period of 14 days. Data analysis of three independent transfections of cells sorted for GFP-positive cells was performed as previously reported [44]. Results were presented as mean ± standard error of means (SEM).
Determination of apoptotic index in cell blocks
To characterize whether LNCaP cells undergo apoptosis in the absence of AR, numbers of apoptotic bodies were quantitated in LNCaP cells transfected with pSiAR-EGFP and control silencer constructs. Transfected cells were tripsinized and centrifuged at 400 × g for 5 minutes to collect the cells. The cell pellets were fixed in 10% formalin for at least 24 hours. They were then mixed with drops of warm 2% agarose in order to form cell-agarose blocks. The cell blocks were paraffin embedded, cut at 4–6 μm, mounted on microscope slides, and baked at 55 °C overnight. The slides were deparaffinized, rehydrated to water, and stained by hematoxylin. Apoptotic bodies were quantitated under a bright field microscope at 400× magnification. Apoptotic bodies were defined as small, roughly spherical or ovoid cytoplasmic fragments, some of which contained nuclear fragments [45]. The density of apoptotic cells was determined by counting 1,000 cells from each transfection. Quantitative analyses of apoptotic changes were recorded as apoptotic body indices (ABI, the number of apoptotic cells/transfection efficiencies).
Statistical Analysis
For statistical analysis, data were presented as mean and SEM from at least 3 independent experiments. Statistical significance was determined when P < 0.05. Student t-test was used for comparing two treatment groups.
Abbreviations
5α-DHT: 5α-dihydrotestosterone; AR: androgen receptor; ARE: androgen response element; ARG: androgen response gene; RNAi: RNA interference; siRNA: small interfering RNA.
Acknowledgements
This work was supported by the NIH grant DK54925 to HKL.
==== Refs
Heinlein CA Chang C Androgen receptor in prostate cancer Endocr Rev 2004 25 276 308 15082523 10.1210/er.2002-0032
Roche PJ Hoare SA Parker MG A consensus DNA-binding site for the androgen receptor Mol Endocrinol 1992 6 2229 2235 1491700 10.1210/me.6.12.2229
Nelson PS Clegg N Arnold H Ferguson C Bonham M White J Hood L Lin B The program of androgen-responsive genes in neoplastic prostate epithelium Proc Natl Acad Sci USA 2002 99 11890 11895 12185249 10.1073/pnas.182376299
Culig Z Hobisch A Cronauer MV Cato ACB Hittmair A Radmayr C Eberle J Bartsch G Klocker H Mutant androgen receptor detected in an advanced-stage prostatic carcinoma is activated by adrenal androgens and progesterone Mol Endocrinol 1993 7 1541 1550 8145761 10.1210/me.7.12.1541
Buchanan G Yang M Harris JM Nahm HS Han G Moore N Bentel JM Matusik RJ Horsfall DJ Marshall VR Greenberg NM Tilley WD Mutations at the boundary of the hinge and ligand binding domain of the androgen receptor confer increased transactivation function Mol Endocrinol 2001 15 46 56 11145738 10.1210/me.15.1.46
Visakorpi T Hyytinen E Koivisto P Tanner M Keinänen R Palmberg C Palotie A Tammela T Isola J Kallioniemi O-P In vivo amplification of the androgen receptor gene and progression of human prostate cancer Nat Genet 1995 9 401 406 7795646 10.1038/ng0495-401
Nazareth LV Weigel NL Activation of the human androgen receptor through a protein kinase A signaling pathway J Biol Chem 1996 271 19900 19907 8702703 10.1074/jbc.271.33.19900
Ueda T Mawji NR Bruchovsky N Sadar MD Ligand-independent activation of the androgen receptor by interleukin-6 and the role of steroid receptor coactivator-1 in prostate cancer cells J Biol Chem 2002 277 38087 38094 12163482 10.1074/jbc.M203313200
Stoner E The clinical development of a 5α-reductase inhibitor, finasteride J Steroid Biochem Mol Biol 1990 37 375 378 1701660 10.1016/0960-0760(90)90487-6
Soloway MS Matzkin H Antiandrogenic agents as monotherapy in advanced prostatic carcinoma Cancer 1993 71 1083 1088 8428332
Scher HI Steineck G Kelly WK Hormone-refractory (D3) prostate cancer: refining the concept Urol 1995 46 142 148 7624983 10.1016/S0090-4295(99)80182-4
Eder IE Culig Z Ramoner R Thurnher M Putz T Nessler-Menardi C Tiefenthaler M Bartsch G Klocker H Inhibition of LncaP prostate cancer cells by means of androgen receptor antisense oligonucleotides Cancer Gene Ther 2000 7 997 1007 10917202 10.1038/sj.cgt.7700202
Chen S Song CS Lavrovsky Y Bi B Vellanoweth R Chatterjee B Roy AK Catalytic cleavage of the androgen receptor messenger RNA and functional inhibition of androgen receptor activity by a hammerhead ribozyme Mol Endocrinol 1998 12 1558 1566 9773979 10.1210/me.12.10.1558
Zegarra-Moro OL Schmidt LJ Huang H Tindall DJ Disruption of androgen receptor function inhibits proliferation of androgen-refractory prostate cancer cells Cancer Res 2002 62 1008 1013 11861374
Cogoni C Macino G Post-transcriptional gene silencing across kingdoms Curr Opin Genet Dev 2000 10 638 643 11088014 10.1016/S0959-437X(00)00134-9
Elbashir SM Harborth J Lendeckel W Yalcin A Weber K Tuschl T Duplexes of 21-nucleotide RNAs mediate RNA interference in cultured mammalian cells Nature 2001 411 494 498 11373684 10.1038/35078107
Zamore PD Tuschl T Sharp PA Bartel DP RNAi: double-stranded RNA directs the ATP-dependent cleavage of mRNA at 21 to 23 nucleotide intervals Cell 2000 101 25 33 10778853 10.1016/S0092-8674(00)80620-0
Wright ME Tsai MJ Aebersold R Androgen receptor represses the neuroendocrine transdifferentiation process in prostate cancer cells Mol Endocrinol 2003 17 1726 1737 12775765 10.1210/me.2003-0031
Elbashir SM Harborth J Weber K Tuschl T Analysis of gene function in somatic mammalian cells using small interfering RNAs Methods 2002 26 199 213 12054897 10.1016/S1046-2023(02)00023-3
Newmark JR Hardy DO Tonb DC Carter BS Epstein JI Isaacs WB Brown TR Barrack ER Androgen receptor gene mutations in human prostate cancer Proc Natl Acad Sci, USA 1992 89 6319 6323 1631125
Brinkmann AO Jenster G Ris-Stalpers C van der Korput JAGM Brüggenwirth HT Boehmer ALM Trapman J Androgen receptor mutations J Steroid Biochem Mol Biol 1995 53 443 448 7626493 10.1016/0960-0760(95)00090-M
Barrack ER Androgen receptor mutations in prostate cancer The Mount Sinai J of Med 1996 63 403 412
Gou D Jin N Liu L Gene silencing in mammalian cells by PCR-based short hairpin RNA FEBS Letters 2003 548 113 118 12885417 10.1016/S0014-5793(03)00630-6
Ko YJ Devi GR London CA Kayas A Reddy MT Iversen PL Bubley GJ Balk SP Androgen receptor down-regulation in prostate cancer with phosphorodiamidate morpholino antisense oligomers J Urol 2004 172 1140 1144 15311058 10.1097/01.ju.0000134698.87862.e6
Bao BY Hu YC Ting HJ Lee YF Androgen signaling is required for the vitamin D-mediated growth inhibition in human prostate cancer cells Oncogene 2004 23 3350 3360 15048085 10.1038/sj.onc.1207461
Ajiro K Histone H2B phosphorylation in mammalian apoptotic cells. An association with DNA fragmentation J Biol Chem 2000 275 439 443 10617636 10.1074/jbc.275.1.439
Cheung WL Ajiro K Samejima K Kloc M Cheung P Mizzen CA Beeser A Etkin LD Chernoff J Earnshaw WC Allis CD Apoptotic phosphorylation of histone H2B is mediated by mammalian sterile twenty kinase Cell 2003 113 507 517 12757711 10.1016/S0092-8674(03)00355-6
Lee EC Zhan P Schallhom R Packman K Tenniswood M Antiandrogen-induced cell death in LNCaP human prostate cancer cells Cell Death Differ 2003 10 761 771 12815459 10.1038/sj.cdd.4401228
Lin J Adam RM Santiestevan E Freeman MR The phosphatidylinositol 3'-kinase pathway is a dominant growth factor-activated cell survival pathway in LNCaP human prostate carcinoma cells Cancer Res 1999 59 2891 2897 10383151
Hara T Miyazaki J Araki H Yamaoka M Kanzaki N Kusaka M Miyamoto M Novel mutations of androgen receptor: a possible mechanism of bicalutamide withdrawal syndrome Cancer Res 2003 63 149 153 12517791
Godoy-Tundidor S Hobisch A Pfeil K Bartsch G Culig Z Acquisition of agonistic properties of nonsteroidal antiandrogens after treatment with oncostatin M in prostate cancer cells Clin Cancer Res 2002 8 2356 2361 12114440
Zhan P Lee EC Packman K Tenniswood M Induction of invasive phenotype by Casodex in hormone-sensitive prostate cancer cells J Steroid Biochem Mol Biol 2002 83 101 111 12650706 10.1016/S0960-0760(02)00275-3
Brinkmann AO Blok LJ de Ruiter PE Doesburg P Steketee K Berrevoets CA Trapman J Mechanisms of androgen receptor activation and function J Steroid Biochem Mol Biol 1999 69 307 313 10419007 10.1016/S0960-0760(99)00049-7
Taplin ME Bubley GJ Ko YJ Small EJ Upton M Rajeshkumar B Balk SP Selection for androgen receptor mutations in prostate cancers treated with androgen antagonist Cancer Res 1999 59 2511 2515 10363963
Nickerson T Pollak M Bicalutamide (Casodex)-induced prostate regression involves increased expression of genes encoding insulin-like growth factor binding proteins Urology 1999 54 1120 1125 10604720 10.1016/S0090-4295(99)00421-5
Li S Rosenberg JE Donjacour AA Botchkina IL Hom YK Cunha GR Blackburn EH Rapid inhibition of cancer cell growth induced by lentiviral delivery and expression of mutant-template telomerase RNA and anti-telomerase short-interfering RNA Cancer Res 2004 64 4833 4840 15256453
Park WS Hayafune M Miyano-Kurosaki N Takaku H Specific HIV-1 env gene silencing by small interfering RNAs in human peripheral blood mononuclear cells Gene Ther 2003 10 2046 2050 14566364 10.1038/sj.gt.3302099
Iguchi K Otsuka T Usui S Ishii K Onishi T Sugimura Y Hirano K Zinc and metallothionein levels and expression of zinc transporters in androgen-independent subline of LNCaP cells J Androl 2004 25 154 161 14662799
Singh S Sadacharan S Su S Belldegrun A Persad S Singh G Overexpression of vimentin: role in the invasive phenotype in an androgen-independent model of prostate cancer Cancer Res 2003 63 2306 2311 12727854
Hobisch A Hoffmann J Lambrinidis L Eder IE Bartsch G Klocker H Culig Z Antagonist/agonist balance of the nonsteroidal antiandrogen bicalutamide (Casodex) in a new prostate cancer model Urol Int 2000 65 73 79 11025427 10.1159/000064843
Chen CD Welsbie DS Tran C Baek SH Chen R Vessella R Rosenfeld MG Sawyers CL Molecular determinants of resistance to antiandrogen therapy Nat Med 2004 10 33 39 14702632 10.1038/nm972
Miyagishi M Taira K U6 promoter-driven siRNAs with four uridine 3' overhangs efficiently suppress targeted gene expression in mammalian cells Nat Biotechnol 2002 20 497 500 11981564 10.1038/nbt0502-497
Lee NS Dohjima T Bauer G Li H Li MJ Ehsani A Salvaterra P Rossi J Expression of small interfering RNAs targeted against HIV-1 rev transcripts in human cells Nat Biotechnol 2002 20 500 505 11981565
Nunlist EH Dozmorov I Tang Y Cowan R Centola M Lin H-K Partitioning of 5α-dihydrotestosterone and 5α-androstane-3α, 17β-diol activated pathways for stimulating human prostate cancer LNCaP cell proliferation J Steroid Biochem Mol Biol 2004 91 157 170 15276623 10.1016/j.jsbmb.2004.02.008
Kerr JFR Harmon BV Tomei LD, Cope FO Definition and incidence of apoptosis: an historical perspective Apoptosis The Molecular Basis of Cell Death 1991 Plainview, NY: Cold Spring Harbor Laboratory Press 5 29
| 15813967 | PMC1087496 | CC BY | 2021-01-04 16:40:10 | no | Cancer Cell Int. 2005 Apr 6; 5:8 | utf-8 | Cancer Cell Int | 2,005 | 10.1186/1475-2867-5-8 | oa_comm |
==== Front
Curr Control Trials Cardiovasc MedCurrent Controlled Trials in Cardiovascular Medicine1468-67081468-6694BioMed Central 1468-6708-6-51581397110.1186/1468-6708-6-5ReviewThe Blood Pressure "Uncertainty Range" – a pragmatic approach to overcome current diagnostic uncertainties (II) Pater Cornel [email protected] Chippenham, UK2005 6 4 2005 6 1 5 5 4 3 2005 6 4 2005 Copyright © 2005 Pater; licensee BioMed Central Ltd.2005Pater; 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.
A tremendous amount of scientific evidence regarding the physiology and physiopathology of high blood pressure combined with a sophisticated therapeutic arsenal is at the disposal of the medical community to counteract the overall public health burden of hypertension. Ample evidence has also been gathered from a multitude of large-scale randomized trials indicating the beneficial effects of current treatment strategies in terms of reduced hypertension-related morbidity and mortality.
In spite of these impressive advances and, deeply disappointingly from a public health perspective, the real picture of hypertension management is overshadowed by widespread diagnostic inaccuracies (underdiagnosis, overdiagnosis) as well as by treatment failures generated by undertreatment, overtreatment, and misuse of medications.
The scientific, medical and patient communities as well as decision-makers worldwide are striving for greatest possible health gains from available resources.
A seemingly well-crystallised reasoning is that comprehensive strategic approaches must not only target hypertension as a pathological entity, but rather, take into account the wider environment in which hypertension is a major risk factor for cardiovascular disease carrying a great deal of our inheritance, and its interplay in the constellation of other, well-known, modifiable risk factors, i.e., attention is to be switched from one's "blood pressure level" to one's absolute cardiovascular risk and its determinants. Likewise, a risk/benefit assessment in each individual case is required in order to achieve best possible results.
Nevertheless, it is of paramount importance to insure generalizability of ABPM use in clinical practice with the aim of improving the accuracy of a first diagnosis for both individual treatment and clinical research purposes. Widespread adoption of the method requires quick adjustment of current guidelines, development of appropriate technology infrastructure and training of staff (i.e., education, decision support, and information systems for practitioners and patients). Progress can be achieved in a few years, or in the next 25 years.
==== Body
Introduction
During the past decades, hypertension, denoting abnormal elevation of blood pressure, has commonly been assigned a distinct disease quality. The majority of the medical community as well as renowned medical textbooks have considered it a pathological entity requiring diagnostic and appropriate treatment in most individuals having it.
Inherent in the 100-year old approach of discriminating between normal and abnormal blood pressure is, however, an arbitrary threshold established currently at 140/90 mmHg for mild hypertension and two higher cut-off values to define moderate and severe hypertension. This classification, still maintained as such by the ESH [1] and WHO [2] guidelines, has been in use for many decades both for management of hypertension in individual subjects and for defining patient population samples targeted for testing of new antihypertensive drugs.
It is common knowledge, however, that up to 95% of hypertensive individuals have high blood pressure of unknown etiology denoted essential hypertension. The majority of these have mild hypertension while some 10% of them have either moderate or severe degrees of hypertension.
Classically, the diagnosis of essential hypertension has been regarded as pretty straightforward, as long as a secondary hypertension could be ruled out with confidence.
While this latter entity can, on justifiable grounds, be considered a disease, essential hypertension is a quantitative expression of the fluctuations of a biologic variable – the blood pressure – that should be considered a major risk factor for cardiovascular disease at best, rather than a disease by itself.
Nevertheless, with increasing awareness that the trade-off between normality and abnormality on a blood pressure curve is confounded by large diurnal and random variation of the blood pressure variable, by individual factors such as age, sex, race as well as by a great number of potential errors occurring during blood pressure measurement per se, the medical and scientific community are desperately in search for approaches to increase the accuracy of diagnosis and management of hypertension and to shift the weight of decision making from the current "BP-value"-focused attitude to the appraisal of any patient's absolute risk, rather than his/her individual risks (i.e., high blood pressure, hypercholesterolemia, smoking, overweight, etc.) [3-5].
The driver behind the need for radical change is at least two fold:
1. The widespread awareness that the hypertension-related risk – in terms of cardiovascular complications, stroke and renal disease – rises in relation to increases of both systolic and diastolic blood pressure [6-13].
2. The hypertension diagnosis currently implies not only exclusion of secondary causes of hypertension but as well, careful consideration is to be given to differentiating entities like: white coat hypertension [14-17], white coat effect [18,19], masked hypertension [20-22] and prehypertension, from the genuine, sustained hypertension [23].
The clustering of these entities around the current diagnostic cut-off point (140/90 mmHg) generates a BP uncertainty range (130/85-160/95 mmHg) reflecting a universally widespread inaccuracy of a first diagnosis based on office BP measurements and a consequent inappropriate long-term management of the subjects assessed.
Blood pressure measurement as a diagnostic test – a clinical epidemiology perspective
Commonly, variation in clinical medicine may be due to fluctuations of biologic variables or the presence or absence of disease as well as the nature of that disease and its severity; it may also be due to differences in measurement technique, errors in measurement, observer bias, and to a great extent to random variation.
The blood pressure variable happens to be a typical example that displays all the aforementioned variation parameters. An attempt to understand this variation through gaining insight in relative simple epidemiology and statistical concepts and then acting on the basis of the new acquired knowledge, might be the key to account for the blood pressure overall variability, to accurately interpret the prognostic significance of any blood pressure values and confidently manage pharmacological therapy in all hypertensive patients.
In the general population blood pressure values follow a smooth bell-shaped distribution as displayed in Fig. 1 for the SBP and in Fig. 2 for the DBP. The two histograms attempt to account for the known general prevalence of hypertension worldwide [24], with higher figures for the SBP as compared with DBP due to their divergent pattern at higher ages (and both sexes) [25]. The black bars highlight the frequency distribution of the blood pressure values considered abnormal (according to the arbitrary threshold of 140/90 mmHg).
Figure 1 Normal distribution of SBP values in the general population.
Figure 2 Normal distribution of DBP values in the general population.
The distribution of SBP is slightly skewed to the right due to the high proportion of people with BP values in the range 140-160 mmHg [26].
Certainly, both systolic and diastolic blood pressure belong to one and the same, simplified, theoretical normal distribution (Fig. 3) assumed to describe the underlying population from which relevant data might be derived [27,28]. Fig. 3 suggests that all normal blood pressure values measured in the population are comprised by the interval given by the mean ± two standard deviations (M ± 2SD).
Figure 3 Illustration of the normality concept in a normal distribution.
Assuming that the commonly used office blood pressure measurement was an ideal test (a "gold standard", which is not the case), it could separate all healthy people from those who have the disease (i.e., people with BP values that require treatment from those who do not). In such a hypothetical case the blood pressure measurement would be 100% sensitive and 100% specific, with no false-positive or false-negative results.
In reality, things are much different; there are two normal distributions of blood pressure "test" results. One is for individuals free of disease and one for individuals who have the disease. Framing the reality in this way is still misleading because in practice many patients may have raised blood pressure levels (but are "disease free", insofar as they may not need pharmacologic treatment – e.g., white coat hypertension) [29,30] and, many patients may display apparently normal blood pressure values but, definitely need pharmacologic treatment (because of high absolute risk for CVD, the presence of a compelling indication or of masked hypertension) [29,30].
In fact, there is an overlap of the two distributions and that overlap is rather large; it predisposes to both over- and underdiagnosis, i.e., false-positive and false-negative results. The immediate consequence under the current circumstances of conventional office BP measurement is that setting any test value (i.e., BP threshold) as cut-off point to distinguish between "normal" and "abnormal", will misclassify a great proportion of the patients falling into the overlap area (Fig. 4) – a real uncertainty range exposing to misdiagnosis (i.e., under- and overdiagnosis) [30].
Figure 4 Result alternatives by office BP measurement and the trade-off displayed by the sensitivity and specificity of the method as related to the cut-off point selected (140/90 mmHg).
As Fig. 4 illustrates, for our diagnostic "test" (the office BP measurement) which does not behave as a "gold standard", there are four distinct alternatives [31]:
1. True negatives (TN) – subjects without the disease who test negative (i.e., normal BP and no indication for blood pressure lowering treatment).
2. True positives (TP) – subjects who have the disease and test positive (i.e., sustained hypertension requiring pharmacologic treatment).
3. False-negatives (FN) – subjects who have the disease but test negative (i.e., normal BP, however, in need to get pharmacologic treatment (e.g., masked hypertension).
4. False-positives (FP) – subjects who do not have the disease but test positive (i.e., subjects with high blood pressure values but in no need for pharmacologic treatment (e.g., white coat hypertension).
Fig. 4 also illustrates the dynamic complexity of what goes on around the cut-off point selected. As mentioned in the first part of this paper, many national hypertension societies still maintain a cut-off point for blood pressure normality/abnormality by 160/95 mmHg [32]. For those societies which decided to select 140/90 mmHg as cut-off point, the switch from 160/95 to 140/90 mmHg has meant a substantial change of several important diagnostic parameters (including the prevalence of the condition) [33].
Namely, the sensitivity of the test has increased, however, with a simultaneous decrease in its specificity. The number of TPs has increased as compared to the number of FNs. Further, the move of the cut-off point to the left on the BP curve has caused an increase in FPs as compared to the FNs.
With other words, a high cut-of point on the BP curve implies less false-positive results (and less overdiagnosis) with the reverse occurring if the cut-off point is low.
Selection of "best cut-off point" can be enhanced by constructing a receiver operating characteristic (ROC) curve (Fig. 5) [34]. Such a curve displays sensitivity on the X-axis and the false positive error rate (1-specificity) on the Y-axis. These two parameters can be computed for different threshold values on the basis of points plotted on the graph, as results of blood pressure measurements in individuals with known health status (hypertension/no hypertension).
Figure 5 Receiver operating characteristic curve exploring the proper cut-off point for blood pressure measurement.
A cut-off point set at 0 would mean a "test" (i.e., BP measurement) with 100% sensitivity and ability to detect all patients with genuine hypertension (i.e., sustained hypertension). However, all normal individuals (i.e., without hypertension) would screen positive, as reflected by a 100% false-positive error rate (overdiagnosis). The corresponding point would be placed in the right upper corner of the graph.
In the other extreme, a hypothetical cut-off point of 350 mmHg, for example, would imply that virtually all hypertensive patients were missed (underdiagnosis), reflected by 0% sensitivity. This point would be placed in the lower left corner of the graph.
Using a similar reasoning, the sensitivity and the false-positive error rate can be computed for increasing threshold values. Connecting the points plotted on the graph would generate a ROC curve.
As the left upper corner represents a sensitivity of 100% and 0% false-positive rate, the real best cut-off point is the one lying closest to it.
Fig. 5 suggests that the 140/90 mmHg is, as a matter of fact, the best cut-off point, a fact that should satisfy all those who are sceptical as to the value of this particular threshold value. Obviously, its use does not preclude a refined physician-patient dialog aiming at accounting for the patients view regarding a particular antihypertensive treatment, as Campbell and Murchie put it: "Appropriate management of blood pressure should be guided by an informed dialogue between patients and doctors and not by blind pursuit of blood pressure targets"[35].
In fact, despite the drawbacks mentioned above, of an arbitrarily selected office blood pressure cut-off point for practical purposes, the office BP measurement as a method as such, need not be entirely discarded.
On the contrary, there are three different instances in which it may be used with certainty to decide whether to treat or not to treat patients. The first two instances are discernable from Fig. 4 and pertain to individuals without any compelling indication (i.e.., diabetes mellitus or renal dysfunction):
1). The great majority of individuals with office BP > 160/95 mmHg are genuine hypertensives. They may have sustained essential hypertension (or a secondary form of hypertension) [36]. Approximately 50% of myocardial infarctions and one third of strokes are known to be associated with BP > 160/95 mmHg. The presence of one or several other risk factors further increase the certainty that these patients need pharmacological treatment and careful, long-term follow-up [5,37].
2). Likewise, the great majority of individuals with office BP < 130/80 mmHg are likely to be free of hypertension and/or the need for pharmacologic blood pressure lowering treatment [38]. Regular health checks, for example at two years intervals, are likely to capture propensity toward increased risk, particularly among individuals who, according to the current classification belong to the high-normal prehypertension category (130/85-139/89 mmHg).
In this context, it is worth re-emphasizing that blood pressure values below the 140/90 mmHg cut-off point do not confer total protection against a cardiovascular events. Data from Framingham Study have shown that more than one half (57%) of all heart attacks and almost one half of all strokes in some population studies occur in persons with normal office blood pressure [39-41].
3). The third instance pertains to patients with known diabetes mellitus and renal dysfunction. For these patients, the office blood pressure measurement functions merely as a quasi-"gold standard" test.
Fig. 6 illustrates the concept of two population distributions: one of diabetics without hypertension and another of diabetics with associated hypertension. Given the widespread consensus that patients with diabetes and associated hypertension should get pharmacological treatment whenever BP > 130/80 mmHg [42,43], the two distributions do not appear to have any degree of overlap, with the cut-off point of 130/80 mmHg discriminating well between those supposed to be treated pharmacologically, from those who are not.
Figure 6 Hypothetical, non-overlapping distributions of diabetics and diabetics with associated hypertension patients.
A justifiable question that may arise is why a somewhat lower cut-off point, i.e., 130/80 mmHg, can be considered as more reliable than the 140/90 mmHg, given that office BP measurement is likely to carry the same sort of problems in both cases.
The answer is pretty simple. An assumed overlap of the two distributions (Fig. 6), leading to FP assessments (overdiagnosis) and FNs (underdiagnosis), would cause neither diagnosis nor management concerns. Namely, patients with diabetes and associated hypertension need aggressive antihypertensive treatment meant to lower BP as much as possible [44-47]. Blood pressure measurement errors around the cut-off point value of 130/80 mmHg are, therefore, likely to have neither clinical nor prognostic relevance, as long as the diabetes mellitus diagnosis is certain.
Furthermore, Fig. 6 depicts a second cut-off point relevant for the population of diabetics with associated hypertension: the 150/90 mmHg, which, according to current guidelines is an indication for use of combination therapy (two drugs from the start).
At closer scrutiny of the uncertainty range (≥ 130/85 to <160/95 mmHg) (see Fig. 7) from its lower bound upward, as assessed by office blood pressure measurements, it appears to be a mix of populations consisting of:
Figure 7 The uncertainty range and its mix of hypertension categories. Accurate diagnosis is only possible by ABPM.
• Subjects with prehypertension (high-normal BP ≥ 130/85-139/89 mmHg)
• Patients with masked hypertension
• Subjects with white coat hypertension
• Patients with sustained hypertension
Fig. 8 attempts to depict the frequency distribution of the aforementioned patient categories in a single, common population distribution. It suggests that a great part of the total population lies either in the in the uncertainty range or under the area consisting of genuine hypertensive patients (i.e., with sustained hypertension).
Figure 8 Frequency distribution of hypertensive categories as parts of a common, normal distribution.
A recent analysis of the worldwide global burden of hypertension [48] indicated that 26.4% of the adult population in 2000 had hypertension and that 29.2% were projected to have the condition by 2025. The hypertension prevalence is, however, considerably much higher in economically developed countries as compared with economically developing countries (37.3% versus 29.9%) [48].
A multinational sample surveys carried out in six European countries, Canada and US, in the 1990s indicated the age-adjusted prevalence of hypertension was 28% in the North American countries and 44% in the European countries [24].
About 59 million American adults (29%) fall into prehypertension category (SBP 120-139 mmHg or DBP 80-89 mmHg) [49].
Extrapolating roughly from the above figures, the prevalence of hypertension in economically developed countries (37.3%) and the estimation of prehypertension in US (29%) would add up to a 66.3% of the entire population having a form of hypertension or prehypertension. This average figure might be somewhat lower in America and Canada but may reach 75% in the European countries.
Furthermore, epidemiological data [50] indicate that approximately 25% of the community burden of BP-related CVD is occurring among the population of hypertensives with systolic BP ≥ 160 mmHg (representing only approximately 5% of the total number of hypertensives). About 33% of all BP-related CVD events are likely to occur in the persons within normotensive BP range (<140 mmHg) while more than 66% of the same burden can be attributed to the patient population with systolic BP ≥ 140 mmHg, i.e., belonging to the uncertainty range. Given that the lower bound of the uncertainty range extends to 130/85 mmHg (to include the high-normal prehypertension), it might encompass as much as 80% of the total community burden of BP-related CVD risk.
Prehypertension (PH)
The new JNC-7-hypertension category [49] was subject to heavy criticism, primarily by European scientists who argued that the new entity might have negative psychological impact and generate a wave of more or less unnecessary investigations in fairly healthy individuals.
This is, certainly, not a strong argument even if it may contain a grail of truth. Obviously, the sudden and unexpected awareness that one, despite knowing him/herself to be in good health, belong to a group of people "at risk for heart disease", might trigger a more or less uncomfortable feeling.
However, people with BP values in the range of 130-139/85-89 mmHg, known to have high normal blood pressure and labelled as being prehypertensives, are also known to run a considerable higher risk for cardiovascular disease than people with optimal (<120/80 mmHg) and with normal BP (120-129/80-84 mmHg) [49,51,52].
In a survey of 9845 Framingham Heart Study participants over a 4-year period, 43% of the total of 1907 individuals (19% of the original sample) who had high-normal blood pressure at the initial screening have developed hypertension [53]. In contrast, only 6% of the subjects with optimal blood pressure and 20% of those with normal blood pressure have developed hypertension [53]. Hypertension incidence in all three categories increased with age, reaching 37% among subjects with high-normal BP who were 35-64 years old and 50% among subjects aged 65 or older, likewise, with high-normal blood pressure at screening. Worthwhile emphasizing, about 15% of the subjects with high-normal blood pressure have progressed to stage II or greater degree of hypertension over 4 years of follow-up.
Overall, compared with optimal blood pressure, high-normal blood pressure was associated with a 5- (age 35-64 years) to 12-fold (age 65 years or over) elevated odds of hypertension on follow-up [53], justifying inclusion of these subjects in the uncertainty area and thereby, the consequent need for accurate diagnosis from the outset as well as appropriate long-term management of these individuals.
Masked Hypertension (MH)
As depicted by Fig. 7, individuals with MH fall into the false negative area of the uncertainty range, suggesting that the condition per se is perceived as normotension as assessed by office blood pressure measurement. However, MH is genuine hypertension, as assessed by daytime ABP of >135/85 mmHg [54-58].
The reported prevalence rates of MH are 9% [59], 14% [60], 23% [61] and 31% [62]. Individuals with masked hypertension were shown to be similar with true hypertensive patients in terms of left ventricular characteristics, carotid artery wall thickness, and prevalence of discrete atherosclerotic plaques [63] but to be different on several demographic and lifestyle variables (greater proportion of males, older, greater degree of alcohol consumption, past smoking) [60,63].
Missed diagnosis of MH in these patients leaves them untreated and at risk for long-term consequences of hypertension.
Selenta et al. [60] have demonstrated that the BP values falling in the borderline range (10 mmHg above and below 140/90 mmHg) are particularly inaccurate. Most participants presumed healthy in the 10 mmHg range below 140/90 mmHg have hypertensive ABP, i.e., MH, commonly missed by the office BP measurement. The authors of the same article concluded that only those office readings averaging 20 points above or below the 140/90 mmHg cut-off, represent safe diagnostic information.
The high prevalence rates of MH and the high level of misdiagnosis rate by office BP measurement of the condition calls for generalization of ABPM use in clinical practice, for diagnosis and management purposes of this large patient population.
White coat hypertension (WCH)
Subjects with white coat hypertension have a normal average daytime blood pressure outside a medical setting [64,65] but present with high BP in the medical environment [66,67]. By definition [68,69], white coat hypertension is diagnosed as such, if the conventional BP is persistently ≥ 140/90 mmHg and the average daytime ambulatory BP is below 135/85 mmHg.
The prevalence rate of the WCH is reported to be in the area of 15 to 35% of patients in whom hypertension is diagnosed [70,71] and in nearly 30% of pregnant women [72].
The use of a distinct cut-off point is important for diagnostic and management purposes. It also distinguishes WCH from the white coat effect, the latter being a quantitative measure of the blood pressure rise in the presence of a physician. This transient blood pressure rise has been quantified by Mancia et al. [73] who demonstrated mean value increases of 27 mmHg for both systolic and diastolic pressure when measurement was done in the presence of a physician. The white coat effect is commonly defined as an office BP exceeding mean daytime ambulatory BP by at least 20 mmHg systolic and/or 10 mmHg diastolic [74]. Such a large white coat effect has been found in as many as 73% of treated hypertensive subjects and it may occur more frequently in women than in men [75,76].
As illustrated by Fig. 7, both WCH as a distinct entity and the white coat effect as a transitory BP rise, fall into the false positive area of the uncertainty range, i.e., beyond the 140/90 mmHg cut-off point as assessed by office BP measurement, and count thereby as overdiagnosis.
Indeed, the consequences of failing to identify white coat hypertension are considerable. People may be penalized for insurance and pension policies, and for employment [77].
Long-term treatment may be prescribed unnecessarily with all the risks derived from potential treatment-emergent adverse reactions [78-80].
Given the seriousness as well as the high risk for such consequences in the clinical practice, several hypertension guidelines recommend use of ABPM for diagnosis of WCH [81-83].
In 2001, the CMS in USA has selected "patients with suspected WCH" as having indication for ABPM, with the use of method itself being reimbursed [84].
Speculations on what clinical characteristics might suggest the presence of WCH and thereby the need for ABPM, in an attempt to preclude indiscriminate use of the method, is certainly not justified anymore [85]. On the contrary, current evidence argues in favour of ABPM use in all patients with office blood pressure falling in the uncertainty range (≥ 130/85 to <160/95 mmHg).
Once WCH has been diagnosed, ABPM should be repeated at annual or biannual intervals [86] with the aim to capture increased cardiovascular risk that would justify initiation of drug treatment [87,88].
A holistic approach toward hypertension management
Fig. 7 suggests that ABPM should be used for further diagnostic work up of subjects whose office-measured BP values fall in the uncertainty range. Indeed, combining office blood pressure (OBP) measurement values with ABPM recording results (daytime normality cut-off point of 135/85 mmHg) allows for discrimination among the following diagnostic entities (see Fig. 9):
Figure 9 ABPM assessment of subjects with BP values falling in the uncertainty range.
• Sustained Hypertension (SH) – (ABPM > 140/90 and OBP ≥ 140/90)
• Masked Hypertension (MH) – (ABPM > 135/85 and OBP < 140/90)
• White Coat Hypertension (WCH) – (ABPM < 135/85 and OBP ≥ 140/90)
• Prehypertension (high-normal BP) (PH) – (ABPM < 135/85 and OBP 130/85-139/89)
The specific thresholds used for diagnostic discriminatory purposes can also be used for management purposes. It is widely agreed that poor control of hypertension is defined by BP values > 140/90 mmHg [89] while good control, in terms of ABPM, counts ≤ 135/85 mmHg [90-92]. Likewise, there is widespread agreement that patients with clinic pressures of = 160/95 mmHg need drug treatment [93].
The well-known, multifactorial inaccuracies imbedded in the office blood pressure measurement and, in contrast to that, the widespread agreement as to the superiority and multi-purpose use of ABPM [94-102], as well as reports on its cost-effectiveness [103] [104], make the method to emerge as an additional alternative to the conventional approach of hypertension management. While ABPM can, on no account, be a replacement for the conventional office blood pressure measurement, it is increasingly obvious that its use is a sine qua non condition for accurate diagnosis and management of subjects belonging to a large segment of the general population displaying different forms of high blood pressure-related entities (falling in the uncertainty range).
More recent studies provide evidence that the ABP testing is generally well-accepted and tolerated by patients. A survey of 177 patients, who underwent ABPM in primary care office setting in US over a 2.5 years period [103], showed that 75% of them considered the 24-hour ABPM test as worthwhile, with respect to the time and money incurred by the investigation. Ninety per cent of the patients thought that the results of the test were to provide useful information for the physician's decision making on appropriate therapy. Similar results were derived from a qualitative study of ABPM in UK [71]. Both studies have emphasized, however, the importance of the explanation given by the physician to the patient as to the benefit of undergoing ABPM testing, in order to minimize the perception of discomfort related to the 24-hour recording.
Fig. 10 is an algorithmic approach to the management of all subjects who display office blood pressure falling in the uncertainty range (≥ 130/85 to < 160/95 mmHg). The top part of the figure suggests that two different units are involved in the management/investigation of such patients, a usual GP or specialist office and a laboratory specialised in ABPM recordings and results interpretation.
Figure 10 An algorithmic approach to management of subjects with OBP in the uncertainty range (≥ 130/85 to <160/95 mmHg).
An initial assessment is commonly performed in the GP's office where blood pressure is recorded, according to current guidelines, by the physician or a trained nurse. Patients whose average BP, derived from two measurements per visit, in two or several visits, lies in the uncertainty range should be referred for 24-hour ABP monitoring.
The result of the daytime ABPM assessed together with the OBP generates a diagnostic conclusion matching one of the following alternatives: SH/MH/WCH or PH (high-normal level).
Amazingly, the current conventional cut-off point of 140/90 mmHg looses entirely its value for decision making, at least in the initial stage assessment.
Indeed, as mentioned above, much information may be lost when sensitivity and specificity are defined in relation to a single cut-off point value of a continuous variable (such as the BP). Instead, using a range, i.e., the interval between two cut-off points (≥ 130/85 to <160/95 mmHg) for further decision making, avoids the well-known diagnostic uncertainties.
The 140/90 cut-off point retains, however, its (arbitrary) value in the next stage of ABPM investigation. The value of the OBP, whether below or above the 140/90 mmHg threshold is corroborated with the ABPM result for selection of one out of four different diagnostic alternatives: sustained hypertension, masked hypertension, white coat effect or high-normal prehypertension.
Once diagnosed, the patient returns to his/her physician who remains in charge with the further management decision whatever the diagnosis might be (including lifestyle changes, drug treatment, follow-up, etc.).
Obviously, diagnosis and management of patients who have any form of symptomatic atherosclerotic vascular disease including previous myocardial infarction, by pass graft surgery, angina, stroke or transient ischaemic attack, peripheral vascular disease or atherosclerotic renovascular disease, need treatment of even mild hypertension (≥ 140/90 mmHg) for secondary prevention. Likewise, patients with target organ damage such as LVH, heart failure, proteinuria or renal impairment need treatment of even very mild hypertension.
Patients with type I and II diabetes mellitus and associated mild hypertension (≥ 130/80 mmHg) generally have diabetic nephropathy and should be treated.
All the aforementioned compelling indications require drug treatment for any level of raised blood pressure.
This translates in the need to apply the algorithm in Fig. 10 and perform formal absolute risk assessment only in patients with uncomplicated hypertension (i.e., with BP values falling in the uncertainty range ≥ 130/85 to <160/95 mmHg).
Conclusion
Accurate diagnosis and management of high blood pressure is of paramount importance for the prevention of long-term, cardiovascular, cerebrovascular and renal complications. Aggressive attempts to identify and treat "high blood pressure values" must be balanced carefully with the risks of overdiagnosis and overtreatment in these patients.
ABPM has a proven value not only as a research tool but also as a valuable investigative method for a large segment of the hypertensive population belonging to the uncertainty range.
Practical management of these patients, once accurate diagnosis has been established, should be based on absolute cardiovascular disease risk and on a risk-communication dialog between the physician and patient as well as on their mutual agreement regarding the specific treatment to be initiated and the appropriate long-term follow-up.
Competing interests
The author(s) declare that they have no competing interests.
==== Refs
2003 European Society of Hypertension – European Society of Cardiology guidelines for the management of arterial hypertension Journal of Hypertension 2003 21 1011 1053 12777938 10.1097/00004872-200306000-00001
World Health Organization, International Society of Hypertension Writing Group 2003 World Health Organization (WHO/International Society of Hypertension (ISH) statement on management of hypertension J Hypertens 2003 21 1983 1992 14597836
Anderson KV Odell PM Wilson PWF Kannel WB Cardiovascular disease risk profiles Am Heart J 1991 121 293 298 1985385 10.1016/0002-8703(91)90861-B
Jackson RJ Lawes CMM Bennett DA Milne RJ Rodgers A Treating Individuals 5: Treatment with drugs to lower blood pressure and blood cholesterol based on an individual's absolute cardiovascular risk Lancet 2005 365 434 441 15680460
Volpe M Alderman MH Furberg CD Jackson R Kostis JB Laragh JH Psaty BM Ruilope LM Beyond hypertension. Toward guidelines for cardiovascular risk reduction Am J Hypertens 2004 17 11 1068 1074 15533736 10.1016/j.amjhyper.2004.06.017
Mancia G Grassi G Systolic and diastolic blood pressure control in antihypertensive drug trials J Hypertens 2002 20 1461 1464 12172300 10.1097/00004872-200208000-00001
Prospective Studies Collaboration Age-specific relevance of usual blood pressure to vascular mortality: a meta-analysis of individual data for one million adults in 61 prospective studies Lancet 2002 360 1903 1913 12493255 10.1016/S0140-6736(02)11911-8
Law MR Wald NJ Risk factor thresholds: their existence under scrutiny BMJ 2002 324 1570 1576 12089098 10.1136/bmj.324.7353.1570
Rodgers A Murray CJL, Lopez Ad Quantifying selected major risks to health The world health report 2002: reducing risks, promoting healthy life 2002 Geneva: WHO 47 92
Neaton JD Wenthworth D for the Multiple Risk Factor Intervention Trial Research Group Serum Cholesterol, blood pressure, cigarette smoking, and death from coronary heart disease. Overall findings and differences by age for 316,099 white men Arch Intern Med 1992 152 56 54 1728930 10.1001/archinte.152.1.56
Asia Pacific Cohort Studies Collaboration Cholesterol, coronary heart disease and stroke in the Asia Pacific Region Int J Epidemiol 2003 32 563 572 12913030 10.1093/ije/dyg106
Asia Pacific Cohort Studies Collaboration Blood pressure and Cardiovascular disease in the Asia Pacific Region J Hypertens 2003 21 707 716 12658016
Eastern Stroke and Coronary Heart Disease collaborative Research Group Blood pressure, cholesterol, and stroke in Eastern Asia Lancet 1998 352 1801 1807 9851379 10.1016/S0140-6736(98)03454-0
William BW Ambulatory blood-pressure monitoring in clinical practice New Engl J Med 2003 348 24 2377 2378 12802022 10.1056/NEJMp030057
Little P Barnet J Barnsley L Marjoram J Fitzgerald-Baron A Mant D Comparison of agreement between different measures of blood pressure in primary care and daytime ambulatory blood pressure BMJ 2002 325 254 259 12153923 10.1136/bmj.325.7358.254
Bellomo G Narducci PL Rondoni F Pastorelli G Stagnoni G Prognostic value of 24-hour blood pressure in pregnancy JAMA 1999 282 1447 1452 10535435 10.1001/jama.282.15.1447
Staessen JA Fagard R Thjis L Randomized double-blind comparison of placebo and active treatment for older patients with isolated systolic hypertension Lancet 1997 350 757 764 9297994 10.1016/S0140-6736(97)05381-6
O'Brian E Ambulatory blood pressure monitoring in the management of hypertension Heart 2003 89 571 576 12695477 10.1136/heart.89.5.571
MacDonald MB Laing GP Wilson MP Wilson TW Prevalence and predictors of white-coat response in patients with treated hypertension Can Med Assoc J 1999 161 265 269 10463047
Pickering T Davidson K Gerin W Masked hypertension Hypertension 2002 40 795 796 12468559 10.1161/01.HYP.0000038733.08436.98
Guillaume B Chatellier G Genes N Vaur L Vaisse B Menard J Mallion JM Cardiovascular prognosis of "Masked Hypertension" detected by blood pressure self-measurement in elderly treated hypertensive patients JAMA 2004 291 1342 1349 15026401 10.1001/jama.291.11.1342
Guillaume B Chatellier G Genes N Vaur L Vaisse B Menard J Mallion JM Cardiovascular prognosis of "Masked Hypertension" detected by blood pressure self-measurement in elderly treated hypertensive patients JAMA 2004 291 1342 1349 15026401 10.1001/jama.291.11.1342
Mulrow PJ Izzo JL, Black HR, Goodfriend TL Hypertension: a worldwide epidemic Hypertension primer: the essentials of high blood pressure 1999 2 Baltimore: Williams and Wilkins 721 273
Wolf-Maier K Cooper RS Banegas JR Gianpaoli S Hense HW Joffres M Kastarinen M Poulter N Primatesta P Rodriguez-Artalejo F Stegmayr B Thamm M Tuomilehto J Vanuzzo D Vescio F Hypertension Prevalence and blood pressure levels in 6 European countries, Canada, and the United States JAMA 2003 289 2363 2369 12746359 10.1001/jama.289.18.2363
Kotchen JM McKean HE Kotchen TA Blood pressure trends with aging Hypertension 1982 4 suppl 3 11 129
Franklin SS Larson MG Khan SA Does the relation of blood pressure to coronary heart disease risk change with aging? The Framingham Heart Study Circulation 2001 103 1245 1249 11238268
Pitt B Julian D Pocock S Clinical Trials in Cardiology 1997 WB Saunders Company Ltd., London
Mower WR Evaluating bias and variability in diagnostic test reports Ann Emerg Med 1999 33 85 91 9867892
Patterson RA Horowitz SE Importance of epidemiology and biostatistics in deciding clinical strategies for using diagnostic tests: a simplified approach using examples from coronary artery disease J Am Coll Cardiol 1989 13 1653 2656825
William BW Ambulatory blood-pressure monitoring in clinical practice New Engl J Med 2003 348 24 2377 2378 12802022 10.1056/NEJMp030057
Kottke TE Stroebel R Hofman R JNC-7 it's more than high blood pressure JAMA 2003 289 19 2573 2575 12748200 10.1001/jama.289.19.2573
Mulrow PJ Izzo JL, Black HR, Goodfriend TL Hypertension: a worldwide epidemic Hypertension primer: the essentials of high blood pressure 1999 2 Baltimore: Williams and Wilkins 721 723
Jekel JF Elmore JG Katz DL Epidemiology Biostatistics and Preventive Medicine 1996 WB Saunders Company Ltd., Philadelphia
Mayer D Essential Evidence-Based Medicine 2004 Cambridge University Press, Cambridge
Campbell NC Murchie P Treating hypertension with guidelines in general practice BMJ 2004 329 523 524 15345603 10.1136/bmj.329.7465.523
Wong ND Black HR Gardin JM Preventive Cardiology 2000 McGraw Hill companies, Inc., New York
Alderman M Blood Pressure Measurement. Individualized Treatment Based on Absolute Risk and the Potential for Benefit Ann Intern Med 1993 119 329 335 8328743
Prospective Studies Collaboration Age-specific relevance of usual blood pressure to vascular mortality a meta-analysis of individual data for one million adults in 61 prospective studies Lancet 2002 360 9343
Alderman MH Blood Pressure Management: Individualized Treatment Based on Absolute Risk and Potential for Benefit Ann Intern Med 1993 119 329 335 8328743
Kannel WB Some lessons in cardiovascular epidemiology from Framingham Am J Cardiol 1976 37 269 281 1246956 10.1016/0002-9149(76)90323-4
Kannel WB Wolf PA Verter J McNamara PM Epidemiologic assessment of the role of blood pressure in stroke. The Framingham Study JAMA 1970 214 301 310 5469068 10.1001/jama.214.2.301
Arauz-Pacheco C Parrott MA Raskin P The Treatment of hypertension in adult patients with diabetes Diabetes Care 2002 25 1 134 147 11772914
Bakris GL Weir MR Achieving goal blood pressure in patients with type 2 diabetes: combination versus fixed-dose combination approaches J Clin Hypertens 2003 5 3 202 209
Deedwania PC Hypertension and diabetes: new therapeutic options Arch Intern Med 2000 160 1585 1594 10847251 10.1001/archinte.160.11.1585
American Diabetes Association American diabetes Association: clinical practice recommendations 2002 Diabetes Care 2002 25 suppl 1 S1 S147 11788484
Staessen JA Wang JG Thijs L Cardiovascular protection and blood pressure reduction: a meta-analysis Lancet 2001 358 1305 1315 11684211 10.1016/S0140-6736(01)06411-X
Gradman AH Acevedo C Evolving strategies for the use of combination therapy in hypertension Curr Hypertens Rep 2002 4 5 343 349 12217251
Kearney PM Whelton M Reynolds K Muntner P Whelton PK He J Global burden of hypertension: analysis of worldwide data Lancet 2005 365 217 223 15652604
Chobanian AV BAkris GL Black HR Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure, National Heart, Lung, and Blood Institute; National High Blood Pressure Education Program Coordinating Committee Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure Hypertension 2003 42 1206 1252 14656957 10.1161/01.HYP.0000107251.49515.c2
Stamler J Stamler R Neaton JD Blood pressure, systolic and diastolic, and cardiovascular risks. US population data Arch Intern Med 1993 153 598 625 8439223 10.1001/archinte.153.5.598
World Health Organization-International Society of Hypertension Guidelines for Management of Hypertension. Guidelines Subcommittee J Hypertens 1999 17 151 183 10067786
Hotline Editorial Rates of progression to hypertension among non-hypertensive subjects: implications for blood pressure screening Eur Heart J 2002 23 6067 1070
Vasan RS Larsan MG Leip EP Kannel WB Levy D Assessment of frequency of progression to hypertension in non- hypertensive participants in the Framingham Heart Study: a cohort study Lancet 2001 358 1682 1686 11728544 10.1016/S0140-6736(01)06710-1
Pickering TG Utility of 24-hour ambulatory blood pressure monitoring in clinical practice Can J Cardiol 1995 11 43H 48H 7850664
Staessen J Fagard R Lijnen P Thijs L van Hoof R Amery A Ambulatory blood pressure monitoring in clinical trials J Hypertens Suppl 1991 9 suppl 1 S13 S19 2040902
Staessen J Fagard R Lijnen P Thijs L van Hoof R Amery A Reference values for ambulatory blood pressure: a meta-analysis J Hypertens 1990 8 suppl 6 S57 S64
O'Brien E Murphy J Tyndall A Twenty-four-hour ambulatory blood pressure in men and women aged 17 to 80 years: the Allied Irish Bank Study J Hypertens 1990 9 355 360 1646262
ISH Scientific Committee Consensus document on non-invasive ambulatory blood pressure monitoring J Hypertens 1990 8 suppl 6 S136 S140
Sega R Trocino G Lanzarotti A Carugo S Cesana G Schiavina R Valagussa F Bombelli M Giannattasio C Zanchetti A Mancia G Alterations of cardiac structure in patients with isolated office, ambulatory, or home hypertension: data from the general population (Pressione Arteriose Monitorate E Loro Associazioni [PAMELA] Study) Circulation 2001 104 1385 1392 11560854
Selenta C Hogan BE Linden W How Often Do Office Blood Pressure Measurements Fail to Identify True Hypertension? An Exploration of White-Coat Normotension Arch Fam Med 2000 9 533 540 10862216 10.1001/archfami.9.6.533
Prisant LM Ambulatory blood pressure monitoring in the diagnosis of hypertension Cardiol Clin 1995 13 479 490 8565011
O'Brien E Mee F Atkins N O'Malley K Accuracy of the Spacelabs 90207 determined by the British Hypertension Society Protocol J Hypertens 1991 5 573 574 1653299
Liu JE Roman MJ Pini R Schwartz JE Pickering TG Devereux RB Elevated ambulatory with normal clinic blood pressure ("white coat normotension") is associated with cardiac and arterial target organ damage Ann Intern Med 1999 131 564 572 10523216
Pickering TG James GD Boddie C Harsfield GA Blank S Laragh JH How common is white-coat hypertension? JAMA 1988 259 225 228 3336140 10.1001/jama.259.2.225
Staessen JA Thijs L Fagard R for the Systolic Hypertension in Europe Trial Investigators Predicting cardiovascular risk using conventional vs ambulatory blood pressure in older patients with systolic hypertension JAMA 1999 282 539 546 10450715 10.1001/jama.282.6.539
White WB Schulman P McCabe EJ Dey HM Average daily blood pressure, not office pressure, determines cardiac function in patients with hypertension JAMA 1989 261 873 877 2521522 10.1001/jama.261.6.873
Mancia G Zanchetti A White-coat hypertension: misnomers, misconceptions and misunderstandings. What should we do next? J Hypertens 1996 14 1049 1052 8986902
Verdecchia P Clement D Fagard R Palatini P Parati G Target organ damage, morbidity and mortality Blood Pressure Monitoring 1999 4 303 317 10602535
Pickering TG Coats A Mallion JM Mancia G Verdecchia P Task Force V White-coat hypertension Blood Pressure Monitoring 1999 4 333 341 10602537
William BW Ambulatory blood-pressure monitoring in clinical practice New Engl J Med 2003 348 24 2377 2378 12802022 10.1056/NEJMp030057
Little P Barnet J Barnsley L Marjoram J Fitzgerald-Baron A Mant D Comparison of agreement between different measures of blood pressure in primary care and daytime ambulatory blood pressure BMJ 2002 325 254 259 12153923 10.1136/bmj.325.7358.254
Bellomo G Narducci PL Rondoni F Pastorelli G Stagnoni G Prognostic value of 24-hour blood pressure in pregnancy JAMA 1999 282 1447 1452 10535435 10.1001/jama.282.15.1447
Mancia G Bertineri G Grassi G Effects of blood pressure measured by the doctor on patient's blood pressure and heart rate Lancet 1983 2 695 698 6136837 10.1016/S0140-6736(83)92244-4
Myers MG Haynes B Rabbkin SW Canadian Hypertension Society Guidelines for Ambulatory Blood Pressure Monitoring Am J Hypertens 1999 12 1149 1157 10604495 10.1016/S0895-7061(99)00199-5
Myers MG Reeves RA White-coat phenomenon in patients receiving antihypertensive therapy Am J Hypertens 1991 4 844 849 1747219
Myers MG Reeves RA White-coat effect in treated hypertensive patients: sex difference J Hum Hypertens 1995 9 729 733 8551486
O'Brien E Ambulatory blood pressure monitoring in the management of hypertension Heart 2003 89 571 576 12695477 10.1136/heart.89.5.571
Brawn LA Castleden CM Adverse drug reactions: an overview of special considerations in the management of the elderly patient Drug Safety 1990 5 421 435 2285496
Rochon PA Anderson GM Tu JV Age- and gender-related use of low-dose drug therapy: the need to manufacture low-dose therapy and evaluate the minimum effective dose J Am Geriatr Soc 1999 47 954 959 10443856
Tomlinson B Optimal dosage of ACE inhibitors in older patients Drugs Aging 1996 9 262 273 8894524
Joint National Committee on Detection, Evaluation, and Treatment of High Blood Pressure The Sixth Report of the Joint National Committee on detection, evaluation, and treatment of high blood pressure Arch Intern Med 1997 157 2413 1446 9385294 10.1001/archinte.157.21.2413
Guidelines Subcommittee 1999 WHI/ISH Guidelines for the management of hypertension J Hypertens 1999 17 151 183 10067786
Ramsay LE Williams B Johnston GD MacGregor GA Poston L Potter JF Poulter NR Russel G British Hypertension Society Guidelines for hypertension management 1999: summary Br Med J 1999 519 17630 635
Centers for Medicare & Medicaid Services Medicare coverage policy – decisions ABPM monitoring (#CAG-00067N) 2001
Krishan G Isolated Office Hypertension: Ambulatory BP Monitoring Is the Only Answer Arch Intern Med 2002 162 13 1526 10.1001/archinte.162.13.1526
Staessen JA O'Brien E Thijs L Fagard RH Modern approaches to blood pressure measurement Occup Environ Med 2000 57 510 520 10896957 10.1136/oem.57.8.510
Ferrucci L Furberg CD Penninx BW Treatment of isolated systolic hypertension is most effective in older patients with high-risk profile Circulation 2001 104 1923 1926 11602495
Pocock SJ McCormack V Gueyffier F Boutitie F Fagard RH Boissel JP A score for predicting risk of death from cardiovascular disease in adults with raised blood pressure, based on individual patient data from randomized controlled trials BMJ 2001 323 75 81 11451781 10.1136/bmj.323.7304.75
O'Rourke J Richardson SW Evidence based management of hypertension: what to do when blood pressure is difficult to control BMJ 2001 322 1229 1232 11358779 10.1136/bmj.322.7296.1229
Pickering TG Ambulatory blood pressure monitoring Curr Hypertens Rep 2000 2 558 564 11062602
Prasad N Isles C Ambulatory blood pressure monitoring: a guide for general practitioners BMJ 1996 313 1535 1541 8978234
O'Brien E Coats A Owens P Petries J Padfield P Littler W Use and interpretation of ambulatory blood pressure monitoring: recommendations of the British Hypertension Society BMJ 2000 320 1128 1134 10775227 10.1136/bmj.320.7242.1128
Ramsey L Williams B Johnson G MacGregor GA Poston L Potter J Guidelines for management of hypertension: report of the third working party of the British Hypertension Society J Hum Hypertens 1999 13 569 592 10482967 10.1038/sj.jhh.1000917
Devereux RB Pickering TG Relationship between the level, pattern and variability of ambulatory blood pressure and target organ damage in hypertension J Hypertens Suppl 1991 9 suppl 8 S34 S38 1839035
Giaconi S Levanti C Fommei E Microalbuminuria and casual and ambulatory blood pressure monitoring in normotensives and in patients with borderline and mild essential hypertension Am J Hypertens 1989 2 259 261 2706093
Shimada K Kawanato A Matsubayishi K Ozawa T Silent cerebrovascular disease in the elderly: correlation with ambulatory pressure Hypertension 1990 16 692 699 2246035
Mancia G Zanchetti A Agebiti-Rosei E Ambulatory blood pressure is superior to clinic blood pressure in predicting treatment-induced regression of left-ventricular hypertrophy Circulation 1997 95 1464 1470 9118514
Ohkubo T Imai Y Tsuji I Prediction of mortality by ambulatory blood pressure monitoring versus screening blood pressure measurements: a pilot study in Ohasama J Hypertens 1997 15 357 364 9211170 10.1097/00004872-199715040-00006
Staessen JA Thijs L Fagard R Predicting cardiovascular risk using conventional vs ambulatory blood pressure in older patients with systolic hypertension JAMA 1999 282 539 546 10450715 10.1001/jama.282.6.539
Perloff D Sokolow M Cowan R The prognostic value of ambulatory blood pressures JAMA 1983 249 2792 2798 6842787 10.1001/jama.249.20.2792
Redon J Campos C Narciso ML Rodicio JL Pascual JM Ruilope LM Prognostic value of ambulatory blood pressure monitoring in refractory hypertension: a prospective study Hypertension 1998 31 712 718 9461245
Verdecchia P Porcellati C Schillaci G Ambulatory blood pressure, an independent predictor of prognosis in essential hypertension Hypertension 1994 24 793 801 7995639
Ernst EE Bergus GR Favorable patient acceptance of ambulatory blood pressure monitoring in a primary care setting in the United States: a cross-sectional survey BMC Family Practice 2003 4 15 14533981 10.1186/1471-2296-4-15
| 15813971 | PMC1087497 | CC BY | 2021-01-04 16:47:32 | no | Curr Control Trials Cardiovasc Med. 2005 Apr 6; 6(1):5 | utf-8 | Curr Control Trials Cardiovasc Med | 2,005 | 10.1186/1468-6708-6-5 | oa_comm |
==== Front
Genet Vaccines TherGenetic Vaccines and Therapy1479-0556BioMed Central London 1479-0556-3-21583678110.1186/1479-0556-3-2MethodologyElectroporation by nucleofector is the best nonviral transfection technique in human endothelial and smooth muscle cells Iversen Nina [email protected] Baard [email protected] Kari [email protected] Srdjan [email protected] Department of Medical Genetics, Ullevål University Hospital, Oslo, Norway2005 18 4 2005 3 2 2 7 12 2004 18 4 2005 Copyright © 2005 Iversen et al; licensee BioMed Central Ltd.2005Iversen 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 this study was to determine the optimal non-viral transfection method for use in human smooth muscle cells (SMC) and endothelial cells (EC).
Methods
Coronary Artery (CoA) and Aortic (Ao) SMC and EC were transfected with a reporter plasmid, encoding chloramphenicol acetyltransferase type 1 (CAT), with seven different transfection reagents, two electroporation methods and a photochemical internalization (PCI) method. CAT determination provided information regarding transfection efficiency and total protein measurement was used to reflect the toxicity of each method.
Results
Electroporation via the nucleofector machine was the most effective method tested. It exhibited a 10 to 20 fold (for SMC and EC, respectively) increase in transfection efficiency in comparison to the lipofection method combined with acceptable toxicity. FuGene 6 and Lipofectamine PLUS were the preferred transfection reagents tested and resulted in 2 to 60 fold higher transfection efficiency in comparison to the PCI which was the least effective method.
Conclusion
This study indicates that electroporation via the nucleofector machine is the preferred non-viral method for in vitro transfection of both human aortic and coronary artery SMC and EC. It may be very useful in gene expression studies in the field of vascular biology. Through improved gene transfer, non-viral transfer techniques may also play an increasingly important role in delivering genes to SMC and EC in relevant disease states.
ElectroporationGene TherapyLiposomesLipofectionPhotochemical InternalizationNucleofectionTransfection
==== Body
Background
Several methods have been described to introduce DNA expression vectors into mammalian cells in vitro and in vivo: calcium phosphate precipitation, microinjection, electroporation, receptor-mediated gene transfer, particle guns, viral vectors, and lipofection [1-3]. Each system has benefits and limitations, and to date there is no ideal method for gene transfer.
Viral vector systems, derived from modified animal or human viruses, resulting in replication-deficient vectors [4], represent a powerful transfection tool. Nevertheless, their immunogenicity, oncogenic properties, inactivation of vector, development of replication-competent virions and need for a relatively large-scale infrastructure for their production are serious disadvantages [5].
The use of cationic liposome/DNA complexes (lipofection) for gene transfer into somatic cells has become a popular method of delivering genes. Interaction between cationic lipids and DNA through ionic interaction leads to forming cationic lipoplexes [1,4]. The resulting complexes fuse with the anionic surfaces of cells, delivering DNA into the cells via endocytosis. However, the final transport of DNA into the nucleus is still not fully understood. Although inferior, transfection using lipofection offers some advantages over viral vectors, such as simplicity of production, low toxicity and low immunogenicity.
Another transfection method, electroporation [6], also termed electrotransfer [7] or electropermeabilization [8], is an experimental technique involving the application of brief electric pulses to cells or tissues in order to increase cellular permeability to macromolecules. This method has been reported to increase naked DNA expression by 100-fold or more [6-8]. Finding the balance between the best possible transfection efficiency and survival rate is very important, therefore we investigated the optimization of this technique using two different electroporation instruments.
Photochemical internalization (PCI) was reported as a procedure for site-specific delivery of several types of membrane impermeable macromolecules from endocytotic vesicles to the cytosol [9]. This technology is based on the cytosolic release of endocytosed macromolecules from endosomes and lysosomes which become localized to these vesicles upon exposure of cells to photosensitizing compounds and light. PCI has several advantages over other conventional applications for the cytosol delivery of membrane impermeable molecules [10]. One advantage is that there are no restrictions on the type and size of the molecule to be internalized, as long as the molecule of interest can be endocytosed. We examined the applicability of PCI technology to our cells of interest.
In this study we present extensive investigations performed with transfection reagent mediated transfections, electroporation and PCI. The aims of the study were to evaluate the efficiency and safety of optimized novel non-viral transfection techniques for our four cell types of interest: coronary artery (CoA) SMC, aortic (Ao) SMC, CoAEC and AoEC. Our results showed that electroporation via the nucleofector machine turned out to be the most effective non-viral method for in vitro transfection of both human SMC and EC, while FuGene6 and Lipofectamine PLUS appeared as best performing lipofection reagents. These results also provided useful informations regarding optimization and selection of transfection conditions for the cell types tested.
Methods
Cell cultures
Human Coronary Artery (#CC-2583) and Aortic (#CC-2571) SMC were obtained from Clonetics Corporation (Walkersville, MD) together with human Coronary Artery (#CC-2585) and Aortic [#CC-2535] EC. The cells had been isolated from normal human tissue and cryopreserved in smooth muscle cell media, SmGM-2 (#CC-3182) and endothelial cell media, EGM-2-MV (#CC-3202) respectively, supplemented with 10% FCS (Gibco BRL, Gathersburg, MD) and 10% dimethyl sulfoxide in order to improve cell viability and seeding efficiency upon thawing. Cells were cultivated in modified Sm basal medium (SmBM; #CC-3181) supplemented with SmGM-2 Single Quots and growth factors (#CC-4149) or, for EC in EBM-2 basal medium (#CC-3156) supplemented with EGM-2-MV Single Quots and growth factors (#CC-4147) (Clonetics Corporation, Walkersville, MD) and 5% FCS. Cells were incubated at 37°C in a humidified atmosphere with 95% air and 5% CO2. Medium was changed every second day and the protocols from producer were strictly followed. For the transfection experiments, low-passage cells (passages 4 to 8) at 80% confluency were used.
Plasmid vectors
The bacterial enzyme, CAT, encoded by Tn9, has no eukaryotic equivalent and has become one of the standard markers used in transfection experiments.
The pRc/CMV2/CAT plasmid supplied by Invitrogen (Carlsbad, CA, USA) was used in this study. We amplified the plasmid using competent E. coli cells from One Shot chemical transformation kits supplied by Invitrogen (Carlsbad, CA, USA). Bacteria were grown and the plasmid was isolated using GigaPrep kit, QIAGEN (Valencia, CA, USA).
Transfection reagents
Seven commercially available transfection reagents were used:
• FuGENE 6 (Roche, Mannheim, Germany), a non-liposomal transfection reagent, proprietary blend of lipids and other compounds,
• Lipofectamine PLUS (Invitrogen, Carlsbad, CA, USA), a 3:1 liposome formulation of the polycationic lipid 2,3-dioleyloxy-N(2(sperminecarboxamido)ethyl)-N,N-dimethyl-1-propanaminium trifluoroacetate (DOSPA) and the neutral lipid dioleoyl phosphatidylethanolamine (DOPE) in membrane-filtered water. PLUS reagent is used to pre-complex DNA prior to the preparation of the transfection complexes,
• Metafectene (Biontex, Munich, Germany), a polycationic transfection reagent that encompasses "repulsive membrane acidolysis" which ensures destabilization of the DNA-coating lipid membrane by repulsive electrostatic forces in the weakly endosomal acidic environment and release of the DNA into the cell protoplasm,
• Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA), a cationic lipid that allows high transfection efficiencies and protein expression levels,
• GenePORTER (Gene Therapy Systems Inc., San Diego, CA, USA), a formulation of the neutral lipid DOPE and a proprietary cationic lipid derived from hydrophilic conjugation technology,
• LipoGen (InvivoGen, San Diego, CA, USA), a formulation of a unique lipid that combines in its structure the characteristics of both a cationic lipid and a fusogenic lipid, such as DOPE, which works via the unsaturated hydrocarbon chains of DOPE which destabilize membrane bilayers, thereby facilitating delivery of lipid/DNA complexes into the cells, and
• Lipofectin (Invitrogen, Carlsbad, CA, USA) a 1:1 liposome1 liposome formulation of cationic lipid N-(1-(2,3-dioleyloxy)propyl) -n,n,n-trimethylammonium chloride (DOTMA) and DOPE in membrane filtered water.
Transfection by reagents
Low-passage cells were cultivated and used in 6-well plates 18 h before transfection. Approximately 3 × 105 cells per well (80% confluence) were used in transfections.
The transfections, using reporter vector complexed with each of the tested reagents, were performed according to the manufacturer's protocols.
Plasmid DNA (0.8–6 μg CAT) at different DNA:liposome ratios (1:3 – 1:5) was diluted in separate tubes containing 100 μl – 1000 μl of serum-free media, mixed and incubated 15–45 min at room temperature. Media was removed and transfection solutions were added to each well (100 μl – 1000 μl). After 3 – 6 hrs incubation at 37°C and 5% CO2, 1 ml fresh media (with FCS and supplements) was added to each well and transfection continued for 24 hours.
Transfection by electroporation
Two different methods of electroporation were tested, each using a different instrument. Firstly, electroporation was conducted with ECM 630 electroporator (BTX, San Diego, CA, USA) and secondly, the nucleofector instrument, (Amaxa Biosystems, Cologne, Germany) was tested.
Cells were grown in T175 bottles, trypsinized, collected by centrifugation (200 × g, 10 min) and resuspended in medium containing 10% FCS for EC and Hanks solution for SMC. 0.4 ml containing approximately 2 × 106 cells and 20 μg CAT plasmid (1 μg/μl) was placed in a sterile electroporation cuvette (BTX 0,2 cm gap). Cells were subjected to high-voltage at a setting that had been optimized for each cell type. After electroporation, the cells were immediately plated out using pre-warmed growth media supplemented with 10% FCS in 6 well plates.
For transfection with the Nuclefector instrument, a specific optimized electroporation method and a specific nucleofector solution were used for each cell type. For SMC the human AoSMC Nucleofector™ kit was used (VPC-1001). Cells were grown in T175 bottles, trypsinized, collected by centrifugation (200 × g, 10 minutes) and resuspended in the HCAEC nucleofector solution at two cell suspensions of 5 × 105 and 1 × 106 cells per 100 μl and 1–10 μg DNA (1 μg/μl CAT). Program U-25 was applied. For CoAEC the human HCAEC Nucleofector™ kit (VPB-1001) was used. CoAEC were treated as SMC, except that they were tested at a single concentration of 5 × 105 cells per 100 μl. 100 μl of cell suspension and 1–10 μg DNA (1 μg/μl CAT) were mixed and transferred to a cuvette. Program S-05 was used. After treatment, the cells were immediately plated out in pre-warmed medium, supplemented with 10% FCS, into 6 well plates.
Transfection by photochemical internalization (PCI)
Photochemical internalization was conducted with a LumiSource™ (PCI Biotech AS, Oslo, Norway). Reagents (LumiTrans and p(Lys)) were also provided from PCI Biotech.
For this method 7 × 104 were cells plated into 12-well culture plates. The next day media was removed and the cells were treated with 0.4 ml of the photosensitizer LumiTrans in medium containing 10% FCS (2 μg/ml) for 16–18 hours at 37°C. The cells were washed three times with medium. For Optimization of light dose, 0.8 ml fresh medium was added to cells before exposure to the LumiSource for 20 to 200 sec. Cell lysates were harvested after 24 hours and total protein measurement was carried out. The light dose that gave 50% survival was set as the highest dose and a range of lower light doses was used for optimization of the PCI method.
Photochemical transfection
Plasmid-p(Lys) complexes were formed by gentle mixing of 75 μl cell suspension with 2–20 μg CAT plasmid (1 μg/μl), water with 5.35 μl of p(Lys) (1 μg/μl) and 69.65 μl of water. The resulting solution was incubated for 30 minutes at room temperature before being diluted to 1 ml with medium. Cells were incubated with 0.4 ml of the plasmid mixture for 4 hours at 37°C. When the cells were washed once with medium, fresh medium (0.8 ml) was added and the cells were exposed to LumiSource light doses. The cells were exposed to increasing light doses before the transfection.
Post-transfection cell treatment
24 hrs after transfection, media was removed and cells were washed 3 times with 1 × PBS and lysed in 1 or 2 ml CAT lysis buffer (supplied in CAT ELISA kit, Roche, Mannheim). Cell lysates were used for CAT determination and total protein measurement assay.
CAT ELISA Measurements
Concentrations of CAT in cell lysate were measured by CAT-ELISA (Roche, Mannheim, Germany) as recommended by the producer. All measurements were done in duplicate and concentrations of unknowns were determined from standards run with each plate.
Cell Survival calculations
Cellular total protein was measured by an improved Lowry assay (Bio-Rad DC Protein Assay, Bio-Rad Laboratories, Hercules, CA, USA). When comparing the results from test and control wells, it was assumed that cells in the control well were unaffected by the experiment. Test results were then compared to the control results and a percentage survival was calculated.
These measurements were confirmed using a Cytotoxicity Detection Kit (Roche, Mannheim), which measures lactate dehydrogenase (LDH) activity released from damaged cells (results not shown).
Reporting of results
In order to effectively compare the results from each of the three methods, we standardized the results according to the number of cells used : transfection reagents requiring only 3 × 105 cells while electroporation and PCI use 1–2 × 106 cells per well. To standardize, we used a ratio of CAT produced (ng) divided by total protein of surviving cells (ng), thereafter called transfection efficiency (the amount of CAT produced per living cell). This value was then multiplied by 1 × 106 to make the numbers more manageable. This calculation does not take into account the differences in cell survival, and that is why this should be considered as well. for the comparisons of transfection efficiency
Results
Transfection by reagents
In order to determine the preferred transfection reagent for each cell type, we comparatively considered the following: the amount of CAT produced, the ratio between CAT/total protein and the cell survival. When considering the results obtained in the four cell types used, the results show that the three best performing reagents were FuGENE 6, Lipofectamine PLUS and GenePORTER (Table 1 and Figure 1). As presented in Table 1, the values display a range across the four cell types used. Individual results are reported in the text below and in Figure 1, where the results found using the optimal concentration of plasmid for each reagent are displayed.
Table 1 Summary of the results obtained from cells transfected with chloramphenicol acetyl transferase using the seven different transfection reagents tested. Results are given as a range across all cell types.
Liposome Manufacturer DNA amount Liposome: DNA ratio Transfection Efficiency % Cell Survival
FuGENE 6 Roche 1 and 2 μg 3:1 3.0 – 16.4 65 – 80
Lipofectamine 2000 Invitrogen 2 and 4 μg 3:1 0.4 – 34.0 9 – 27
Lipofectamine PLUS Invitrogen 0.8 and 1.6 μg 3:1 3.4 – 18.3 18 – 61
Metafectene Biontex 2 and 4 μg 3:1 0.6 – 8.6 25 – 50
Lipofectin Invitrogen 1 and 2 μg 4:1 0.0 – 7.1 65 – 100
GenePORTER Gene Therapy Solutions 3 and 6 μg 5:1 1.6 – 21.9 24 – 55
LipoGen Invivo 1 and 2 μg 3:1 0.0 – 13.9 10 – 90
Figure 1 Figure (a) shows the amount of chloramphenicol acetyltransferase (CAT) produced in each of the cell lines, when the different transfection reagents were used, at optimal plasmid amount. Figure (b) shows the corresponding % survival when each of these reagents and plasmid amounts were used. Note: Results are shown as a mean +/- SD of two individual experiments (performed in duplicate).
In CoASMC, use of FuGENE 6 achieved the best results. It produced almost twice as much CAT per ml media than cells transfected using the second best performing reagent, Lipofectamine PLUS (Figure 1). When 1 μg and 2 μg plasmid were used, ratios of 3–5 were obtained and the cell survival rate was between 69 and 74%, respectively (Figure 1). In AoSMC, Lipofectamine PLUS gave the best results. It produced more CAT per ml media than cells transfected with the next best performing reagent, FuGENE 6 (Figure 1). When 0.8 μg and 1.6 μg plasmid was used, ratios of 10–18 were obtained and the cell survival rate was between 53 and 58%, respectively (Figure 1).
In CoAEC, best transfection efficiency was achieved by FuGENE 6 : it produced more than double the amount of CAT per ml than cells transfected using the other reagents (Figure 1). When 1 μg and 2 μg of plasmid were used, ratios of 8–11 were obtained and the cell survival rate was between 64 and 73%, respectively (Figure 1).
FuGENE 6 gave the best results in AoEC, as well : it produced more CAT per ml media than cells transfected with the second best liposome, Lipofectamine PLUS (Figure 1). When 1 μg and 2 μg of plasmid was used, ratios of 7–16 were obtained and the cell survival rate was between 79 and 88%, respectively (Figure 1).
Transfection by electroporation
Electroporator
To optimize the electroporation procedure, a range of voltage, capacitance and resistance settings were used. For SMC the initial resistance and capacitance settings were 725Ω and 125 μF and for EC they were 950Ω and 25 μF. The voltage settings tested varied from 400 – 500 V. The optimal voltage in all four cell types was 450 V, illustrated by AoSMC (Figure 2).
Figure 2 Figure (a) shows the transfection efficiency obtained in AoSMC when voltage settings were varied using the ECM 630 electroporator. Capacitance and Resistance were held constant at 125 μF and 725Ω respectively. Figure (b) shows the % survival obtained at the corresponding settings. Results represent mean of triplicates +/- SD of a typical experiment.
After the voltage settings had been established the optimal resistance and capacitance were found. For CoA and Ao SMC the best resistance setting was found to be in the area 725–900Ω (Figure 3), but best capacitance varied between the two cell types. In CoASMC, the best capacitance setting was 75 μF (Ratio 2.5 and 70% survival). In some experiments, we achieved a ratio of up to 6 when 125 μF was used, but survival dropped to around 30%. Nevertheless, we choose 75 μF as the best setting because it resulted in higher cell survival. In AoSMC the best results were obtained when 125 μF were used (Ratio of 0.92 and a survival of 30%) (Figure 4). The higher the capacitance settings was, the lower become the cell survival (Figure 4b).
Figure 3 Figure (a) shows the transfection efficiency obtained in AoSMC when resistance settings were varied using the ECM 630 electroporator. Voltage and capacitance were held constant at 450 V and 125 μF respectively. Figure (b) shows the % survival obtained at the corresponding settings. Results represent mean of triplicates +/- SD of a typical experiment.
Figure 4 Figure (a) shows the transfection efficiency obtained in AoSMC when different capacitance settings were used on the ECM 630 electroporator (BTX, San Diego, USA). Voltage and resistance were held constant at 450 V and 800Ω, respectively. Figure (b) shows the % survival obtained at the corresponding settings. Results represent mean of triplicates +/- SD of a typical experiment.
Both CoA and Ao EC reacted similarly to the different settings. Resistance was tested between 850–1050Ω and at 900Ω a ratio of 25 was obtained (55% survival). We tested capacitance varying from 25 – 75 μF. When 50 μF was used, we obtained a ratio of 40 and a survival of 38%. However, 25 μF was the best setting since it resulted in better cell survival (61%) (results not shown).
Nucleofector
Optimized nucleofector protocols were available for AoSMC and CoAEC. These methods were tested and the results were compared with the electroporation results. For AoSMC we tested two cell suspensions, 5 × 105 and 1 × 106cells per reaction. Both the ratio and the survival increased by increasing the number of cells used (Figure 5a). At the highest plasmid dose, the cell survival was 80% (Figure 5b).
Figure 5 Figure (a) shows the transfection efficiency obtained in AoSMC when different amounts of CAT plasmid were transfected into different cell numbers using the Nucleofector instrument, program U-25. Figure (b) shows the % survival obtained at the corresponding plasmid amounts. Results represent mean of duplicates +/- SD.
In CoAEC, we observed a dose-response for the CAT/protein ratio when 1–10 μg plasmid was used (Figure 6a), and at the highest plasmid dose of 10 μg, 30–46 % cell survival was achieved (Figure 6b).
Figure 6 (a) and (b): Figure (a) shows the transfection efficiency obtained in CoAEC when different amounts of CAT plasmid were transfected, using the Nucleofector instrument (Amaxa Biosystems, Cologne, Germany), program S-05. Figure (b) shows the % survival at the corresponding plasmid amounts. Results represent mean of triplicates +/- SD of a typical experiment.
Transfection by PCI
The initial experiments with PCI were aimed to find the light dose at which we obtained at least 50 % survival. For AoSMC this was observed to be 100 sec. In further experiments light doses varying from 25 to 100 seconds were used. A low transfection effect, ratio of 0.3, was achieved when the cells were exposed to light before the transfection of 5 μg plasmid (Table 2).
Table 2 The best results obtained in AoSMC and CoAEC when different transfection methods were used.
Methods Best Transfection Efficiency Corresponding % Cell Survival
SMC EC SMC EC
Transfection Reagent 18 11 53 74
ECM 630 5.5 25 38 55
Nucleofector 200 209 80 30
PCI 0.3 4.7 84 55
The light dose that gave 50% survival in CoAEC was between 40 and 50 seconds, and for AoEC it was 32 seconds. The best transfection effect obtained had a ratio of 4.7 and 55% survival, when the cells were given 5 μg plasmid before exposure to light for 25 seconds (Table 2). None effect was seen when the cells were exposed to light after addition of plasmid.
Discussion
Improvement of the delivery efficiency of genes into SMC and EC and the development and optimization of transfection methods has increasingly become an important research objective. In this study we found that transfection by electroporation, using the nucleofector instrument, was comparatively the most effective transfection method combining both high efficiency and acceptable survival rate for both smooth muscles cells and endothelial cells (Table 2). Enhancement of transfection efficiency by transfection reagents and the ECM 630 instrument also worked well, but not to the same extent as nucleofection (Table 2).
Transfection using the nucleofector is a patented commercial technique requiring special buffers and programs, the constituents of which are a secret. Nevertheless, we developed "in house" methods for ECM 630 electroporator machine. Optimizing these methods is possible, but many variables have to be taken into account. In this study we used constant buffer, cell numbers and plasmid amounts in order to test and optimize the variables available on the instrument (voltage, capacitance and resistance). From our findings we can conclude that transfer efficiencies could be greatly improved. We believe that electroporation by nucleofection is an easy and effective method for transfecting human EC and SMC, although the high number of cells and high plasmid amounts required could be considered a weakness.
On the other hand, the use of transfection reagents in in vitro cell transfection is easy, affordable and requires low cell numbers. Enhancement of transfer efficiency was achieved by reagents in all the cell types tested. However, the best performing transfection reagents found in certain cell types are not necessarily the optimal reagents for other cell types. Nevertheless, FuGENE 6 reagent has shown similar transfection efficiency across all four cell types. Of further importance is to notice that results in this study reflect those setups recommended by each reagent manufacturer. As expected, optimization of transfection condition by these reagents may also lead to an improvement in transfection efficiencies. For optimal transfer efficiencies, higher DNA concentrations require higher amounts of liposomes, which will inevitably increase cell death. FuGENE 6 achieved relatively good transfer efficiencies combined with low toxicities.
Another transfection reagent, GenePORTER, demonstrated high transfer efficiencies, although accompanied with relatively high toxicity. This may be attributed to GenePORTER requirements for high amounts of DNA and liposome to be successful.
Generally, the observed differences in the transfer efficiency and the optimal DNA/liposome ratios may depend on the readiness of cells to take up DNA/liposome complexes [11-14]. Our findings suggest that the ratios could not be generalized, and they have to be specified for each cell type and liposome used. The negative polarity of the cell surface appears to play a key role in the process. Therefore, transfection using reagents/liposomes needs to be optimized for each targeted cell type and reagent used [15].
PCI was yet another method we tested, but considering comparatively lower transfer efficiency, the time-consuming and demanding procedure, this method could not be evaluated as the most suitable one. It should be noted though, that PCI has shown promising applications in cancer therapy [9].
Several studies have reported optimization of non-viral gene transfer techniques to individual cells [16-18]. Up to our knowledge, this is the first report that compared lipofection, electroporation and PCI at the same experimental settings on human Ao and CoA SMC and EC.
Electroporation and PCI may prove difficult to use in a clinical setting. Regarding electroporation, tissues would have to be electroporated by using methods that are yet not well established. It has been reported [19,20] that in vivo electroporation has been implicated as the major cause of muscle damage in studies with electrical trauma. Moreover, clinical use of this method resulted in transient to permanent alterations in membrane permeability [19,20]. Furthermore, the extent of muscle damage may depend on pulsing parameters and electrode design [8,21-23]. Hartikka et al. [24] reported extensive lesions containing necrotic myofibers and heavily populated with infiltrating inflammatory cells.
Through improved gene transfer, non-viral therapeutic techniques may play an increasingly important role in delivering genes to cells in relevant disease states. One example is the cardiovascular field [25-27], and this is the reason why we tested and optimized these techniques using aortic and coronary artery SMC and EC. A strong theoretical advantage of cardiovascular gene therapy is the ease of access and, in some conditions only a temporary expression of the transfected gene is needed to achieve a beneficial biological effect [28]. Therefore, non-viral transfection techniques might offer a therapeutic option and prove suitable for the treatment of cardiovascular disease, and informations regarding optimization of transfection conditions in order to improve transfection efficiency and reduce cytotoxicity, such from our study, may be valuable for use in gene therapy studies.
Conclusion
From the results achieved in this study it is evident that electroporation by the nucleofector instrument is the preferred transfection method for all four cardiovascular cell types. Nucleofection technique exhibited a high transfection efficiency and acceptable cell survival rate and should be very useful in gene expression studies in cardiovascular biology. As a step toward further development of gene therapy strategy, extensive in vitro studies with novel techniques presented within this work are essential in definition of the most suitable transfer methods.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
N.I. participated in the design and coordination of the study, established the transfection methods electroporation, nucleofection and PCI, and supervised the following optimalization of these methods. She was responsible for writing the manuscript. B.B. and K.T. were responsible for performing experiments independently, and interpretation of the results, while SD concieved the study, and participated in the design and coordination of it, and supervised liposomal transfection experiments.
All authors critically read and approved the final version of the manuscript.
Acknowledgements
This work was funded by the grant from the Norwegian Board of Health (st.prp. nr. 61). We thank Jamie Cameron and Therese Lundin for skilful laboratory assistance.
==== Refs
Felgner PL Clifton, NJ Cationic liposome-mediated transfection with lipofectin reagent. E.J. Murray Editor, Gene transfer and expression protocols, Methods of molecular biology 1991 7 Humana Press 81 89
Blaese RM Mullen CA Ramsey WJ Strategies for gene therapy Pathol Biol 1993 41 672 676 8290310
Eastman SJ Siegel C Tousignant J Smith AE Cheng SH Scheule RK Biophysical characterization of cationic lipid: DNA complexes Biochim Biophys Acta 1997 1325 41 62 9106482
Guenzburg WH Salmons B Virus vector design in gene therapy Mol Med Today 1997 1 410 417
Yang Y Jooss KU Su Q Ertl HC Wilson JM Immune responses to viral antigens versus transgene product in the elimination of recombinant adenovirus-infected hepatocytes in vivo Gene Ther 1996 3 137 144 8867861
Aihara H Miyazaki J Gene transfer into muscle by electroporation in vivo Nat Biotechnol 1998 16 867 870 9743122
Mir LM High-efficiency gene transfer into skeletal muscle mediated by electric pulses Proc Natl Acad Sci USA 1999 96 4262 4267 10200250
Mathiesen I Electropermeabilization of skeletal muscle enhances gene transfer in vivo Gene Ther 1999 6 508 514 10476210
Berg K Selbo PK Prasmickaite L Tjelle TE Sandvig K Moan J Gaudernack G Fodstad O Kjølsrud S Anholt H Rodal GH Rodal SK Høgset A Photochemical internalization: a novel technology for delivery of macromolecules into cytosol Cancer Res 1999 59 1180 1183 10096543
Selbo PK Hogset A Prasmickaite L Berg K Photochemical internalization: a novel drug delivery system Tumour Biol 2002 23 103 12 12065848
Escriou V Ciolina C Helbling-Leclerc A Wils P Scherman D Cationic lipid-mediated gene transfer: analysis of cellular uptake and nuclear import of plasmid DNA Cell Biol Toxicol 1998 14 95 104 9553720
Pelisek J Engelmann MG Golda A Fuchs A Armeanu S Shimizu M Mekkaoui C Rolland PH Nikol S Optimization of nonviral transfection: variables influencing liposome-mediated gene transfer in proliferating vs. quiescent cells in culture and in vivo using a porcine restenosis model J Mol Med 2002 80 724 36 12436349
Brisson M Huang L Liposomes: conquering the nuclear barrier Curr Opin Mol Ther 1999 1 140 146 11715938
Simoes S Pires P Duzgunes N Pedrosa de Lima MC Cationic liposomes as gene transfer vectors: barrier to successful application in gene therapy Curr Opin Mol Ther 1999 1 147 157 11715939
Armeanu S Pelisek J Krausz E Fuchs A Groth D Curth R Keil O Quilici J Rolland PH Reszka R Nikol S Optimization of liposomal gene transfer of vascular smooth muscle cells in vitro and in vivo Mol Ther 2000 1 366 375 10933955
Uchida E Mizuguchi H Ishii-Watabe A Hayakawa T Comparison of the efficiency and safety of non-viral vector-mediatet gene transfer into a wide range of human cells Biol Pharm Bull 2002 25 25891 897
Schakowski F Buttgereit P Mazur M Marten A Schottker B Gorschluter M Schmidt-Wolf IG Novel non-viral method for transfection of primary leukemia cells and cell lines Genet Vaccines Ther 2004 12 1 11 14715084
Kiefer K Clement J Garidel P Peschka-Suss R Transfection efficiency and cytotoxicity of nonviral gene transfer reagents in human smooth muscle and endothelial cells Pharm Res 2004 21 1009 17 15212167
Block TA The 1995 Lindberg Award. Nonthermally mediated muscle injury and necrosis in electrical trauma J Burn Care Rehabil 1995 16 581 588 8582934
Lee RC Canaday DJ Hammer SM Transient and stable ionic per-meabilization of isolated skeletal muscle cells after electrical shock J Burn Care Rehabil 1993 14 528 540 8245107
Bureau MF Gehl J Deleuze V Mir LM Scherman D Importance of association between permeabilization and electrophoretic forces for intramuscular DNA electrotransfer Biochim Biophys Acta 2000 1474 353 359 10779687
Vicat JM Muscle transfection by electroporation with high-voltage and short- pulse currents provides high-level and long-lasting gene expression Hum Gene Ther 2000 11 909 916 10779167
Ghel J In vivo electroporation of skeletal muscle: threshold, efficacy and relation to electric field distribution Biochim Biophys Acta 1999 1428 233 240 10434041
Hartikka J Sukhu L Buchner C Hazard D Bozoukova V Margalith M Nishioka WK Wheeler CJ Manthorp M Sawdey M Electroporation-facilitated delivery of plasmid DNA in skeletal muscle: plasmid dependence of muscle damage and effect of poloxamer 188 Mol Ther 2001 4 407 15 11708877
Nikol S Pelisek J Engelmann MG Rolland PH Armeanu S Prevention of restenosis using the gene for cecropin complexed with DOCSPER liposomes under optimized conditions Int J Angiol 2000 9 87 94 10758203
Nikol S Huehns TY Krausz E Armeanu S Engelmann MG Winder D Salmons B Höfling B Needle injection catheter delivery of the gene for an antibacterial agent inhibits neointimal formation Gene Ther 1999 6 737 748 10505096
Isner JM Arterial gene transfer of naked DNA for therapeutic angiogenesis: early clinical results Adv Drug Delivery Rev 1998 30 185 197
Yla-Herttuala S Martin JF Cardiovascular gene therapy Lancet 2000 15 213 22 10675133
| 15836781 | PMC1087498 | CC BY | 2021-01-04 16:39:09 | no | Genet Vaccines Ther. 2005 Apr 18; 3:2 | utf-8 | Genet Vaccines Ther | 2,005 | 10.1186/1479-0556-3-2 | oa_comm |
==== Front
Health Qual Life OutcomesHealth and Quality of Life Outcomes1477-7525BioMed Central London 1477-7525-3-241581712710.1186/1477-7525-3-24ResearchCombination of qualitative and quantitative methods for developing a new Health Related Quality of Life measure for patients with anogenital warts Badia Xavier [email protected] Jose Antonio [email protected] Nuria [email protected] M Angels [email protected] Luis [email protected] los Terreros Miguel Sainz [email protected] Jose Antonio [email protected] Juan Jose [email protected] Health Outcomes Research Europe Group, Avda. Diagonal, 618 1° C-D, 08021Barcelona, Spain2 Hospital Universitario Puerta del Mar, Cádiz, Spain3 Departamento Médico, 3M España, División Farmacéutica, Madrid, Spain4 Hospital Clínico, Madrid, Spain5 Unidad ITS, Gijón, Spain6 Hospital General Universitario, Valencia, Spain2005 7 4 2005 3 24 24 26 1 2005 7 4 2005 Copyright © 2005 Badia et al; licensee BioMed Central Ltd.2005Badia 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
Anogenital warts are the most easily recognized sign of genital Human Papilloma Virus infection. The objective was to develop a short, valid and reliable questionnaire to measure Health Related Quality of Life (HRQL) in patients with anogenital warts.
Methods
First a literature review was performed to identify relevant papers describing the impact of anogenital warts in HRQL; second the main domains were identified by some experts in a focus group, and third in-depth-semi-structured interviews were conducted in patients with anogenital warts to identify the initial set of items. A qualitative reduction of the initial set of items was performed based on the mean scoring of the experts for the three scales: clarity, frequency and importance. The initial questionnaire was pilot tested in 135 patients. Rasch analysis was performed with the results of the questionnaire in order to refine the instrument. Spearman's correlation was calculated between the initial questionnaire and the reduced version. Additionally the measurement properties (validity and reliability) of the resulting final questionnaire were tested and compared using standard procedures (Cronbach's Alpha and item-total correlation).
Results
the main domains identified as affected in patient's life were: sexual, colleagues and partner relationships. After a proper qualitative reduction the initial set of 134 items was reduced to 22. The questionnaire was pilot tested in 135 patients and two dimensions were identified after the multifactorial analysis: emotional dimension and sexual activity dimension. As a result of the Rasch analysis the questionnaire was reduced to 10 items. High correlation was found between the initial and the reduced version for the two dimensions. Cronbach's alpha values were acceptable (0.86).
Conclusion
The initial 22 items questionnaire was reduced by Rasch analysis to a version of 10 items, with two dimensions: emotional and sexual. The results suggest the adequacy of the 10 items to evaluate HRQL of patients with anogenital warts in a valid and reliable way.
==== Body
Background
Any health problem may place significant restrictions on the normal development of physical, emotional and social aspects of a patient's life. Health Related Quality of Life (HRQL) is a multidimensional construct referring to patient's perceptions of the impact of disease and treatment on their physical, psychological, and social function and well-being [1]. The increasing interest in measures reflecting the personal viewpoint of patients' health has led to an extended demand for reliable and valid standardized questionnaires of Health-Related Quality of Life (HRQL) [2,3].
Although the use of HRQL questionnaires is more common in clinical research [4], their use in clinical practice can also help clinicians to obtain standardized information on the impact of the disease and its treatment on patients' HRQL; information that cannot be obtained using traditional measures, and that could be of a great use in clinical decision making.
Taking into account features like item content, scope and target population, instruments can be basically classified as generic or disease-specific [5]. Disease-specific questionnaires, those that only contain items specifically designed for a particular condition, are more likely to be relevant and sensitive to patients in areas that clinicians may wish to monitor [6,7].
Human Papilloma Virus (HPV) is one of the most common sexually transmitted infections. Its prevalence has increased many fold throughout the world, and it is estimated that a total of 1% of the sexually active population are infected. One study in Korea showed that among sexually active students, HPV prevalence was 38.8% in females and 10.6% in males [8]. Another study in Hungary showed an overall prevalence of HPV infection in women of 23% [9]. The number of sexual partners, frequency of sexual intercourse and the presence of genital warts in the partner, have been identified as factors that are related to HPV infection in women. In men only the number of sexual intercourses have shown to be related to the infection [10,11].
Genital warts (sometimes called condylomata acuminata or venereal warts) are the most easily recognized sign of genital HPV infection. Many people, however, have a genital HPV infection without genital warts. Genital warts are soft, moist, or flesh colored and appear in the genital area within weeks or months after infection. They sometimes appear in clusters that resemble cauliflower-like bumps, and are either raised or flat, small or large. Genital warts can show up in women on the vulva and cervix, and inside and surrounding the vagina and anus. In men, genital warts can appear on the scrotum or penis.
Quality of life of patients with genital warts is potentially affected [12]. Given the lack of a disease-specific instrument to measure HRQL in patients with anogenital warts, the objective of the study was to develop a short, valid and reliable questionnaire to measure quality of life in patients with anogenital warts.
Methods
The process of questionnaire elaboration was divided in two phases:
• Phase I: Generation of quality of life issues (domains) and items
A qualitative study was performed to assess the quality of life issues and to develop the items to be included in the anogenital warts disease-specific questionnaire.
The strategy was to first carry out a literature review to identify relevant papers describing the impact of anogenital warts in HRQL as well as to describe any questionnaires previously designed to measure HRQL in patients with anogenital warts. A semi-structured questionnaire was designed based on the results of the literature search. Secondly, a group of experts answered the semi-structured questionnaire in order to identify those domains of anogenital warts they believed would cause problems in patients' life. Afterwards a focus group was organized with the same ten experts. The moderator of the focus group was responsible for leading the discussion among experts in order to reach a final consensus about the dimensions to be considered in the final questionnaire.
• Phase II
This was divided in the following steps:
1. Identification of the initial set of items through semi-structured interviews to a sample of patients.
2. Qualitative reduction of the initial set of items
3. Questionnaire design
4. Administration of the questionnaire to a sample of patients
5. Quantitative reduction of items, Rasch analysis.
1. Identification of the initial set of items through semi-structured interview
With the aim of evaluating the impact of the symptoms of genital warts and its treatment in the quality of life, a group of patients were interviewed, using in-depth semi-structured-questionnaires. The interviewer made sure that the patient took into account all the dimensions previously identified by the experts and also tried to identify new dimensions that could arise during the interview. The interview focused on the impact of the disease on the physical, psychological and social functioning of the patient.
All semi-structured interviews were tape-recorded and transcribed to paper for the qualitative content analysis. A first qualitative analysis was based on the redundancy and repetition of expressions. As a result a first set of items was obtained.
2. Qualitative reduction of the initial set of items
In a second meeting, the experts scored each one of the items. The criteria to include or exclude each one of the 134 items analyzed were based on the mean scoring given by the experts for the three scales: "clarity of wording", "frequency of occurrence", and "importance" among patients with anogenital warts, using a 1–5 Likert-type scale. Items were considered for final inclusion if clarity, frequency and importance were considered high, if clarity and importance were high together with low frequency, and also if frequency and importance were high and clarity low; in this latter situation the item was included but modified to enhance clarity. Low values were defined as those ones below percentile 50 of the distribution of the three scales, and high values where those ones above percentile 50. Based on the results of the scoring of each item a first item reduction was obtained.
3. Questionnaire design and administration to a sample of patients
The pre- identified items were put together in a questionnaire format. The questionnaire was administered to a sample patients with anogenital warts.
4. Quantitative reduction of the items
A multifactorial analysis was performed to define possible dimensions within the questionnaire. In order to refine the instrument a further quantitative reduction was performed with Rasch analysis (dichotomous logistic response model) [13], using the answers to the questionnaire from a group of patients with anogenital warts. The Rasch model constructs a line of measurement with the items placed hierarchically and provides fit statistics to indicate just how well different items describe the group of subjects responding to a questionnaire [14]. In this study a Infit and Outfit value below 1.3 was accepted. No definitive rules exist regarding what is considered acceptable and unacceptable fit, but values greater than 1.3, in a sample smaller than 500, are usually diagnosed as potential misfits to Rasch model condition [15].
A Spearman's correlation was obtained between the original questionnaire and the final reduced version. Additionally the measurement properties (validity and reliability) of the resulting final questionnaire were tested and compared using standard procedures (Cronbach's Alpha and item-total correlation).
Results and Discussion
This study attempts to develop a disease-specific quality-of-life questionnaire for adults with anogenital warts. The fact that is a disease specific one allows the comparison between groups of patients within the same disease.
Phase I: Generation of quality of life domains and items
Pubmed data base was searched [16]. No specific instrument to measure HRQL in patients with anogenital warts was found after the literature review. Only a few studies were found that had considered which dimensions of HRQL were affected in patients with anogenital warts. Based on the results of the literature research a semi-structured questionnaire was elaborated.
Three dermatologists and one gynaecologist answered the semi-structured questionnaire, as well as two experts in the elaboration of HRQL instruments. Given that anogenital warts are mainly seen by dermatologists and gynaecologists in Spain, these two specializations were chosen in order to get the widest perspective of the population affected with anogenital warts and the aspects that impact their quality of life. Three aspects were identified as having impact on HRQL of patients with anogenital warts: sexual relationships, relationships with colleagues, and within the couple relationship. The aspects that worry the patient mentioned by the participants were: contagion, cancer, guilt, anticonceptive methods, pregnancy, infidelity and low performance. In men the most affected aspects were sexual relationships (continuity of sexual relationships, fear to contagion and use of condom). In women the most affected aspects were maternity (fear to pregnancy and possibility to contagion for the baby), fear to cancer, sexual relationships and within couple relationship.
Finally all participants agreed that the aspects that most impact HRQL of patients with anogenital warts are: couple relationship, guilt, and fear of contagion. It is worth pointing out that at this initial state none of the experts mentioned the patient's worry about the effectiveness of the treatment, while all the aspects were more related to patient's relationships.
With all the information gathered in phase I a guideline for the semi-structured interview of patients was elaborated.
Phase II
Identification of the initial set of items
A total of 20 in-depth semi-structured interviews of patients with anogenital warts were obtained between June and July 2003. The sample of 20 patients is considered to be more than enough in qualitative research at the stage of identifying dimensions [17]. The participants were 11 men and 9 women, mean age 34.2, ranging from 19 to 53. Mean time since diagnosis was 25.3 months, ranging from 15 years (1 patient) to 1 month (2 patients). The treatment received by the patients was: criotherapy in 9 cases, inmunomodulators in 2 cases, and no treatment in 9 cases. The interviews were conducted by two experienced psychologists.
The focus group technique could have also been used in this step. The advantages of focus group is that they do not discriminate against patients who cannot read or write and they can encounter participation from people reluctant to be interviewed. [18]. In this case a semi-structured interviewed was chosen because the initial domains were already identified, therefore the semi-structured interviewed was good to provide information on concrete issues allowing open answers.
The dimensions identified were: physical, psychological, social, sexual, everyday life activities, symptoms, perception of the disease, perception of health and perception of treatment. After the qualitative analysis, based on the redundancy and repetition of expressions, a first set of 134 items was obtained.
The selection of items was based on patient's opinion and health professional's views. The aim of this methodological approach is to ensure that the content and validity (specificity and sensibility) of the tool under construction is appropriate for the target population. In other words, whether the questionnaire deals with relevant questions on disease and treatment and does not omit important patient issues [19]. As we see in this study, treatment issues were not initially mentioned by experts while the issue was raised by the patients later on.
Regarding the patient issues the results do not differ from the ones found in the literature. In a survey by the American Social Health Association among people with human papillomavirus infection, more than three-quarters of respondents reported feelings of depression and anger, and two-thirds feelings of shame. Sexual enjoyment and sexual activity were also negatively affected by genital warts [20]. Other studies have shown a negative impact in the social life of the patient [12].
It has been demonstrated that the help seeking behaviour in patients with sexually transmitted diseases is determined by the shame and stigma that they feel [21-23]. This fact might limit the results of the study because those patients who did not seek for health care due to shame and stigma were not able to participate in the study.
Qualitative reduction of the initial set of items
The criteria to include or exclude each one of the 134 items analyzed were based on the mean scoring given by the experts for the three scales: "clarity of wording", "frequency of occurrence", and "importance". A total of 49 items were included.
A further qualitative review of the pre-selected items based on prior experience with other similar instruments concluded in the initial 22 -item questionnaire, which was called CECA 22 (Cuestionario Específico en Condilomas Acuminados, in Spanish) (Table 1). The questions referred to the previous 7 days before the visit, and each question allowed an answer in a 5 options Likert scale (always, almost always, sometimes, rarely, never) with the exception of items 3, 5, and 21 that also allowed the option of "not applicable". The higher the score the better the quality of life.
Table 1 Results of the administration of the 22 item questionnaire to a sample of patients. The items of the final 10-item questionnaire are highlighted in bold
N° of Item Item n Range Mean scoring (SD)
Min Max
1 The discomfort that they cause me affects my daily activites. 135 1 5 3.5 (1.2)
2 I worry too much about my personal hygiene 135 1 5 1.7 (0.9)
3 I am afraid that the lesions won't disappear 135 1 5 2.4 (1.4)
4 I am anxious to know whether I am going to recover from the infection for good 134 1 5 1.7 (1.1)
5 I worry about whether the warts will get worse or if there will be some complications 135 1 5 2.1 (1.3)
6 I have feelings of guilt 135 1 5 3.3 (1.4)
7 I am anxious to know who infected me and when 135 1 5 2.2 (1.4)
8 I worry that I might infect other people 135 1 5 1.4 (0.8)
9 My state of mind is upset (anxiety, depression, sadness, uneasiness...) 135 1 5 2.9 (1.2)
10 I feel more insecure 135 1 5 2.9 (1.3)
11 Knowing that I have the illness affects me in my daily life 135 1 5 3.3 (1.3)
12 I worry about whether I might have problems in having children in the future 135 1 5 3.7 (1.5)
13 I worry about the lack of knowledge concerning my illness and its treatment 135 1 5 2.7 (1.5)
14 I worry about people finding out about my illness (at work, at leisure centres, etc.) 135 1 5 2.4 (1.6)
15 My personal relationships have been affected (partner, friendship, 134 1 5 3.1 (1.4)
16 My sexual drive has decreased 135 1 5 3.4 (1.4)
17 I feel worried during the act 135 1 5 2.8 (1.4)
18 I avoid sexual relations 134 1 5 3.3 (1.4)
19 My sexual relations have decreased in quality and/or frequency 135 1 5 3.0 (1.4)
20 I worry about having to use barrier methods (condom, diaphragm) to have sexual relation 135 1 5 3.5 (1.6)
21 The discomfort produced by treatment affects my daily life 135 1 5 3.7 (1.3)
22 I worry about having to follow a treatment for a long time 135 1 5 2.4 (1.4)
Administration of the questionnaire to a sample of patients
The 22 item questionnaire was pilot tested in a sample of 135 patients diagnosed of anogenital warts from different centers throughout Spain. Mean age (SD) was 31.1 (7.3), 51% of them were women. A total of 76.3% of the patients were heterosexual and 17.8% homosexual. In 32.1% of the patients the anogenital warts were not visible at the moment of the visit. In 6% the extension of the wart was large (>200 mm2), in 29.6% was medium (50–200 mm2) and small in 35.1% (< 50 mm2). The anogenital warts were mostly located in the penis in men and vulva in women. The diversity of the patients' disease situation enhanced the possibility of getting as many broad results as possible.
The questionnaire was completely fulfilled by the 132 patients (response rate 132/135). The results of the administration are shown in table 1. Sexual activities and everyday life activities, as well the worry for having children were the dimensions that obtained higher scoring, meaning that they were the main worry for the patient.
Quantitative reduction of the items
The multifactorial analysis identified two possible dimensions in the preliminary CECA questionnaire. The first one (emotional dimension) was integrated by the first 15 items, and the second (sexual activity dimension) was integrated by the 7 remaining items. Afterwards the Rasch analysis was performed for each one of the dimensions. Rasch model enumerates the items in relation to gravity and selects the items according to: adjustment to the model, redundancy, discriminant validity, and content. As a result of it, the first dimension initially formed by 15 items was reduced to a total of 6 items including: item 3, 4, 5, 9, 10 and 11. There was evidence of the overlapping of item 9 and 10 but the decision of excluding or keeping any of them was totally qualitative. The final decision was to keep the six items. The second dimension (7 items) was reduced to 4 items, including : item 16, 17, 18 and 19. Therefore the final version of the questionnaire had a total of 10 items (Table 1). Global scoring range was 10 to 50, going from 6 to 30 in the emotional dimension and from 4 to 20 in sexual activity dimension. All dimensions were standardized for a scoring between 0 (worst HRQL) and 100 (the best HRQL) in order to facilitate the interpretation and comprehension.
Afterwards a global final analysis was performed to evaluate the adequacy of the total of the selected items in the final reduced version. The correlation between the reduced version and the two dimensions (emotional and sexual activity) in the 22 item questionnaire was high (Table 2). The results of the item-total correlation and the reliability (Cronbach's alpha) for the reduced version as well as for the initial 22-items one are shown in table 3. Item total correlation is the correlation coefficient between the score on an individual item and the total score for the whole set of items of which the individual item is a part. It provides a clue of the internal consistency, the higher the value the higher the consistency (minimum recommended value 0.3).
Table 2 Correlation between the reduced version and the two dimensions (emotional and sexual activity) in the 22 item questionnaire
CECA10 CECA6 (emotional) CECA4 (sexual activity)
CECA6 (emotional) r = 0.851*
n = 133
CECA4 (sexual activity) r = 0.830*
n = 133 r = 0.433*
n = 133
CECA22 r = 0.928*
n = 133 r = 0.845*
n = 132 r = 0.719*
n = 132
* p < 0.01
CECA 22: initial 22-item questionnaire
CECA 10: 10-item questionnaire
CECA 6: emotional dimension in the 10-items questionnaire
CECA 4 sexual activity dimension in the 4-item questionnaire
Table 3 Results of the item-total correlation and the reliability (Cronbach's alpha) for the reduced version as well as for the initial 22-item
CECA22* CECA10 CECA6 CECA4
N° of items 22 10 6 4
Scoring distribution
Valid observations 132 133 134 134
Mean 44.9 44.6 38.8 53.4
SD 17.5 22.3 23.9 29.3
Percentile 50 43.7 40 37.5 53.1
% scoring 0 0 0 5 3 (2.2%)
% scoring 100 0 1 2 8 (5.9%)
Correlation item-global scoring 0.21–0.72 0.53–0.73 0.65–0.79 0.78–0.89
Reliability (Cronbach's alpha) 0.87 0.86 0.84 0.86
Results of Rasch analysis Person separation 2.28 2.21 1.87 1.79
Person reliability 0.84 0.83 0.78 0.76
CECA 22: initial 22-item questionnaire
CECA 10: 10-item questionnaire
CECA 6: emotional dimension in the 10-items questionnaire
CECA 4: sexual activity dimension in the 4-item questionnaire
The mean (SD) scoring for CECA-10, in a 0 to 100 scale, was 44,6 (Table 4). The CECA-6 factor showed a mean (SD) of 38,8 (23.9) and the factor CECA-4 53.4 (29.3). Patients showed worse quality of life (more impact) in the emotional dimension.
Table 4 Global scoring of the CECA 10 questionnaire and for each one of the dimensions
Mean Median Standard deviation Minimum Maximum N Valid
Global scoring CECA-10 0-worse HRQL 100 better HRQL 44.61 40.00 22.33 2.50 100.00 133
Global scoring CECA-6 38.84 37.50 23.93 .00 100.00 134
Global scoring CECA -4 53.6 56.25 29.8 .00 100.00 134
CECA-6: emotional dimension
CECA-4: sexual activity dimension
Even though the final questionnaire was initially designed to be equally valid for both genders, it would be interesting in future research to develop gender specific questionnaires given the characteristics of the disease.
Conclusion
The original CECA questionnaire composed by 22 items was reduced by Rasch analysis to a version of 10 items. Two dimensions were identified in the final questionnaire: emotional and sexual activity. The results suggest the adequacy of the 10 items to evaluate in a valid and reliable way the HRQL of patients with anogenital warts.
==== Refs
Guyatt GH Feeny DH Patrick DL Measuring health-related quality of life Ann Intern Med 1993 118 622 629 8452328
Patrick DL Erickson P Health Status and Health Policy 1993 Oxford University Press, New York
Shumaker SA Naughton MJ Shumaker SA, Berzon R The international assessment of health-related quality of life: a theoretical perspective Quality of Life: theory, translation, measurement & analysis 1995 Oxford: Rapid communications of Oxford Ltd 3 10
Sanders C Egger M Donovan J Tallon D Frankel S Reporting on quality of life in randomised controlled trials: bibliographic study BMJ 1998 317 1191 1194 9794853
Mc Dowell I Newell C Measuring Health: A Guide to Rating Scales and Questionnaires, 1996 2 Oxford University Press, New York
Patrick DL Deyo RA Generic and disease-specific measures in assessing health status and quality of life Med Care 1989 27 S217 232 2646490
Schipper H Clinch JJ Olweny CLM Spilker B Quality of life studies: definitions and conceptual issues Quality of lie and pharmacoeconomics in clinical trials 1996 2 Philadelphia: Lippincott-Raven 11 23
Shin HR Franceschi S Vaccarella S Roh JW Ju YH Oh JK Kong HJ Rha SH Jung SI Kim JI Jung KY Van Doorn LJ Quint W Prevalence and determinants of genital infection with papillomavirus, in female and male university students in busan, South Korea J Infect Dis 2004 190 468 476 15243918 10.1086/421279
Nyari TA Kalmar L Deak J Szollosi J Farkas I Kovacs L Prevalence and risk factors of human papilloma virus infection in asymptomatic women in southeastern Hungary Eur J Obstet Gynecol Reprod Biol 2004 115 99 100 15223174 10.1016/j.ejogrb.2003.10.003
Hippelainen M Syrjanen S Hippelainen M Koskela H Pulkkinen J Saarikoski S Syrjanen K Prevalence and risk factors of genital human papillomavirus infections in healthy males: a study on Finnish conscripts Sex Transm Dis 1993 20 321 328 8108754
Kataia V Syrjanen S Yliskoski M Hippelinen M Vayrynen M Saarikoski S Mantyjarvi R Jokela V Salonen JT Syrjanen K Risk factors associated with cervical human papillomavirus infections: a case control study Am J Epidemiol 1993 138 735 745 8237988
Schofield MJ Minichiello V Mishra GD Plummer D Savage J Sexually transmitted infections and use of sexual health services among young australias women: women's health Australia study Int J STD AIDS 2000 11 313 323 10824940 10.1258/0956462001915949
Rasch G Probabilistic Models for Some Intelligence and Attainment Tests 1993 Mesa Press, Chicago
Wright BD Master GN Rating Scale Analysis 1982 MESA Press, Chicago
Smith RM Schcumacker RE Bush MJ Using item mean squares to evaluate fit to the Rasch model J Outcome Measurement 1998 2 66 78
PubMed [Internet data base] Bethesda (MD): National Library of Medicine (US) c1993- [ref 29 Mar 2005]
Badia X Webb SM Prieto L Lara N Acromegaly Quality of Life Questionnaire (AcroQoL) Health Qual Life Outcomes 2004 2 13 14987332 10.1186/1477-7525-2-13
Kitzinger J Qualitative research. Introducing focus group BMJ 1995 311 299 302 7633241
Arranz P Remor E Quintana M Vilar A Díaz JL Moreno M Monteagudo J Ugarriza A Soto I Perez R Chacon J Garcia-Luaces M Cid A Balda I Lopez MF Gutierrez MJ Martinez E Marrero C Prieto M Sedano C Vaca R Altisent C Hernandez-Navarro F Hemofilia-QoL Group Development of a new disease-specific quality-of-life questionnaire to adults living with haemophilia Haemophilia 2004 10 376 382 15230953 10.1111/j.1365-2516.2004.00918.x
Clarke P Ebel C Catotti DN Stewart S The psychosocial impact of human papillomavirus infection: implications for health care providers Int J STD AIDS 1996 7 197 200 8799782 10.1258/0956462961917618
Fortenberry JD The effects of stigma on genital herpes care-seeking behaviours Herpes 2004 11 8 11 15115631
Liu H Detels R Li X Ma E Yin Y Stigma, delayed treatment, and spousal notification among male patients with sexually transmitted disease in China Sex Transm Dis 2002 29 335 343 12035023
Faxelid E Ahlberg BM Ndulo J Krantz I Health-seeking behaviour of patients with sexually transmitted diseases in Zambia East Afr Med J 1998 75 232 236 9745841
| 15817127 | PMC1087499 | CC BY | 2021-01-04 16:38:14 | no | Health Qual Life Outcomes. 2005 Apr 7; 3:24 | utf-8 | Health Qual Life Outcomes | 2,005 | 10.1186/1477-7525-3-24 | oa_comm |
==== Front
J Immune Based Ther VaccinesJournal of Immune Based Therapies and Vaccines1476-8518BioMed Central London 1476-8518-3-11580435110.1186/1476-8518-3-1Original ResearchBCG vaccination at three different age groups: response and effectiveness Briassoulis George [email protected] Irene [email protected] Vasilis [email protected] Athina [email protected] Department of Pediatrics, Markopoulo District Health Centre, Markopoulo Attikis, Greece2 Current affiliation Pediatric Intensive Care Unit, University Hospital of Heraklion, Faculty of Medicine, Heraklion, Crete, Greece3 Microbiology and Transfusion Departments, NIMTS Hospital, Athens, Greece2005 1 4 2005 3 1 1 13 12 2004 1 4 2005 Copyright © 2005 Briassoulis et al; licensee BioMed Central Ltd.2005Briassoulis 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 protection, which some BCG vaccines could confer against the development of tuberculosis (TB) in childhood, might be indirectly reflected by the subsequent development of BCG immune response. The objectives of the study were to examine effectiveness and possible differences of post-vaccination reaction to a lyophilized BCG at different age groups and to evaluate its protection against TB in a decade's period.
Methods
We studied the post-vaccination PPD-skin reaction and scar formation at three different school levels, corresponding to ages of 6, 12 and 15 years old, vaccinated by a lyophilized BCG vaccine (Pasteur Institute), currently used in our country. During a 10-year follow up the reported TB cases in vaccinated and non-vaccinated adolescences up to 24-years old were analyzed and compared to the number of cumulative cases observed in the adult population of two neighboring territories (vaccinated and non-vaccinated).
Results and Discussion
There was a significant correlation (r2 = 0.87, p < 0.0001) between tuberculin induration and scar formation. There was no statistically significant difference between the three age groups (6, 12, and 15 year-old, respectively) in regard to the diameter of tuberculin induration or scar formation. Although 34% of 10-year later indurations were unpredictably related to the initial ones (increased or decreased), they were significantly correlated (r2 = 0.45, p = 0.009). The relative percentage of TB for the 14–24 years-age group to the adult studied population was significantly lower among the immunized children compared to the non-immunized population of the same age group (17/77, 22% vs. 71/101, 70%, p < .0001).
Conclusion
Our data suggest that the lyophilized BCG vaccine used for BCG programs at different age groups is equally effective and may confer satisfactory protection against tuberculosis in puberty.
==== Body
Background
As one third of the world's population is already infected with mycobacterium tuberculosis (TB), efficacious control of TB, one of the world's major health threats, is best achieved by a combination of chemotherapy and vaccination. Bacillus Calmette-Guerin (BCG) vaccination is compulsory in 64 countries and recommended in others [1]. Recently, the World Health Organization expanded programs of immunization recommended BCG at 3 months [2], while in many areas there is vaccination at birth [3], at school entry and in adolescence [4]. The policy in Greece for some years has been to recommend BCG vaccination routinely in schools to children aged 11 to 13 years, but adjusted to a current tuberculous infection index of 3.4%, this schedule has been recently endorsed by offering the vaccine at school entry (ages 5 to 7 years), until the risk of infection is low everywhere in Greece. The justification for continuing vaccination throughout the country even although TB is now rarely encountered in some areas, is related to the mobility of the population: many young people study, join the army, or seek work in areas where TB is more prevalent [5].
The efficacy of BCG vaccination, however, has been strongly questioned [6]. Studies in older children and adults showed 77% protection in Britain [7]), only 14% in southern USA [8], and none in Madras [9]. Furthermore, in a retrospective study of 22 children with TB of the spine in a developing country, all gave a history of BCG vaccination scars [11]. In addition, several reports recommending the continuation of the policy of BCG vaccination offered routinely in schools, are now concerned about the influence of the quality of the vaccine, its transportation, and the technique of its application on the protection obtained [10]. Laxity in the TB control programmes and widespread of HIV could also appear to have a role in the recent reurgence of TB infection worldwide [11]. However, methodological and statistical reappraisal has showed that different biological and environmental conditions in individual studies and, principally, biases or inadequate statistical power may have contributed to the conflicting data [12]. Recently, findings at 15 years showed that even among populations with high infection rates and high non-specific sensitivity where BCG did not offer any protection against adult forms of bacillary pulmonary tuberculosis, BCG offered some level of overall protection (up to 50%) in children [13].
BCG induced tuberculin sensitivity is a quantitative characteristic and has been used to compare vaccine efficacy. It has been also suggested that the protection which some BCG vaccines could confer against the development of TB in childhood, might be indirectly reflected by the subsequent development of BCG immune response [14,15]. The preliminary research, however, has been restricted to examine the effectiveness of freeze dried vaccines only, and has not been extended to study the immunologic properties of the lyophilized BCG vaccines, currently used in Europe. Furthermore, those limited studies were properly oriented to evaluate the protective effect of BCG in either infant Asians or high risk neonates, thus including too few details to permit assessment of the main clinical expressions of acquired immunity and to draw the indisputable conclusions about the significance and its derived implications among the general population.
As part of a prospective assessment of the efficacy of a lyophilized BCG vaccine used in current BCG schemes in Greece, we had the opportunity to evaluate its immunologic response to study the interrelations between the post-vaccination tuberculin sensitivity and scar formation, and to analyze their epidemiological data in children who had received BCG in three different age groups. During the next decade, we comparatively recorded the 10-year reported TB cases in vaccinated and non-vaccinated adolescents up to 24-years old in our territory and in a neighboring area not covered by similar preventive program, and compared them to the 10-year number of cumulative cases observed in the adult population of the two areas.
Methods
Participants
Intradermal BCG vaccination has been offered routinely to schoolchildren living in the district area covered by our health center since 1988 and continues up to date. Children participating in the preventive program from 1/10/1988 to 1/10/1993 were enrolled in the study. To restrict escapes, BCG scheme was annually applied at three different school levels, corresponding to ages of about 6, 12 and 15 years old, respectively. All children monitored in each of the cohorts were documented to be receiving their first BCG vaccination when they were in their respective age groups and had not previously been vaccinated (there is no neonatal BCG program in Greece and also all children should have had negative tuberculin testing before been enrolled in the study). Children, who had received BCG vaccination at birth or at any other time because of known contact with a case of TB, were excluded from the study. Likewise, notified subjects given chemoprophylaxis, whether tuberculin positive or negative, were also excluded from analysis. Of a total of 1,124 vaccinated schoolchildren included in the final analysis (group A), 394 received their first vaccination at age 6 and accordingly were classified in the 6 year -old group, 483 in the 12 year-old first vaccination group and 247 in the 15 year-old first vaccination group.
Methodology
Vaccine was given by a doctor and team of health visitors experienced in vaccination technique. Lyophilized vaccine (0,1 ml, Pasteur inst.) was injected intra-dermaly over the insertion of the left deltoid muscle to produce a weal of about 7-mm diameter, using a separate syringe and a 27-G needle for each person [10]. The reconstituted vaccine used at our center contains about 0.15 mg moist weight of Galmette-Guerin organisms per ml, which implies a concentration of colony-forming units of viable organisms of 6 × 106/ml. In all cases using 10 UI of PPD was performed three days before vaccination along with a thorough clinical examination. Tuberculin testing was repeated three months after vaccination when children were again reviewed. The result was read three days later and induration of 6 mm or more, when read across the forearm, was regarded as positive result [14]; induration of 4–5.9 mm as a weak positive and induration less than 4 mm as a PPD-negative. Scar formation was classified as real scar (>2 mm), tiny (≤ 2 mm), or not visible [10]. No other immunologic studies were done to correlate immunity to BCG with assessment of other in vitro responses to vaccination, since in such a massive program it would not be feasible to do many things in many individuals in a short time, especially, in a school environment. As the study did not affect patient care, the institutional review board waived the need for informed parental consent.
Sub cohorts
Those who gave a negative result retested within one month. In proved tuberculin negative responders (Mantoux < 4 mm), revaccination was suggested and if consent was obtained, the vaccine was given apart from the site of the scar of the previous BCG vaccination (group R) as it was previously advised [10].
Controls
Children in schools not participating in the BCG prophylaxis program (schools in neighboring areas not covered by our health center) were used as controls for the follow up period (non vaccinated adolescences). In total 1340 control children of comparative age groups were recorded the same time period. The two territories had similar population according to the latest data (2001) available by the National Statistics Center (vaccinated 375024 and non-vaccinated 348236).
Follow-up
Tuberculin testing was repeated ten years after vaccination in a small proportion of children, still attending school and if an informed parental consent had previously been obtained. Also, during the intervened decade, we comparatively recorded the 10-year reported TB cases in vaccinated and non-vaccinated adolescents from more than 14 up to 24-years old in the two territories. Data were drawn from the Health Centre's records and the National Statistics Center of Infectious Diseases, and subsequently were analyzed and compared to the 10-year number of cumulative cases observed in the adult population (more than 25 years old) in the same areas (vaccination area covered by our health center and non-vaccinated neighboring area not covered by our health center). There were not any epidemiologic data available regarding the incidence of TB in the 6 up to 14-year age group.
Statistical analysis
Methods for assessing data significance included the two-sample null hypothesis for a two-tailed test, the Mann-Whitney U test and the x2 test with Yates' correction, using a standard statistical package.
Results
Tolerability
In all subjects tuberculin testing before vaccination gave a negative result and on the basis of history or clinical examination there was no contraindication to perform BCG vaccination. The procedure was well tolerated with no cases of osteopathy and only very few of reactions such as, three cases of transient lymphadenopathy and one of local subcutaneous abscess.
Age groups
Details of tuberculin testing and scar formation at each age group are shown in Figures 1 and 2. There was no statistically significant difference between the three age groups (6, 12, and 15 year-old, respectively) in regard to the diameter of tuberculin induration or scar formation. The mean (SD) diameter of the tuberculin reaction was 11.13 (4.20) mm. The distribution of Mantoux induration diameters appeared to be the one of normal curve with the 95% confidence limits set at 2.73 mm and 19.53 mm (binomial distribution). Extensive search revealed no evidence of active TB in those schoolchildren with a tuberculin induration greater than or equal to 20 mm. Although 1034 (92.2%) subjects gave a positive post-vaccination tuberculin reaction, 45 (4%) gave a weak positive induration, and 43 (3.8%) gave a PPD-negative test. Real scar formation has been developed in 96% of the vaccinated schoolchildren and a tiny scar (≤ 2 mm) in 3.5%.
Figure 1 Post-vaccination tuberculin testing in three school-age groups
Figure 2 BCG scar formation in three school-age groups
Tuberculin induration and scar formation
There was a significant correlation (r2 = 0.87, p < 0.0001) between tuberculin induration and scar formation (Figure 3). Table 1 shows that none of the children without scar formation has developed PPD-positive reaction. Instead, 99.4% of the children with positive tuberculin skin testing exhibited scar formation compared to only 39.5% of those who were negative for tuberculin reactivity (p < 0.001). Accordingly, the possibility for a schoolchild to not develop a visible scar has been significantly higher among Mantoux negative children than among those who presented with either a positive Mantoux (p < 0.001), or a weak positive tuberculin testing (p < 0.001). The real estimated percentage of the possibility for absence of scar formation in Mantoux negative children has been included between the 95% limits of 14% and 38%, the most possible value been an estimated percentage of 23 % (binomial distribution).
Figure 3 Correlation between tuberculin induration and scar formation
Table 1 No (%) of BCG scar formations in three groups of different post vaccination tuberculin reactions among school children
Mantoux Children Evaluation of scar formation [n (%)]
Induration (mm) n Real scar (>2 mm) Tiny scar (≤ 2 mm) No scar (?)
Negative (0–3,9) 43 17 (39.5)* 20 (46.5)**,*** 6 (14)***,***
Weak positive (4–5,9) 45 32 (71.1) 13 (28.9)**,*** 0 (0)***,***
Positive (>6) 1036 1030 (99.4)* 6 (0.6)***,*** 0 (0)***,***
Total 1124 1079 (96.0) 39 (3.5) 6 (0.5)
p values apply to differences among negative, weak positive and positive indurations for the same scar formation groups (real scar, tiny or not visible scar).
* < 0.01, ** p < 0.025, *** p < 0.001.
Tuberculin negative subjects were more likely to develop a tiny scar than the subjects with a positive (p < 0.001) or weak positive Mantoux test (p < 0.025), respectively. There has been a higher percentage of tiny scars in children with tuberculin indurations of 4–5,9 mm than in those with the next range (6–9 mm) of induration diameters (p < 0.001). The real estimated percentage for the presence of a tiny scar in children with a tuberculin reaction of less than 6 mm has been included between the 95% limits 42% to 54% (mean 47%, binomial distribution)]
Negative reactors
Of the 43 tuberculin negative reactors, 32 cases were reviewed. Lack of reactivity was confirmed and all 32 had the BCG vaccine repeated. Surprisingly, following revaccination, reactivity levels to 10 UI of PPD (90.6%) were similar to the reactivity rates documented in the initial vaccination group A cohort (92.2%, NS), corresponding to 100% scar formation. There was not only an apparently similar distribution of tuberculin indurations (p < 0.42), but also a similar mean between groups (group R 12.76 (5.40) mm vs. 11.13 (4.20) mm for group A, p = 0.1).
Follow-up
A. Tuberculin induration
Ten years later the cohorts became inconsistent and many children disappeared (studies, army, new families) or – when found – most denied having a repeated Mantoux test. Among 183 16-year old schoolchildren (former 6 year -old group) been visited by health visitors at schools in 2005, 110 denied to participate in the follow-up of the study. Thus, approximately 10 years later (2005), we only managed to repeat Mantoux in 73 children of the 6 year-old group of the 1993–5 year period. There was no statistically significant difference between the two year periods in regard to the diameter of tuberculin indurations (mean (SD) diameter of the tuberculin reaction 8.4 (5) mm in 2005 vs. 8.1 (6) mm in 1995). The mean paired difference of the paired samples test was non significant: .25 (6) mm (95% confidence interval of the difference: lower: -1.9, upper: 2.4 mm). Although the paired sample correlation (Figure 4) was significant (r2 .45, p = 0.009) and new individual values were almost identical to the old ones in 66% of children (± 2 mm), a great proportion of the 10-year later values were unpredictably related (12% decreased, 22% increased) to the initial ones (Figure 5).
Figure 4 Paired sample correlation (quadratic regression) of tuberculin indurations between the two time – periods (initial study period and 10-year later follow up)
Figure 5 Follow up of recent individual indurations (2005) correlated to initial Mantoux values (10-year time interval)
B. TB cases
Data available in the National Statistics Center of Infectious Diseases and in the Health Centre's records showed that the TB cases recorded among vaccinated adolescents – more than 14 up to 24 years-old – during the last decade in our territory were lower compared to the non-vaccinated adolescents (17 vs. 71 cases, in a decade). These numbers would have resulted in the wrong estimation of 1.3% vs. 6.3% TB incidence in adolescenthood for the vaccinated schoolchildren and controls, p < 0.0001, if they had only been restricted to the studied population. However, the recorded TB cases were extended to a decade's population, including at least three times more similarly treated children in each group, making therefore statistical analysis impossible. The numbers of the recorded TB cases for the adult population (>25 years old) in the two territories recorded during the same period were slightly lower for the vaccinated area compared to non-vaccinated territory (77 vs. 101 cases). Thus, the relative percentage of TB for the 14–24 years-age group to the responded adult population, which could provide us with the only comparable numbers, was significantly lower among the immunized children (17/77*100) compared to the non-immunized population (71/101*100) of the same age group (22% vs. 70% of all ages, x2 = 14.7, p < 0.0001).
Discussion
Because of the conflicting results of major controlled trials, BCG vaccination against TB remains controversial despite more than 50 years of use. The wide range of BCG protective efficacy values reported by the trials extended from 75%, reflecting substantial protection by the vaccine [3] to negative ones indicating a higher rate of TB among vaccinees than controls [16].
Although the tuberculin state after vaccination was not generally thought to influence the degree of protection offered by BCG, children who did not become sensitive to tuberculin either died subsequently of disseminated TB [17] or developed tuberculous meningitis [18]. Furthermore, the relative increase of TB in the mainly unimmunized cohorts born in a European country after 1975, compared with the mainly BCG immunized cohorts born there in the period 1969–1974 was, by the end of 1984, estimated at 6 (95% confidence interval, 2.3 to 16.1) [19]. Subsequently, when a high level of post-vaccination immune response was achieved, the estimated protective efficacy for the vaccine was shown to be 64% with 95% confidence limits of 43% and 77% and the prevented fraction 0.50 [15].
Conflicting results from two centers in the United Kingdom showed that only 45–46% of vaccinated children were Mantoux positive when tested between the ages of 3 months and 2 years [20]; 25% had no visible scar [21]. Another study carried out in a third center, however, showed that 353 (98%) out of 361 Asian neonates given BCG were tuberculin positive when tested three months later [14]. Tuberculin conversion rates of 93% or 88% were also established following BCG vaccine in 15 term and 8 pre-term infants, respectively [22]. In addition, a more recent study among 193 Asian vaccinees revealed a positive post-vaccination tuberculin testing in 184 of them (95%) [15]; these results are consistent with those of the present study confirming a 94% prevalence of immune response to BCG in a sample of 1124 vaccinated Greek schoolchildren. If impaired immunity was the cause of diminished tuberculin sensitivity after vaccination, then a fairly constant proportion of Asian or Greek children would be expected to be non-reactors to tuberculin after vaccination. In another study, significantly higher proportions of infants given Japanese BCG were found to be tuberculin convertors (74.7%) when compared to those given British BCG (51.4%). In our study groups there was no difference between initially or repeatedly vaccinated children (92% vs. 90%), and this was also reflected in the frequency of non-visible scar formations. In the negative reactors, however, BCG vaccination could have triggered protective (Lister type) rather than antagonistic (tuberculin or Koch type) reactions, which have been speculated to be the most protective [23]. Similar to other study findings done in infants [24], we also did not observe any significant difference in mean tuberculin reaction, tuberculin positivity and mean scar size according to different age group at administration.
Results of our study confirm a highly positive correlation between scar formation and post vaccination tuberculin sensitivity among schoolchildren at various ages. This finding is remarkably similar to that suggested by reports in India and the United States, which have shown that the size of the BCG scar was associated with considerable enhancement in sensitization to tuberculin [25,26]. Similarly, in a recent study the prevalence of skin test positivity was also consistently higher among individuals compared to those without a BCG scar [27]. Our data confirm this trend, since children who failed to produce a BCG scar did not have evidence of a positive post-vaccination immune response to tuberculin skin testing. It's important to note however that negative skin test reactivity can still be associated with real (46.5%) or tiny scar (39.5%) formation (37 or 43 patients, total 86%). It must be also noticed that in contrast to most previous studies, this study uses a lyophilized vaccine, demonstrating the basic characteristics of a recently used BCG vaccination in a western country upon the entry of the new millennium.
The 1079 subjects with a scar are assumed to have mounted an initial response to the vaccination, as reflected by a strong positive immune response in the 1030 of them (99.4%). Possibly, a few negative or weak responses were due to physiological implications such as general health nutritional status, prevalence and especially virulence of atypical mycobacteria [23], an infection with M. Kansasii approximating 80% of the potential protection offered by BCG [28]. Furthermore, in a population with a low level of sensitivity to PPD (median skin reactions 3 mm), 36% of the subjects gave a greater response to an atypical antigen such as PPD-B [29]. It is unlikely, however, that all negative results were caused by this, since most of the Mantoux negative children turned to become positive reactors after they subsequently underwent repeat BCG vaccination. This finding which has been further enhanced by an observed similar to the expected, according to our results, distribution of tuberculin indurations, is in agreement with recent reports, suggesting that revaccination resulted in a significant increase in positivity to tuberculin 10 and to other reagents tested [30].
Accordingly, it seems unlikely that the impaired reaction to tuberculin in some of the recently vaccinated subjects had been caused by a defect of initial recognition of the antigen or an inability to retain this information or even by a failure of sensitized lymphocytes to react because the individual was malnourished or had a serious infection [23]. Instead, there is circumstantial evidence that even when given at birth, BCG achieves tuberculin conversion in a high proportion of neonates [31], irrespective of race [32], ethnic origin [33] or prematurity [24]. Since age has not been a significant factor in influencing BCG immune response in the present study, the age mass vaccine must be offered at should be ideally adjusted to the one, shortly before the infection rate is going to be accelerated [6].
Although the mean (SD) diameter of postvaccination tuberculin reactors in our study was unexpectedly high, (11.13 (4.20) mm), this was significantly lower than a mean of 17.9 mm calculated for naturally infected subjects [33,34] and was not significantly different from the mean postvaccination reaction of 9.4 (2.7) mm reported by others [15]. Surprisingly, also, a much wider distribution range of tuberculin indurations has been included between the 95% confidence limits for the sample (2.73–19.53 mm). In fact, a 71.1% of those with a weak positive Mantoux (4–5.9 mm) exhibited a scar of >2 mm, which has been previously suggested to adequately represent a positive immune response to BCG vaccination [35]. The rest of them (28.9%) exhibited a tiny BCG scar, presumably underlining a rather smoother and wider transition zone between PPD positive and negative reactions. Examining two different batches of new tuberculin, Stanford JL and Tala-Heikkila found that the mean size of the BCG scar was 8.1 (SD 4.8) mm, and there was a trend associating smaller BCG scars with smaller tuberculin responses, which did not reach statistical significance [36]. Although such an explanation for the right end of the curve appears to be in disagreement with the assumption that reactions larger than 10 mm in diameter are likely to represent infection [30], a similar wide transition zone for variable positive reactions is also strongly supported in our study by the demonstration of a previous negative PPD testing, immediately followed by the BCG vaccination which has been shortly preceded the tuberculin inversion. Thus, the large tuberculin reaction achieved in a substantial proportion of children shortly after BCG might simply represent a strong cell-mediated immune response among individualized expressions of recently acquired immunity. Besides, BCG vaccination at birth and for school age children causes reactivity to tuberculin which persists for 20 to 25 years, so that an induration diameter of > 15 mm does not exclude a vaccinal origin [37]. Similarly, in our study among schoolchildren, no significant waning of immunity to BCG was shown 10 years after initial vaccination. We also showed that although the immune response to BCG persisted for so long, individual responses varied widely (increased or decreased) in 1/3 of children, the meaning of which couldn't be extrapolated. In a recent study, prior BCG vaccination had a strong influence on skin test results of <or = 18 mm in diameter among persons <40 years old, compared with the influence of factors predictive of M. tuberculosis infection [38]. For vaccinated subjects with a previous negative tuberculin test, it is also necessary to exclude the booster effect. Thus, according to our results and those of others [39], an induration diameter of > 15 mm does not exclude a vaccinal origin [38]. Additionally, in some cases, such a reaction might also suggest that sensitization to mycobacteria species might have occurred at a very young age. Although such a sensitization might not be detected by the tuberculin test, it can influence response to BCG vaccination [40]. The marked difference of tuberculin reaction rate between Indian towns strongly supports such an influence of exposure to mycobacteria in the environment [27]. Although it has been found that in a population with a high level of sensitivity to PPD (median skin reaction 12 mm), only 7% of the subjects tested gave a greater response to an atypical antigen than to PPD [31], determination of the optimum range of reactions to the BCG vaccination in a given population is a more contentious issue, which cannot be fully resolved without further information about the interactions among the quality of vaccine, the time elapsed, the prevalence and virulence of atypical mycobacteria, the booster effect, the immunologic memory and the mechanisms of its responsiveness to other reagents [40]. It is therefore not possible in the late post-vaccination period to distinguish between a tuberculin reaction caused by virulent supra-infection and one resulted from persistent post-vaccination sensitivity, even in the case of a strong positive reaction to 10 UI of tuberculin PPD.
Repeat BCG vaccination, malnutrition, and BCG with scars present difficulties in making a diagnosis of TB but did not affect PPD reactivity and did highlight the need for thorough clinical evaluation [41]. Although, high tuberculin sensitivity in healthy schoolchildren may be partially maintained by contact with environmental mycobacteria, attributing a 'positive' Mantoux response to past BCG vaccination may be encouraging a false sense of security in contacts recently exposed to an infectious case of TB. Our results confirm the general assumption that the major disadvantage of BCG is that it clouds interpretation of the tuberculin skin test [42]. These results, however, do also suggest that the notification of post-vaccination tuberculin induration diameter about 3 months after BCG, might well be served as a self-control measurement in the case of clinical diagnosis of TB [43], since post-vaccination skin reactions tend usually to decrease with time. Additional factors, such as age of the contact and sputum status of the index case are important determinants of the degree of increased tuberculin sensitivity [44]. However, at the rise of the millennium, new blood assays (QuantiFERON-TB, CSL Limited), which measure gamma interferon production when M. tuberculosis-specific proteins, such as the ESAT-6, are incubated with venous blood samples, are promising for the recognition of infection with Mycobacterium tuberculosis, since they are not influenced by past BCG exposure [45].
Although vaccination does not prevent the establishment of infection in someone exposed to tubercle bacilli, its effect does limit the multiplication and dissemination of tubercle bacilli and the development of lesions after infection. The direct effect of BCG vaccination is defined as prevention of TB in vaccinated persons and the indirect as the reduction of TB in the population as a whole [30]. Results of epidemiological research laboratories, however, showed considerable variation in the current BCG vaccination policy in different districts while the population is highly mobile [37]. To allow the proportion of unvaccinated young people to increase while more than 3,420 new sputum positive cases of TB are reported annually in Greece – now much more increased among AIDS victims -increases the risk of disease, which may well not be diagnosed until others have been infected [46].
Since there is evidence in the neonatal period, but not in childhood, challenging the view that sensitization is essential for protection [47], the assumption that a positive skin test following vaccination is an indicator of BCG-induced immunity against Mycobacterium tuberculosis might not have been a correct one. We thought, however, that because a high rate of PPD skin test conversion following BCG vaccination in schoolchildren and its persistent reactivity for 10 years might both increase the degree of protection offered by BCG, it should be concluded that the particular lyophilized BCG vaccine used for BCG programs in Greek schools may confer satisfactory protection against TB in puberty. Our results are further enhanced by the findings of a recent study, which showed that BCG vaccination at birth and for school age children causes reactivity to tuberculin which persists for 20 to 25 years [48]. The recorded decline of TB among the adolescents in the vaccinated area, highlights the need to sustain and, wherever possible, intensify preventive and case finding measures in all groups at special risk [49], but it might also suggest that the continuation of offering mass BCG vaccination should ideally be maintained everywhere, including regions that currently have a low prevalence of TB [5]. Similarly, in view of the current incidence of TB in Finland and the likelihood that lymph node infections and sensitivity to environmental mycobacteria will increase, continued BCG vaccination at birth has been recommended [37].
One of the main limitations of this study is that the sample size for a comparison of disease occurrence between the study and control groups was too small and thus statistically underpowered for such an analysis. Furthermore, it wouldn't be practically possible to follow up and record disease occurrence in the cohorts, since there was no computerised system importing data from hospitals or insurance agencies into the health centre and patients do not often reveal such medical confidentialities by themselves. Accordingly, since we had covered all schools at three different age-levels in our territory, and there was not any BCG program running in the control territory, we made the logical assumption that vaccination was widespread in our territory, and absent in the other. A severe limitation of this hypothesis, however, is the prerequisite of a second assumption that no significant migration and no significant differences in the prevalence of latent TB infection or the proportion of adult cases with positive sputum smear, must have occurred over the study period. For all these reasons, the prevention part of this study is not so strong and has accordingly not been the primary endpoint of the study. Another limitation of the study is that we did not succeed in following all cohorts further, retesting them some years later to assess age-related differences that may have been substantial in terms of waning responses. Children of the 12 and 15 year-old cohorts had finished school, the cohorts became inconsistent, and even when personal data were still available (addresses, phones) it was found extremely difficult to bring people back to the health centre. The response of the retested patients, however, might offer a realistic idea of what reaction could have been found in the other age groups as well.
Continuation of the BCG programs could easily work as a bridge to the coming era of new generation of TB vaccines. It has been already shown that vaccination with ESAT-6 antigen from mycobacterium tuberculosis, which is a dominant target for cell-mediated immunity in the early phase of tuberculosis, delivered in a combination of monophosphoryl lipid A and dimethyl dioctadecylammonium bromide, which are adjuvant efficients for the induction of cellular and humoral immune responses, elicited a strong ESAT-6-specific T-cell response and protective immunity comparable to that achieved with mycobacterium bovis BCG [50]. As one third of the world's population is already infected with Mycobacterium tuberculosis and because the acquired immune response is mediated by different T-cell sets, two types of vaccine may be required: one for eradication of already established infection and the other for prompt combat of invading microbes [51]. Meanwhile, emphasis should especially be placed on the importance of the quality, transportation and preservation of the vaccine and on the technique of application [10], that may be responsible for the widely divergent results of prospective BCG trials in several countries, that have lead to current doubts about the efficacy of BCG vaccination in TB prevention.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
GB conceived of the study, participated in its design, and drafted the manuscript. IK participated in the collection of the data and performed the statistical analysis. VG collection of the data and coordination and helped to draft the manuscript. AT participated in the design of the study and performed the statistical analysis. All authors read and approved the final manuscript.
Acknowledgements
We thank the health visitors and social workers of our Health Center for their collaboration and support in performing this study.
==== Refs
Anonymous Is BCG vaccination effective? [editorial] Tubercle 1981 62 219 221 7032020 10.1016/0041-3879(81)90011-8
Ten Dam HG Hitze KI Does BCG vaccination protect the newborn and young infants? Bull WHO 1980 5B 37 41 6991146
Curtis HM Leck I Bamford FN Incidence of childhood tuberculosis after neonatal BCG vaccination Lancet 1984 1 145 148 6140451 10.1016/S0140-6736(84)90072-2
Department of Health and Social Security The school BCG vaccination programme 1985 London: DHSS (DA(85)27).
Horne NW Control and prevention of tuberculosis: a code of practice [editorial] Thorax 1984 39 321 325 6740534
Farer LS The current status of tuberculosis control efforts Am Rev Respir Dis 1986 134 402 407
Hart RDA Sutherland I BCG and vole bacillus vaccines in the prevention of tuberculosis in adolescence and early adult life BMJ 1977 ii 293 295 326347
Comstock GW Palmer CE Long term results of BCG vaccination in the Southern United States Am Rev Respir Dis 1966 93 171 183 5908081
Tuberculosis prevention trial Trial of BCG vaccines in South India for tuberculosis prevention Bull WHO 1979 57 819 827 396057
Public Health Laboratory Service Communicable Disease Surveillance Centre BCG vaccination BMJ 1983 286 876 877 6403119
Sayi EN Mlay SM Tuberculosis of the spine in children at Muhimbili Medical Centre, Dar es Salaam East Afr Med J 1995 72 46 48 7781556
Clemens JD Chuong JJ Feinstein AR The BCG controversy. A methodological and statistical reappraisal JAMA 1983 249 2362 2369 6834635 10.1001/jama.249.17.2362
Research Centre (ICMR), Chennai Fifteen year follow up of trial of BCG vaccines in south India for tuberculosis prevention Tuberculosis Indian J Med Res 1999 110 56 69
Handfield JW Allan J Windebank WJ Sensitivity of neonates to tuberculin after BCG vaccination BMJ 1986 292 989 991
Packe GE Innes JA Protective effect of BCG vaccination in infant Asians: a case-control study Arch Dis Child 1988 63 277 281 3258499
BCG Bad news from India Lancet 1980 1 73 74 6101419
Rosenthal SR Loewinsohn E Graham ML Liveright D Thorne MG Johnson V BCG vaccination against tuberculosis in Chicago. A twenty year study statistically analysed Pediatrics 1961 28 622 641 14494079
Chamberlayne EC (ed) Status of immunisation in tuberculosis in 1971 1971 Bethesda: National Institutes of Health (DHEW publication No (NIH) 72–68).
Romanus V Tuberculosis in Bacillus Calmette-Guerin-immunized and unimmunized children in Sweden: a ten-year evaluation following the cessation of general Bacillus Calmette-Guerin immunization of the newborn in 1975 Pediatr Infect Dis J 1987 6 272 280 3575009
Lumb KM Bandaranayake R Beran PJ BCG vaccination in infancy Public Health 1986 100 54 55 3749439
Grindulis H Baynham MID Scott PH Thompson RA Whalton BA Tuberculin response two years after BCG vaccination at birth Arch Dis Child 1984 59 614 619 6465929
Dawodu AH Tuberculin conversion following BCG vaccination in preterm infants Acta Paediatr Scand 1985 74 564 567 4024927
al-Kassimi FA al-Hajjaj MS al-Orainey IO Bamgboye EA Does the protective effect of neonatal BCG correlate with vaccine-induced tuberculin reaction? Am J Respir Crit Care Med 1995 152 1575 1578 7582297
Aggarwal A Dutta AK Timing and dose of BCG vaccination in infants as assessed by postvaccination tuberculin sensitivity Indian Pediatr 1995 32 635 639 8613331
Stanford JL Sheikh N Bogle G Baker C Serries H Mayo P Protective effect of BCG in Admednagar, India Tubercle 1987 68 169 176 3448795 10.1016/0041-3879(87)90052-3
Sepulveda RL Burr C Ferrer X Sorensen RU Booster effect of tuberculin testing in healthy 6-year-old school children vaccinated with Bacillus Calmette-Guerin at birth in Santiago, Chile Pediatr Infect Dis 1988 7 578 581
Fine PE Bruce J Ponnighaus JM Nkhosa P Harawa A Vynnycky E Tuberculin sensitivity: conversions and reversions in a rural African population Int J Tuberc Lung Dis 1999 3 962 975 10587318
Krugman S Katz JL Infectious diseases of children 1981 Seventh St Luis, Toronto, London: CVMosby Co 442 443
Landi S Ashley MJ The significance of a dual skin test with PPD and PPD-B in humans after BCG vaccination Dev Biol Stand 1986 58 631 636 3111907
Shaaban MA Abdul Ati M Bahr GM Stanford JL Lockwood DN McManus IC Revaccination with BCG: its effects on skin tests in Kuwaiti senior school chidlren Eur Respir J 1990 3 187 191 2311743
Anonymous BCG vaccination in the newborn [editorial] BMJ 1980 28 1445 1446
Tidjani O Amedone A TenDam HG The protective effect of BCG vaccination of the newborn against childhood tuberculosis in an African community Tubercle 1986 67 269 281 3499015 10.1016/0041-3879(86)90016-4
Kue Young T Hershfield ES A case-control study to evaluate the effectiveness of mass neonatal BCG vaccination among Canadian Indians Am J Public Health 1986 76 783 786 3717464
Wolinsky E Tuberculosis Baum GE, Wolinsky E Textbook of pulmonary diseases 1983 3 Boston, Toronto: Little Brown Co 507 572
Santiago EM Lawson E Gillenwater K Kalangi S Lescano AG Du Quella G Cummings K Cabrera L Torres C Gilman RH A prospective study of bacillus Calmette-Guerin scar formation and tuberculin skin test reactivity in infants in Lima, Peru Pediatrics 2003 112 e298 e298 14523215 10.1542/peds.112.4.e298
Stanford JL Tala-Heikkila M Good concordance between two batches of new tuberculin in the Mantoux test Tuber Lung Dis 1992 73 326 329 1292711 10.1016/0962-8479(92)90035-I
Miller CL Morris J Pollock TM PHLS inquiry into current BCG vaccination policy BMJ 1984 288 564 6421378
Tissot F Zanetti G Francioli P Zellweger JP Zysset F Influence of bacille Calmette-Guerin vaccination on size of tuberculin skin test reaction: to what size? Clin Infect Dis 2005 40 211 217 15655737 10.1086/426434
Tala-Heikkila M Nurmela T Misljenovic O Bleiker MA Tala E Sensitivity to PPD tuberculin and M. scrofulaceum sensitin in schoolchildren BCG vaccinated at birth Tuber Lung Dis 1992 73 87 93 1643303 10.1016/0962-8479(92)90061-N
Stanford JL Cunningham F Pilkington A Sargeant I Series H Bhatti N Bennett E Mehrotra ML A prospective study of BCG given to young children in Agra, India- a region of high contact with enviromental mycobacteria Tubercle 1987 68 39 49 3660461 10.1016/0041-3879(87)90006-7
Lao LY De Guia T Tuberculin skin testing: determinants and reaction Respirology 1999 4 311 317 10489682 10.1046/j.1440-1843.1999.00198.x
Lifschitz M Value of the tuberculin skin test as a screening test for tuberculosis among BCG vaccinated children Pediatrics 1965 36 624 627 5294418
Nemir RL Teichner A Management of tuberculin reactors in children and adolescents previously vaccinated with BCG Pediatr Infect Dis 1983 2 446 451 6657499
Johnson H Lee B Doherty E Kelly E McDonnell T Tuberculin sensitivity and the BCG scar in tuberculosis contacts Tuber Lung Dis 1995 76 122 125 7780093 10.1016/0962-8479(95)90553-7
Johnson PD Stuart RL Grayson ML Olden D Clancy A Ravn P Andersen P Britton WJ Rothel JS Tuberculin-purified protein derivative-, MPT-64-, and ESAT-6-stimulated gamma interferon responses in medical students before and after mycobacterium bovis BCG vaccination and in patients with tuberculosis Clin Diagn Lab Immunol 1999 6 934 937 10548589
Medical Research Council Tuberculosis and Chest Diseases Unit Tuberculosis in children in a national survey of notifications in England and Wales, 1978–1979 Arch Dis Child 1982 57 734 741 6982683
Comstock GW Does the protective effect of neonatal BCG correlate with vaccine-induced tuberculin reaction? Am J Respir Crit Care Med 1996 154 263 264 8680692
Miret-Cuadras P Pina-Gutierrez JM Juncosa S Tuberculin reactivity in Bacillus Calmette-Guerin vaccinated subjects Tuber Lung Dis 1996 77 52 58 8733415 10.1016/S0962-8479(96)90076-3
Medical Research Council Tuberculosis and Chest Diseases Unit National survey of notifications of tuberculosis in England and Wales in 1983 BMJ 1985 291 658 661 3928070
Brandt L Elhay M Rosenkrands I Lindblad EB Andersen P ESAT-6 Subunit Vaccination against Mycobacterium tuberculosis Infect Immun 2000 68 791 795 10639447 10.1128/IAI.68.2.791-795.2000
Hess J Kaufmann SH Development of novel tuberculosis vaccines C R Acad Sci III 1999 322 953 958 10646089
| 15804351 | PMC1087500 | CC BY | 2021-01-04 16:37:45 | no | J Immune Based Ther Vaccines. 2005 Apr 1; 3:1 | utf-8 | J Immune Based Ther Vaccines | 2,005 | 10.1186/1476-8518-3-1 | oa_comm |
==== Front
Microb Cell FactMicrobial Cell Factories1475-2859BioMed Central London 1475-2859-4-91580434710.1186/1475-2859-4-9ResearchTransient increase of ATP as a response to temperature up-shift in Escherichia coli Soini Jaakko [email protected] Christina [email protected] Christina [email protected]öhm Daniela [email protected] Stefan [email protected] Johanna [email protected] Antti [email protected] Peter [email protected] Bioprocess Engineering Laboratory and Biocenter Oulu, Department of Process and Environmental Engineering, University of Oulu, P.O.Box 4300, FI – 90014 Oulu, Finland2 Institute for Biotechnology, Department of Biochemistry/Biotechnology, Martin-Luther-University Halle-Wittenberg, Kurt-Mothes-Str. 3, D-06120 Halle, Germany2005 1 4 2005 4 9 9 4 2 2005 1 4 2005 Copyright © 2005 Soini 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
Escherichia coli induces the heat shock response to a temperature up-shift which is connected to the synthesis of a characteristic set of proteins, including ATP dependent chaperones and proteases. Therefore the balance of the nucleotide pool is important for the adaptation and continuous function of the cell. Whereas it has been observed in eukaryotic cells, that the ATP level immediately decreased after the temperature shift, no data are available for E. coli about the adenosine nucleotide levels during the narrow time range of minutes after a temperature up-shift.
Results
The current study shows that a temperature up-shift is followed by a very fast significant transient increase of the cellular ATP concentration within the first minutes. This increase is connected to a longer lasting elevation of the cellular respiration and glucose uptake. Also the mRNA level of typical heat shock genes increases within only one minute after the heat-shock.
Conclusion
The presented data prove the very fast response of E. coli to a heat-shock and that the initial response includes the increase of the ATP pool which is important to fulfil the need of the cell for new syntheses, as well as for the function of chaperones and proteases.
==== Body
Background
Bacteria, as all kind of organisms, respond to a sudden increase of the temperature by a heat shock response which induces a specific set of proteins. This stress system belongs to the best-studied cellular responses. As the protein composition of a cell determines substantially the cellular functions, the heat shock response is mainly understood in terms of protein synthesis and composition.
The heat shock response of E. coli is mediated by the alternative sigma factor σ32 (rpoH gene product) and leads to induction of more than 20 heat shock proteins [1,2]. In difference to other stress responses of E. coli which in a regulatory cascade first lead to synthesis of the respective sigma factors which then, consecutively, direct the transcription of their genes, the transcriptional induction of the heat shock proteins can start very fast after a heat shock. This first fast synthesis of heat shock proteins is transient, followed by an adaptation period with a lower rate of heat shock protein synthesis and later reaches a new steady-state level [1,3]. This is because a cell also under non-stress conditions contains a set of a low number of σ32 molecules (about 10–30 molecules; [4,5]). Those molecules although possessing a high binding affinity for the RNA polymerase [6], however, are rarely transcriptionally active, because of sequestering by binding to DnaK, which together with its cochaperones DnaJ and GrpE associates with σ32 [7,8] and mediates its degradation by the FtsH protease [9]. The release of σ32 from the DnaK chaperone is regulated by the amount of missfolded proteins which may accumulate either when existing proteins lose their structure, or when new-synthesised proteins do not fold properly, which is the characteristic intracellular signal of a heat shock [10] but also may occur under strong overexpression of recombinant proteins [11]. Once released from DnaK, σ32 binds to RNA polymerase, is stabilised and initiates the synthesis of the heat shock proteins. This mechanism due to the high affinity of σ32 for binding to RNA polymerase results in the synthesis of heat shock proteins to a concentration of about 5 % of the total cellular protein [1].
The physiological response of E. coli towards a heat shock is more complex than just the synthesis of heat shock proteins, including transient growth stop, changes in the cell membranes due to changes in the ratio of lipids and integral membrane proteins [12], and probably also transient DNA relaxation [13]. In higher eukaryotes, one of the most immediate effects of a heat shock is an extensive disruption of the cytoskeleton [14,15]. From the metabolic side already early studies have shown that the maintenance energy requirement increases with increasing growth temperature [16-19] and the yield coefficient YX/S was observed to decrease, possibly in connection to the higher protein turnover [20]. A number of authors also reported increased respiration rates in relation to the specific growth rate at higher temperatures, e.g. for A. aerogenes (Klebsiella pneumoniae) [21]. (referred by [22])
The group of more than 30 heat shock proteins involves mainly proteases and chaperones, being involved in the degradation of improper folded proteins. The function of these proteins is connected to the hydrolysis of ATP, such as proteolysis by Lon, FtsH and Clp in E. coli [23-25], and the action of the chaperone complexes GroEL/ES and DnaK/J/GrpE [26,27]. Although a number of in vitro studies have been performed revealing this ATP dependence (e.g. [28]), only limited information is available about the response of the in vivo ATP level after a temperature shift. Different authors (e.g. [29-31]) found in higher eukaryotes a fast decrease of the ATP level to less than 50 % after a temperature up-shift parallel to induction of heat shock proteins.
In E. coli the ATP concentration has been mainly followed in steady state conditions. For instance it was reported to increase several fold in correlation with the specific growth rate [32,33]. In response to stress ATP or the total adenylate concentration have been followed after cold shock, where both decrease [34], after UV radiation, where the ATP level transiently increased twofold [35], with a following decrease in recA+ cells, which the authors relate to the action of substrate level phosphorylation. Also the levels of ATP and other nucleotides have been followed after induction of recombinant proteins by IPTG [36].
Studies with an appropriate short-time resolution of the ATP response in heat-shocked E. coli cells have not been published, although those data might be valuable to better understand the in vivo function of heat shock induced chaperones and proteases.
Results
In response to a heat shock E. coli strongly induces the production of a number of proteins with protecting and degrading functions, which are ATP dependent. However, most of the data on protease and chaperone function were obtained by in vitro studies, and there are still nearly no data available indicating, that there is a sufficient level of ATP available after a heat shock. Rather confusing results from eukaryotes suggest, that the ATP level and the energy charge after heat shock quickly decrease, which could be seen as contradictory to the optimum conditions for the function of these proteins. Therefore it was the aim of this study to measure the response of the pool of adenosine nucleotides following a temperature up-shift in E. coli.
Aside from the measurement of the adenylate phosphorylation status we wanted to discuss the results in relation to the cellular metabolism and respiratory activity. Therefore the experiments were performed in a bioreactor, which allowed the measurement of oxygen consumption and carbon dioxide formation, as well as a fast sampling of nucleotides and medium samples.
All cultivations were performed with the E. coli K-12 strain W3110, which was a major strain in the evaluation of the E. coli stress responses. As indicated in the materials and methods section the cultivations were carried out on mineral salt medium with glucose as carbon source to avoid complex reactions in view of amino acid availability. The medium used was originally designed for high cell density fermentations, and therefore insures, that all salts are maintained at appropriate concentrations up to the end of the experiment.
The experiments were performed with a temperature shift from 30°C to 42°C. In this range the viability of the cell is not significantly affected, although the heat shock response is strongly induced. The basic growth behaviour and parameters are shown in Fig. 1 and Table 1, respectively. Cultivations started with a batch phase at 30°C for about 11 hours, corresponding to 6 generations, to ensure a steady state of cellular parameters. The specific growth rate μ was determined to be 0.35 h-1. By following the glucose and acetate concentrations (see Fig. 1) the specific rate for glucose uptake (qS) was determined to be 5.1 mmol g-1 h-1 and the specific acetate production rate (qA) was 2.5 mmol g-1 h-1. The temperature up-shift to 42°C was performed at a cell density of OD500 = 3 (corresponding to a dry cell weight of 0.675 g L-1) within 3 min. After the temperature shift the cultivation was continued for one hour. Rapidly after the temperature shift the cells start to increase the specific growth rate to 0.70 h-1 and correspondingly the rates for glucose uptake and acetate production (see Table 1).
During growth at 30°C the ATP concentration was detected to be about 1.5 μmol per OD500. Concentrations of the other lower phosphorylated adenosine nucleotides were very low during this phase and the resulting adenylate energy charge (EC) was above 0.95. Nucleotide samples were collected in five minute intervals. The ATP concentration increased to 2.75 μmol per OD500 immediately after the temperature shift to 42°C and also the ADP concentration was found to transiently increase significantly (see Fig. 2). After the transient increase the cellular concentrations of both nucleotides decreased and finally we detected a slight increase of AMP towards the end of the cultivation. Interestingly, the elevation of the ATP level was connected to a total increase of all adenosine phosphates, resulting in a slightly lowered EC to 0.8 (see Fig. 3). However, similar to ATP, also the sum of adenosine phosphates increased only transiently, approaching to the pre-shift level about 20 min after the temperature up-shift. The slight increase of AMP to the end of the cultivation influenced the energy charge, which decreased to 0.65. The transient increase of ATP after the temperature up-shift is connected to a high respiratory activity (Fig. 1C) and occurs in the phase, where also the flux of glucose is rapidly increasing.
In alternate series of experiments the temperature was shifted either from 37°C to 48°C, or from 35°C to 55°C which leads to growth inhibition and cell death. In both cases the temperature up-shift caused a strong transient increase of respiration and, at least in the first case we also found the rapid increase of ATP, however, the following reduction of the ATP level was much faster (data not shown).
Further it was investigated how the time kinetics of the adenosine nucleotides are related to the expression profiles of typical heat shock genes. These experiments were performed in shake flasks with on-line monitoring of dissolved oxygen and temperature. A fast temperature up-shift was reached by diluting the culture with pre-heated medium which explains the decrease of OD580 in Fig. 4. Interestingly, the induction of the three investigated heat shock controlled genes was different. The expression of dnaK and ibp were both induced very rapidly and reached their maximum mRNA levels within only 1 min after the heat shock, but lon reached the maximum RNA level after 5 min. A quantitative evaluation of the mRNA amounts showed the highest concentration for ibp, but lower expression of dnaK and the lowest for lon. The level of dnaK mRNA was continuously significantly higher expressed than before the shift, but the mRNAs of ibp and lon were very low or not any more detectable after only 10 min. Interestingly, this shut-off of the expression of these genes is at about the same time when the pre-shift level of nucleotides was re-established.
Discussion
Our studies indicate, that E. coli responds to a temperature up-shift by rapidly increasing respiration and glucose uptake. The increase of glucose uptake is possibly mainly directed to glycolysis and, consequently, to energy production. Concomitantly we detected a significant drop in the pH and an increase of the acetate level. The higher production of acetate per glucose (YA/S) might point to an increase of glycolytic activity, and suggests that the flow to the tricarboxylic acid cycle was not increased correspondingly.
We detected a significant increase of the concentration of the sum of adenosine nucleotides (AXP) directly during the temperature up-shift which possibly is caused by RNA degradation as is concluded from the total RNA concentrations per cell which decreased to less than 50 % if the samples before and after the temperature up-shift are compared (data not shown). This increase of the AXP concentration was connected to a decrease in the energy charge, as the increase of ADP was higher than the increase of ATP. It remains to be investigated whether this relative transient increase of ADP has physiological effects. For example it might be hypothesized that the high concentration of ADP transiently inactivates various chaperones (see e.g.[37,38]) which bind ADP with higher affinity compared to ATP.
After the heat shock the cells strongly increase respiration and CO2 production which should be related to ain increased glycolytic flux, as correspondingly acetate formation is increased. A similar behaviour of increased glycolytic flux we have observed from other cultivations where a stress was caused by strong induction of a recombinant protein and also provoked a transient increase of the ATP level and higher respiration. However, differently from the investigated temperature up-shift, after induction of the recombinant protein the cells lost their viability [39]. Also a lethal heat shock caused by a temperature up-shift from 37°C to 52°C [40] or from 37°C to 48°C (own results, not shown) resulted in a transient increase of ATP, but this was much slower and the increase was only about 30 % with a maximum at about 30 min after the temperature shift.
Our data suggest that similarly as known for the heat shock response, that the response of the nucleotides is transient and the energy charge is restored 15 to 20 min after the heat shock. After this transient induction of the heat shock response cell growth recovers and it may be hypothesised that the cellular synthetic activities and metabolite pools reach a new steady state.
Conclusion
This study shows that the heat shock response in E. coli in connection to the adenosine phosphates is different from the response of eukaryotes. Whereas different authors described in eukaryotes an immediate decrease of the ATP level following a temperature up-shift (e.g. [29-31]), in E. coli we found a transient increase, before the ATP level is lowered. It may be hypothesised, but has to be verified by further experiments that this transient increase of the ATP concentration is important for the action of chaperones and proteases and consequently for the function of the cellular system towards a heat shock response.
Methods
Strains
All experiments were performed with the Escherichia coli K-12 strain W3110 [F-, IN(rrnD rrnE)1, λ-] which was kindly obtained by the E. coli Stock Center (New Haven, USA).
Cultivation medium and cultivation conditions
Phosphate buffered mineral salt medium according to Teich et al. [41] with 10 g L-1 glucose and thiamine hydrochloride (0.1 g L-1) was used in all cultivations. The pH of the medium was set to 7.0 before sterilisation by addition of NaOH. For fermentation experiments two precultivations were carried out at the appropriate temperature of the main cultivation. The first culture was performed in 10 mL of nutrient broth (NBII: 1 % tryptone, 0.6 % yeast extract), whereas the second preculture contained 200 ml of the fermentation medium in a 1000 mL baffled Erlenmeyer flask. Each preculture was used at the exponential growth phase as inoculum for the next cultivation.
The fermentations were performed in a 6 L Biostat ED bioreactor (B. Braun Biotech International, Germany) with a starting volume of 4 L of mineral salt medium pH 7.0). The pH was controlled to drop not below 7.0 by addition of 25 % ammonia. Air flow (0.002 to 2 vvm) and stirrer speed (200 to 800 rpm) were increased stepwise to keep the dissolved oxygen tension above 20 %. After the initial batch phase to an optical cell density (OD500) of 3 the temperature up-shift was performed with a temperature increase from 30 to 42°C within 6 min.
Batch cultivations with on-line monitoring of temperature and oxygen were performed in 1000 mL baffled Erlenmeyer flasks with three side necks for the sensors and a needle for sampling into a syringe. The cultures were performed in 200 mL M9 mineral salt medium with 10 g L-1 glucose, 2 mL L-1 of trace element solution and 0.1 g L-1 thiamine hydrochloride on a magnetic stirrer bar at 1000 rpm in a tempered water bath. The heat shock was performed by addition of 66 mL fresh M9 medium which was preincubated at 100°C and following replacement of the flask into another water bath (42°C). Temperature and oxygen were monitored by sensors (Pt1000, polarographic oxygen electrode from Meredos GmbH), connected to the wireless Senbit system (teleBITcom GmbH) with on-line data collection.
Analytical methods
Cellular growth
Growth of the cultures was followed by measuring the optical density at 500 nm (OD500, for cultivation in the bioreactor) or 540 nm (OD540, for shake flasks), and by determination of the cell dry weight (CDW). Therefore 1.5 mL cell suspensions were centrifuged in pre-weighed 2 mL eppendorf tubes and washed once with 0.9 % (w/v) NaCl solution. After removal of the supernatant the samples were dried to constancy at 60°C for at least 24 h. Colony forming units were analysed by plating at least three dilutions and each of these at triple of the culture broth on nutrient bullion plates, which were incubated for one to three days at 37°C.
Glucose and acetate
Samples from the cultivation were collected by direct filtration from the reactor through a 0.2 μm disk filter. Afterwards the samples were stored at -20°C for further analysis. The analysis of glucose was performed with the hexokinase/glucose-6-phosphate-dehydrogenase method (Kit no. 139106, Roche Diagnostics GmbH, Germany) scaled down to 96-well plates with four parallels for each sample as described by Larsson and Törnkvist [42]. Acetate was analysed with the Kit no. 148261 (Roche Diagnostics GmbH, Germany) correspondingly.
Outgas analysis
The outgas oxygen and carbon dioxide concentrations were measured with an URAS10E (Hartmann & Braun GmbH, Frankfurt, Germany). The specific rates (qO2, qCO2) were calculated continuously over the fermentation by including fitted biomass values.
Analysis of adenosine nucleotides
The adenosine nucleotides were analyzed with an RP-HPLC method as described earlier in detail Meyer et al. [43].
Analysis of mRNAs
10 mL samples of cell broth were immediately chilled with 1350 μL cold ethanol/phenol (95:5, v/v, pre-cooled at -20°C). After centrifugation at 10 000 rpm (4°C, 5 min) the supernatant was carefully removed, the pellet was resuspended in 1 mL of RNA Later (Ambion) and the samples were stored in 250 μL fractions at -70°C until analysis.
Total RNA extraction was performed with a total RNA isolation kit (A&A Biotechnology, Gdynia, Poland) according to the instructions of the manufacturer. Total RNA was quantified with the RiboGreen Quantification kit (Molecular probes). Quantitative analysis of dnaK, ibp and lon mRNAs was performed with a sandwich system earlier described in detail by Rautio et al. [44]. Therefore probes shown in Table 2 were designed using the CloneManager5 program with following submission of the sequences a NCBI BLAST search to exclude alignments with other E. coli genes.
HPLC-purified unlabelled and biotin-labelled oligonucleotide capture probes were purchased from Metabion GmbH (Martinsried, Germany). Dig-tail labelling of the detection probes was performed according to manufacturer's instructions using the Roche Dig-tailing kit (2nd generation, Roche Diagnostics GmbH, Mannheim, Germany).
In vitro RNA standards were designed for the quantitative analysis of each gene. Therefore primers as indicated in Table 3 were used for in vitro transcription (purchased from Sigma-Genosys, Cambridge, UK). Left primers (Primers 1) contained the 25 nucleotide T7 promoter sequence at the 5' side. Target genes were amplified by PCR from total DNA extracts of E. coli. The PCR products were purified using QIAquick PCR purification Kit and then used as templates for in vitro transcription using the MAXIscript T7 transcription kit from Ambion (Austin Texas, U.S.A.). The in vitro standards were quantified by the RiboGreen RNA Quantification Kit (Molecular probes) and a standard calibration curve with the respective standard was performed at the same plate where the samples were analysed.
Streptavidin MagneSphere Paramagnetic Particles (1 mg mL-1 magnetic particles in PBS, 1 mg mL-1 BSA and 0.02 % NaAzide) from Promega (Madison, USA) were used as magnetic beads for immobilisation of the target. Sandwich hybridisation was performed in 96-well plates at 50°C for 30 min with constant shaking (700 rpm). Therefore hybridisation solution with a final concentration of 5 × SSC, 20% formamide, 2% Denhardt reagent, 3% dextran sulphate and 0.2% SDS was added to a final volume of 100 μL to each sample. Each well of the plate contained 1 pmol of DIG-tail labelled detection probe, 5 pmol of biotin-labelled capture probe and appropriate dilutions of in vitro transcripts for the standard curve or of the RNA extracts from cultivation samples.
Immobilisation of RNA molecules was done by adding 20 μl streptavidin coated magnetic beads to each well after hybridisation. DIG labelled probes were bound to Anti-DIG-alkaline phosphatase (added 1:2000 in TBS-buffer) and the enzymatic reaction was carried out with AttoPhos (Promega) as substrate. Fluorescence was measured with the Wallac 2 fluorescence reader (Perkin Elmer Life Sciences) at 25°C using the Attophos1-program.
Authors' contributions
SW and PN carried out the fermentations and nucleotide analyses. JS, CF, CM and DB developed the method for analysing the mRNAs and performed the corresponding analysis. JP and AV were included in the wireless analysis with the Senbit system. JS, PN and AV drafted the manuscript. All authors read and approved the final manuscript.
Acknowledgements
The study was partly supported by the EU project no. QLK3-CT-1999-00533 "Electric DNA chips for bioprocess control" and the TEKES Neobio program (CORF project).
Figures and Tables
Figure 1 Batch cultivation of E. coli W3110 with a temperature up-shift performed at time 0 by switching the temperature set-point from 30 to 42°C. (A) OD500 (○), temperature (—); (B) glucose (□), acetate (△); (C) qO2 (○), qCO2 (▽), RQ (□), (D) pH (△), relative units of ammonia added (+). The grey area indicates the time period of about 5 min during which the temperature increased.
Figure 2 Response of the adenosine nucleotide pool to a temperature up-shift from 30 to 42°C in E. coli W3110. (A) ATP and cultivation temperature, (B) ADP, (C) AMP. The data were obtained during the batch cultivation shown in Figure 1.
Figure 3 Dynamics of the Energy Charge and sum of adenosine phosphates (AXP) during a cultivation of E. coli W3110 with temperature up-shift from 30 to 42°C. Data were calculated from primary concentrations of the adenosine nucleotides shown in Figure 2.
Figure 4 Response of mRNAs in a stirred-flask cultivation of E. coli W3110 with a temperature shock from 30°C to 42°C. For experimental details see Materials and Methods.
Table 1 Cultivation parameters of parameters from batch cultivations of E. coli W3110 at 30°C and after a temperature shift to 42°C. The data are average values from two independent fermentations.
Parameter before temperature shift after temperature shift
μ [h-1] 0.38 0.71
qS [mmol g-1 h-1] 5.1 17.4
YX/S [g mol -1] 75.6 41.4
qA [mmol g-1 h-1] 2.5 8.0
YA/S [mol mol-1] 0.30 0.51
YCO2/S [mol mol-1] 1.86 0.80
YO/S [mol mol-1] 1.76 0.85
Table 2 Probes used in Sandwich hybridization Capture probes were biotin labelled and detection probes contained a Dig tail.
Name
Sequence (5'-3')
Position
dnaK
Helper probe 1 GTCGGGATAGTGGTGTTTTT 1259c
Capture probe AGAACACCTGGCTGTGCTTG 1279c
Helper probe 2 CTGGTTGTCTTCAGCGGTAG 1299c
Detection probe TTGTCGCCTGCTTCTTCAAC 1667c
Ibp
Capture probe ACCAGCCACAGCGATAGCAA 165c (ibpA)
Detection probe TTCCAGTTCGCTCTCAGCAA 186c (ibpA)
Helper probe ACCACCAGCAGATTATCCTG 215c (ibpA)
Lon
Capture probe ACTCCAGATACTCCGCCTTC 355c
Detection probe GCCTGAATGGACTCCTGCAT 1919c
Table 3 Primers for synthesis of in vitro RNA standards. The T7 promoter sequence was added to the 5' side of all primer 1 sequences.
Name
Sequence (5'-3')
Position
dnaK
Primer 1 CTCTTGTGTAGCGATTATGG 39
Primer 2 GTTGTTTGCAGAAGCATCAG 1860c
Ibp
Primer 1 CCGCTTTACCGTTCTGCTA 22 (ibpA)
Primer 2 CTATTTAACGCGGGACGTTC 425c (ibpB)
Lon
Primer 1 ATGAATCCTGAGCGTTCTGA 1
Primer 2 CACAACCTGCATACCAGAC 2342c
T7 promoter CTAATACGACTCACTATAGGGAGA
==== Refs
Arsene F Tomoyasu T Bukau B The heat shock response of Escherichia coli Int J Food Microbiol 2000 55 3 9 10791710 10.1016/S0168-1605(00)00206-3
Gross CA Neidhardt FC, Curtiss III R, Ingraham JL, Lin ECC, Low KB, Magasanik B Function and regulation of the heat shock proteins Escherichia coli and Salmonella typhimurium Cellular and Molecular Biology 1996 Washington, D.C.: American Society of Microbiology 1382 1399
Bukau B Regulation of the Escherichia coli heat-shock response Mol Microbiol 1993 9 671 680 7901731
Craig EA Gross CA Is hsp70 the cellular thermometer? Trends Biochem Sci 1991 16 135 140 1877088 10.1016/0968-0004(91)90055-Z
Ishihama A Functional modulation of Escherichia coli RNA polymerase Annu Rev Microbiol 2000 54 499 518 11018136 10.1146/annurev.micro.54.1.499
Arsene F Tomoyasu T Mogk A Schirra C Schulze-Specking A Bukau B Role of region C in regulation of the heat shock gene-specific sigma factor of Escherichia coli, sigma32 J Bacteriol 1999 181 3552 3561 10348869
Gamer J Bujard H Bukau B Physical interaction between heat shock proteins DnaK, DnaJ, and GrpE and the bacterial heat shock transcription factor sigma 32 Cell 1992 69 833 842 1534276 10.1016/0092-8674(92)90294-M
Gamer J Multhaup G Tomoyasu T McCarty JS Rudiger S Schonfeld HJ A cycle of binding and release of the DnaK, DnaJ and GrpE chaperones regulates activity of the Escherichia coli heat shock transcription factor sigma32 EMBO J 1996 15 607 617 8599944
Tomoyasu T Gamer J Bukau B Kanemori M Mori H Rutman AJ Escherichia coli FtsH is a membrane-bound, ATP-dependent protease which degrades the heat-shock transcription factor sigma 32 EMBO J 1995 14 2551 2560 7781608
Tomoyasu T Ogura T Tatsuta T Bukau B Levels of DnaK and DnaJ provide tight control of heat shock gene expression and protein repair in Escherichia coli Mol Microbiol 1998 30 567 581 9822822 10.1046/j.1365-2958.1998.01090.x
Hoffmann F Rinas U Roles of heat-shock chaperones in the production of recombinant proteins in Escherichia coli Adv Biochem Eng Biotechnol 2004 89 143 161 15217158
Mejia R Gomez-Eichelmann MC Fernandez MS Fatty acid profile of Escherichia coli during the heat-shock response Biochem Mol Biol Int 1999 47 835 844 10365254
Lopez-Garcia P Forterre P DNA topology and the thermal stress response, a tale from mesophiles and hyperthermophiles Bioessays 2000 22 738 746 10918304 10.1002/1521-1878(200008)22:8<738::AID-BIES7>3.0.CO;2-5
Welch WJ Suhan JP Morphological study of the mammalian stress response: characterization of changes in cytoplasmic organelles, cytoskeleton, and nucleoli, and appearance of intranuclear actin filaments in rat fibroblasts after heat-shock treatment J Cell Biol 1985 101 1198 1211 3900086 10.1083/jcb.101.4.1198
van Bergen en Henegouwen PM Jordi WJ van Dongen G Ramaekers FC Amesz H Linnemans WA Studies on a possible relationship between alterations in the cytoskeleton and induction of heat shock protein synthesis in mammalian cells Int J Hyperthermia 1985 1 69 83 2426373
Esener AA Roels JA Kossen NWF The influence of temperature on the energetics of Klebsiella pneumoniae Biotechnol Bioeng 1983 25 2093 2098 10.1002/bit.260250819
Harder W Veldkamp H A continuous culture study of an obligately psychrophilic Pseudomonas species Arch Microbiol 1967 59 123 130
Pennock J Tempest DW Metabolic and energetic aspects of the growth of Bacillus stearothermophilus in glucose-limited and glucose-sufficient chemostat culture Arch Microbiol 1988 150 452 459 10.1007/BF00422286
Topiwala H Sinclair CG Temperature relationship in continuous culture Biotechnol Bioeng 1971 13 795 813 4944360
Dawes EA Ribbons DW Some aspects of the endogenous metabolism of bacteria Bacteriol Rev 1964 28 126 149 14174839
Senez JC Some considerations on the energetics of bacterial growth Bacterial Rev 2005 26 95 107
Heitzer A Kinetic and physiological aspects of bacterial growth at superoptimum temperatures Diss ETH no 9217, Zürich, Switzerland 1990
Porankiewicz J Wang J Clarke AK New insights into the ATP-dependent Clp protease: Escherichia coli and beyond Mol Microbiol 1999 32 449 458 10320569 10.1046/j.1365-2958.1999.01357.x
Schumann W FtsH – a single-chain charonin? FEMS Microbiol Rev 1999 23 1 11 10077851 10.1016/S0168-6445(98)00024-2
Suzuki CK Rep M van Dijl JM Suda K Grivell LA Schatz G ATP-dependent proteases that also chaperone protein biogenesis Trends Biochem Sci 1997 22 118 123 9149530 10.1016/S0968-0004(97)01020-7
Horwich AL Weber-Ban EU Finley D Chaperone rings in protein folding and degradation Proc Natl Acad Sci U S A 1999 96 11033 11040 10500119 10.1073/pnas.96.20.11033
Braig K Chaperonins Curr Opin Struct Biol 1998 8 159 165 9631288 10.1016/S0959-440X(98)80033-X
Sparrer H Lilie H Buchner J Dynamics of the GroEL-protein complex: effects of nucleotides and folding mutants J Mol Biol 1996 258 74 87 8613994 10.1006/jmbi.1996.0235
Lilly MB Ng TC Evanochko WT Katholi CR Kumar NG Elgavish GA Loss of high-energy phosphate following hyperthermia demonstrated by in vivo 31P-nuclear magnetic resonance spectroscopy Cancer Res 1984 44 633 638 6581861
Findly RC Gillies RJ Shulman RG In vivo phosphorus-31 nuclear magnetic resonance reveals lowered ATP during heat shock of Tetrahymena Science 1983 219 1223 1225 6828852
Jones KA Findly RC Induction of heat shock proteins by canavanine in Tetrahymena. No change in ATP levels measured in vivo by NMR J Biol Chem 1986 261 8703 8707 3722169
Gaal T Bartlett MS Ross W Turnbough CL JrGourse RL Transcription regulation by initiating NTP concentration: rRNA synthesis in bacteria [see comments] Science 1997 278 2092 2097 9405339 10.1126/science.278.5346.2092
Petersen C Moller LB Invariance of the nucleoside triphosphate pools of Escherichia coli with growth rate J Biol Chem 2000 275 3931 3935 10660546 10.1074/jbc.275.6.3931
Rhodes MW Anderson IC Kator HI In situ development of sublethal stress in Escherichia coli: effects on enumeration Appl Environ Microbiol 1983 45 1870 1876 6349527
Barbe J Villaverde A Guerrero R Evolution of cellular ATP concentration after UV-mediated induction of SOS system in Escherichia coli Biochem Biophys Res Commun 1983 117 556 561 6362671 10.1016/0006-291X(83)91236-6
Lin HY Hanschke R Nicklisch S Nietsche T Jarchow R Schwahn C Merten O-W, Mattanovich D, LANG C, Larsson G, Neubauer P, Porro D Cellular responses to strong overexpression of recombinant genes in Escherichia coli DNA relaxation and cell death after induction of α-glucosidase Recombinant Protein Production with prokaryotic and eukaryotic cells A comparative view on host physiology 2001 Dortrecht, The Netherlands: Kluwer Academic Publisher 55 74
Mayer MP Bukau B Hsp70 chaperones: Cellular functions and molecular mechanism Cell Mol Life Sci 2005 62 670 684 15770419 10.1007/s00018-004-4464-6
Zolkiewski M Kessel M Ginsburg A Maurizi MR Nucleotide-dependent oligomerization of ClpB from Escherichia coli Protein Sci 1999 8 1899 1903 10493591
Neubauer P Lin HY Mathiszik B Metabolic load of recombinant protein production: inhibition of cellular capacities for glucose uptake and respiration after induction of a heterologous gene in Escherichia coli Biotechnol Bioeng 2003 83 53 64 12740933 10.1002/bit.10645
Tkachenko AG Pshenichnov MR Salakhetdinova OI Nesterova LI [Role of putrescine and potassium transport in regulating the topological state of DNA during adaptation of Escherichia coli to temperature stress] Mikrobiologiia 1998 67 601 606 9891292
Teich A Lin HY Andersson L Meyer S Neubauer P Amplification of ColE1 related plasmids in recombinant cultures of Escherichia coli after IPTG induction J Biotechnol 1998 64 197 210 9821676 10.1016/S0168-1656(98)00108-4
Larsson G Törnkvist M Rapid sampling, cell inactivation and evaluation of low extracellular glucose concentrations during fed-batch cultivation J Biotechnol 1996 49 69 82 8879165 10.1016/0168-1656(96)01534-9
Meyer S Noisommit-Rizzi N Reuss M Neubauer P Optimized analysis of intracellular adenosine and guanosine phosphates in Escherichia coli Anal Biochem 1999 271 43 52 10361003 10.1006/abio.1999.4119
Rautio J Barken KB Lahdenpera J Breitenstein A Molin S Neubauer P Sandwich hybridisation assay for quantitative detection of yeast RNAs in crude cell lysates Microb Cell Fact 2003 2 4 12780940 10.1186/1475-2859-2-4
| 15804347 | PMC1087501 | CC BY | 2021-01-04 16:05:47 | no | Microb Cell Fact. 2005 Apr 1; 4:9 | utf-8 | Microb Cell Fact | 2,005 | 10.1186/1475-2859-4-9 | oa_comm |
==== Front
Nutr JNutrition Journal1475-2891BioMed Central London 1475-2891-4-141582631310.1186/1475-2891-4-14ReviewWhy nutraceuticals do not prevent or treat Alzheimer's disease Fisher Anna EO [email protected] Declan P [email protected] School of Pharmacy and Biomolecular Sciences, University of Brighton, Cockcroft Building, Moulsecoomb, Brighton, UK2005 12 4 2005 4 14 14 23 11 2004 12 4 2005 Copyright © 2005 Fisher and Naughton; licensee BioMed Central Ltd.2005Fisher and Naughton; 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.
A great deal of research has pointed to deleterious roles of metal ions in the development of Alzheimer's disease. These include: i) the precipitation and aggregation of amyloid β (Aβ) peptides to form senile plaques and neurofibrillary tangles, and/or ii) the augmentation of oxidative stress by metal ion mediated production and activation of hydrogen peroxide. The growing trend in nutraceutical intake is in part a result of the belief that they postpone the development of dementias such as Alzheimer's disease. However, pathogenic events centred on metal ions are expected to be aggravated by frequent nutraceutical intake. Novel therapeutic approaches centred on chelators with specificity for copper and iron ions should be fully explored.
==== Body
Introduction
Dementia currently affects over 750,000 people in the UK, with Alzheimer's disease (AD) accounting for 55 % of those cases. As a result of increasing life expectancy, the number of people predicted to suffer from dementia is expected to rise to over 1.8 million by 2050. Currently there are no known cures for AD. Existing treatments at best only delay the progression of disease and accurate diagnosis can only be made with certainty at autopsy. Hence, there is an urgent need for new therapeutic approaches and for reliable and noninvasive methods to aid in the early diagnosis of AD and to give an indication of disease progression. Advanced AD is evident from the apparent progressive decline in cognitive function as a result of loss of neurons. These are accompanied by pathological changes including extracellular amyloid β plaques and intracellular changes called neurofibrillary tangles, which are abnormal deposits of hyperphosphorylated tau proteins [1].
Recent reports ascribe a major role for the metal ions, iron, copper and zinc in AD. Specifically the redox active metal ions Fe2+ and Cu+ can generate the highly reactive and oxidizing species, the hydroxyl radical via the Fenton reaction. This species is thought to account for a large proportion of oxidative modifications that are seen in AD brains including lipid peroxidation, nucleic acid oxidation and protein carbonylation. Also metal ions have been shown to be central to Aβ pathogenicity.
The Aβ peptide is a low molecular weight protein consisting of 39–43 amino acids. It has two high affinity binding sites, one specific for copper and one specific for zinc. In the Aβ1–42 peptide (the predominant Aβ species found in amyloid plaques in AD brains) the high affinity binding site binds Cu2+ were estimated at log K = 17.2 and 16.3 at pH 7.4 and pH 6.6 respectively, and the low affinity binding site at log K = 8.3 and 7.5 at pH 7.4 and 6.6 respectively [2]. At physiological pH values, soluble Aβ is precipitated by Zn2+ in vitro. However, at pH levels reflecting physiological acidosis, which is characteristic of AD, in vitro precipitation of Aβ is induced by cupric and ferric ions. Cupric ions induce greater precipitation than ferric ions, which is thought to be a result of its higher binding affinity, with Aβ aggregation enhanced to ~85 % [3]. In addition, Cu2+ ions are bound at the low affinity site completely displacing the Zn2+ ions in physiological acidosis [2]. Levels of copper, iron and zinc have been analyzed in the rims and cores of senile plaques of AD patients and concentrations were found to be highly elevated. Concentrations exhibited in the senile plaques cores were 22.7 ± 6.5, 52.4 ± 14.5 and 67.0 ± 13.0 μg/g for copper, iron and zinc respectively [4]. In addition serum copper levels have been shown to be markedly elevated in AD [5].
The catalytic effects of transition metals on the formation of reactive oxygen species are becoming evident as a major factor in AD. Aβ binds to Cu2+ reducing it to Cu+. The Aβ-Cu+ complex can then trap dioxygen and reduce it to hydrogen peroxide (H2O2). One study demonstrated that 1 μM Cu2+ when bound to Aβ could catalytically generate 10 μM H2O2, indicating continual redox cycling [6]. In conditions where this reaction is not favorable, biological reducing agents such as vitamin C have been shown to enhance this reaction [7]. The generation of H2O2 has been shown to be neurotoxic, which is considered to be a result of its reduction to the hydroxyl radical via Fenton chemistry. Cupric ion binding is dependent on histidine and tyrosine amino acid residues. Upon generation of hydroxyl radicals localized to the Cu2+ ion, the tyrosine residues can cross-link to form dityrosine. This further stabilizes the Aβ plaques conferring resistance to proteolysis [8]. Further studies are required to establish if Aβ bound to Cu2+ and Zn2+ exhibits significant anti-oxidant enzyme activities (superoxide dismutase and catalase activities) in vivo, as many researchers have likened Aβ to the native CuZn superoxide dismutase enzyme in both its conformation and its metal binding affinities [9].
Magnetite and maghemite deposits have been found in brains of patients suffering from AD and studies are underway to correlate enhanced levels with disease progression [10]. However, magnetite detection is only possible after the onset of AD. A comprehensive understanding of the levels and nature of redox-active ferric ions is warranted to assess the mechanisms of magnetite deposition. AD brains exhibit increased levels of oxidatively modified nucleic acids, specifically RNA in vulnerable neurons [11]. Haem oxygenase-1 (HO-1), an enzyme which catalyses the breakdown of haem to biliverdin with release of iron and carbon monoxide, is over-expressed in the brains of AD patients. So it could be postulated that this would be one source of redox-active iron. Immunostaining of neurons demonstrated that HO-1 expression was co-localized to the neurofibrillary tangles [12]. However oxidative damage to nucleic acids was reduced in neurons exhibiting neurofibrillary tangles demonstrating an antioxidant mechanism for HO-1 and an involvement in tau proteins expression. Further studies on the oxidative modifications of RNA in AD demonstrated it to be localized to iron rich lysosomes called lipofuscins originating from the mitochondria. This indicates a disruption in mitochondrial function leading to accumulation of redox-active iron in the neuronal cytoplasm [13] and subsequent reactive oxygen species generation.
Therapeutic Implications
Oxidative stress is one of the key mechanisms of neurotoxicity in AD, and studies have shown that Aβ neurotoxicity can be partially or completely attenuated by the antioxidant enzyme catalase [6,7]. Plasma levels of antioxidants including vitamin E, uric acid and vitamin C are depleted in AD patients [14]. Therefore, many people have proposed the supplementation of antioxidants such as vitamin E and vitamin C. Several studies have reported that increased intake either through diet or supplementation results in reduced incidence of AD [15-17]. However, other studies do not report such success [18]. In fact, some studies have shown that vitamin C which can also act as a pro-oxidant can also induce neuronal oxidative stress via its interaction with metal ions [19].
A long-term imbalance in metal ion homeostasis is rarely treatable by alterations in diet. Therefore, research is focusing on the development of new methods for therapeutic intervention, specifically chelators to remove the deleterious metal ions, with concomitant dissolution of the Aβ plaques. In vitro studies on the aggregation of Aβ induced by metal ions, showed that this was completely reversible by metal ion chelation [3]. As a result of these observations new chelators are being developed as novel therapies for AD, however many are restricted by their poor target specificity. Clioquinol a Cu/Zn chelator was tested in APP2576 transgenic mice and appeared to selectively bind the Aβ-metal complex resulting in dissolution of Aβ plaques, reducing brain Aβ by 49 % [20]. No neurotoxicity was exhibited. Clioquinol subsequently was tested in a clinical trial with 36 AD patients. It affected a rise in plasma concentrations of zinc, and also decreased plasma concentrations of Aβ 1–42 [21]. However more studies are required to determine its importance in treating AD. Many other metal ion chelators are currently under investigation for treatment of AD [22-25].
We have developed a series of metal ion chelators for both the study and treatment of metal deposition disorders. These include the development of the first-ever, reactive oxygen and nitrogen species probe for the simultaneous detection of multiple RONS and the role of redox-active metal ions in their generation and suppression [26]. Alteration of the basic structure of these probes affords anti-oxidant enzyme mimics [27]. In addition the chelator EGTA, which has previously been shown to dissolve Aβ plaques [28] was also demonstrated to have substantial superoxide dismutase activity when complexed to redox-active metal ions [29]. Thus, addition of a chelator that generates a superoxide dismutase or catalase mimic could 'dissolve' the magnetite deposits and Aβ plaques and negate oxidative damage arising from them.
Conclusion
Metal specific chelators afford a major opportunity for the control of metal ion mediated pathogeneses, which is not currently provided by nutraceuticals. In particular, administration of "prodrug" chelators that mimic anti-oxidant enzymes with high specificity for Cu2+ and Fe3+ ions will: i) counteract oxidative stress, ii) help to control metal ion deposition and iii) dissolve Aβ plaques and, maghemite and magnetite residues.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
The authors contributed equally to this work.
Acknowledgements
We are grateful to the University of Brighton and the Engineering and Physical Sciences Research Council for financial support.
==== Refs
Luchsinger JA Mayeux R Dietary factors and Alzheimer's disease Lancet Neurol 2004 3 579 587 15380154 10.1016/S1474-4422(04)00878-6
Atwood CS Scarpa RC Huang X Moir RD Jones WD Fairlie DP Tanzi RE Bush AI Characterization of copper interactions with Alzheimer amyloid β peptides: identification of an attomolar-affinity copper binding site on amyloid β1–42 J Neurochem 2000 75 1219 1233 10936205 10.1046/j.1471-4159.2000.0751219.x
Atwood CS Moir RD Huang X Scarpa RC Bacarra NM Romano DM Hartshorn MA Tanzi RE Bush AI Dramatic Aggregation of Alzheimer Abeta by Cu(II) Is Induced by Conditions Representing Physiological Acidosis J Biol Chem 1998 273 12817 12826 9582309 10.1074/jbc.273.21.12817
Lovell MA Robertson JD Teesdale WJ Campbell JL Markesbery WR Copper, iron and zinc in Alzheimer's disease senile plaques J Neurol Sci 1998 158 47 52 9667777 10.1016/S0022-510X(98)00092-6
Squitti R Lupoi D Pasqualetti P Dal Forno G Vernieri F Chiovenda P Rossi L Cortesi M Cassetta E Rossini PM Elevation of serum copper levels in Alzheimer's disease Neurology 2002 59 1153 1161 12391342
Huang X Cuajungco MP Atwood CS Hartshorn MA Tyndall JD Hanson GR Stokes KC Leopold M Multhaup G Goldstein LE Cu(II) potentiation of alzheimer Abeta neurotoxicity. Correlation with cell-free hydrogen peroxide production and metal reduction J Biol Chem 1999 274 37111 37116 10601271 10.1074/jbc.274.52.37111
Opazo C Huang X Cherny RA Moir RD Roher AE White AR Cappai R Masters CL Tanzi RE Inestrosa NC Metalloenzyme-like activity of Alzheimer's disease beta-amyloid. Cu-dependent catalytic conversion of dopamine, cholesterol, and biological reducing agents to neurotoxic H2O2 J Biol Chem 2002 277 40302 40308 12192006 10.1074/jbc.M206428200
Atwood CS Perry G Zeng H Kato Y Jones WD Ling KQ Huang X Moir RD Wang D Sayre LM Copper mediates dityrosine cross-linking of Alzheimer's amyloid-beta Biochemistry 2004 43 560 568 14717612 10.1021/bi0358824
Curtain CC Ali F Volitakis I Cherny RA Norton RS Beyreuther K Barrow CJ Masters CL Bush AI Barnham KJ Alzheimer's disease amyloid-beta binds copper and zinc to generate an allosterically ordered membrane-penetrating structure containing superoxide dismutase-like subunits J Biol Chem 2001 276 20466 20473 11274207 10.1074/jbc.M100175200
Dobson J Nanoscale biogenic iron oxides and neurodegenerative disease FEBS Lett 2001 496 1 5 11343696 10.1016/S0014-5793(01)02386-9
Nunomura A Perry G Pappolla MA Wade R Hirai K Chiba S Smith MA RNA oxidation is a prominent feature of vulnerable neurons in Alzheimer's disease J Neurosci 1999 19 1959 1964 10066249
Schipper HM Heme oxygenase expression in human central nervous system disorders Free Radic Biol Med 2004 37 1995 2011 15544918 10.1016/j.freeradbiomed.2004.09.015
Perry G Nunomura A Hirai K Zhu X Prez M Avila J Castellani RJ Atwood CS Aliev G Sayre LM Is oxidative damage the fundamental pathogenic mechanism of Alzheimer's and other neurodegenerative diseases? Free Radic Biol Med 2002 33 1475 1479 12446204 10.1016/S0891-5849(02)01113-9
Rinaldi P Polidori MC Metastasio A Mariani E Mattioli P Cherubini A Catani M Cecchetti R Senin U Mecocci P Plasma antioxidants are similarly depleted in mild cognitive impairment and in Alzheimer's disease Neurobiol Aging 2003 24 915 919 12928050 10.1016/S0197-4580(03)00031-9
Kontush A Mann U Arlt S Ujeyl A Luhrs C Muller-Thomsen T Beisiegel U Influence of vitamin E and C supplementation on lipoprotein oxidation in patients with Alzheimer's disease Free Radic Biol Med 2001 31 345 354 11461772 10.1016/S0891-5849(01)00595-0
Engelhart MJ Geerlings MI Ruitenberg A van Swieten JC Hofman A Witteman JCM Breteler MMB Dietary intake of antioxidants and risk of Alzheimer disease JAMA 2002 287 3223 3229 12076218 10.1001/jama.287.24.3223
Morris MC Evans DA Bienias JL Tangney CC Bennett DA Aggarwal N Wilson RS Scherr PA Dietary intake of antioxidant nutrients and the risk of incident Alzheimer disease in a biracial community study J Am Med Assoc 2002 287 3230 3237 10.1001/jama.287.24.3230
Luchsinger JA Tang MX Shea S Mayeux R Antioxidant vitamin intake and risk of Alzheimer disease Arch Neurol 2003 60 203 208 12580704 10.1001/archneur.60.2.203
White AR Barnham KJ Huang X Voltakis I Beyreuther K Masters CL Cherny RA Bush AI Cappai R Iron inhibits neurotoxicity induced by trace copper and biological reductants J Biol Inorg Chem 2004 9 269 280 14758525 10.1007/s00775-004-0521-8
Bush AI Metal complexing agents as therapies for Alzheimer's disease Neurobiol Aging 2002 23 1031 1038 12470799 10.1016/S0197-4580(02)00120-3
Doraiswamy PM Finefrock AE Metals in our minds: therapeutic implications for neurodegenerative disorders Lancet Neurol 2004 3 431 434 15207800 10.1016/S1474-4422(04)00809-9
Cherny RA Barnham KJ Lynch T Volitakis I Li QX McLean CA Multhaup G Beyreuther K Tanzi RE Masters CL Bush AI Chelation and intercalation: complementary properties in a compound for the treatment of Alzheimer's disease J Struct Biol 2000 130 209 216 10940226 10.1006/jsbi.2000.4285
Dedeoglu A Cormier K Payton S Tseitlin KA Kremsky JN Lai L Li X Moir RD Tanzi RE Bush AI Preliminary studies of a novel bifunctional metal chelator targeting Alzheimer's amyloidogenesis Exp Gerontol 2004 39 1641 1649 15582280 10.1016/j.exger.2004.08.016
Zheng H Weiner LM Bar-Am O Epsztejn S Cabantchik ZI Warshawsky A Youdim MBH Fridkin M Design, synthesis, and evaluation of novel bifunctional iron-chelators as potential agents for neuroprotection in Alzheimer's, Parkinson's, and other neurodegenerative diseases Bioorg Med Chem 2004
Lee JY Friedman JE Angel I Kozak A Koh JY The lipophilic metal chelator DP-109 reduces amyloid pathology in brains of human beta-amyloid precursor protein transgenic mice Neurobiol Aging 2004 25 1315 1321 15465629 10.1016/j.neurobiolaging.2004.01.005
Fisher AEO Naughton DP Novel peptoids for the detection and suppression of reactive oxygen and nitrogen species Biochem Soc Trans 2003 31 1302 1304 14641048
Fisher AEO Maxwell SC Naughton DP Catalase and superoxide dismutase mimics for the treatment of inflammatory diseases Inorg Chem Comm 2003 6 1205 1208 10.1016/S1387-7003(03)00223-5
Maskos Z Rush JD Koppenol WH The hydroxylation of tryptophan Arch Biochem Biophys 1992 296 514 520 1321587 10.1016/0003-9861(92)90605-V
Fisher AEO Hague TA Clarke CL Naughton DP Catalytic superoxide scavenging by metal complexes of the calcium chelator EGTA and contrast agent EHPG Biochem Biophys Res Comm 2004 323 163 167 15351716 10.1016/j.bbrc.2004.08.066
| 15826313 | PMC1087502 | CC BY | 2021-01-04 16:39:31 | no | Nutr J. 2005 Apr 12; 4:14 | utf-8 | Nutr J | 2,005 | 10.1186/1475-2891-4-14 | oa_comm |
==== Front
Nutr Metab (Lond)Nutrition & Metabolism1743-7075BioMed Central London 1743-7075-2-101581713110.1186/1743-7075-2-10ResearchExogenous recombinant human growth hormone effects during suboptimal energy and zinc intake Rising Russell [email protected] Julio F [email protected] Conrad [email protected] Rozalia [email protected] Debora [email protected] Fima [email protected] EMTAC Inc., 514 Santander Ave, #5, Coral Gables, FL 33134 USA2 Lawndale Medical Clinic, 7109-B Lawndale Ave., Houston, TX 77023 USA3 Department of Pediatrics, Emory University, 2040 Ridgewood Drive NE, Atlanta, GA 30322 USA4 71-50 Parsons Blvd, #5B, Flushing, NY 11365 USA5 Children's Hospital Boston, Dept. of Gastroenterology, 300 Longwood Avenue, Boston, MA 02115 USA6 Sansum Medical Research Institute, 2219 Bath Street, Santa Barbara, CA 93105 USA2005 7 4 2005 2 10 10 25 1 2005 7 4 2005 Copyright © 2005 Rising et al; licensee BioMed Central Ltd.2005Rising 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
Energy and Zinc (Zn) deficiencies have been associated with nutritional related growth retardation as well as growth hormone (GH) resistance. In this study, the relationship between suboptimal energy and/or Zn intake and growth in rats and their response to immunoreactive exogenous recombinant human GH (GHi), was determined.
Results
Rats treated with GHi and fed ad-libitum energy and Zn (100/100) had increased IGFBP-3 (p < 0.05) as compared with NSS (215 ± 23 vs. 185 ± 17 ng/ml) along with similar body weight gain. Rats treated with GHi and fed suboptimal energy and full Zn (70/100) had significantly increased weight gain (109.0 ± 18.2 vs. 73.8 ± 11.0 g) and serum IGF-I levels (568 ± 90 vs. 420 ± 85 ng/ml), along with decreased total body water (TBW; 61.0 ± 1.6 vs. 65.7 ± 2.1%) as compared to NSS controls. However, body weight gain was reduced (p < 0.05) as compared with rats fed ad-libitum energy. Growth hormone treated rats fed only suboptimal Zn (100/70), had increased weight gain (217.5 ± 13.2 vs. 191.6 ± 17.9 g; p < 0.05) compared to those given NSS. These rats gained weight in similar amounts to those fed full Zn. Rats treated with GHi and fed both suboptimal energy and Zn (70/70) showed similar results to those fed suboptimal energy with appropriate Zn (70/100), along with significant increases in IGFBP-3 levels (322 ± 28 vs. 93 ± 28 ng/ml). All restricted rats had reduced 24-h EE (kcal/100 g BW) and physical activity index (oscillations/min/kg BW) and GHi did not overcome these effects.
Conclusion
These results suggest that GHi enhances weight gain in rats with suboptimal energy and Zn intake but does not modify energy expenditure or physical activity index. Suboptimal Zn intake did not exacerbate the reduced growth or decrease in energy expenditure observed with energy restriction.
==== Body
Introduction
Nutritional growth retardation or nutritional dwarfing (ND) denotes a linear growth deterioration accompanied by inadequate weight gain [1]. It has been associated with short stature since growth deceleration represents an adaptive response to suboptimal nutrition. In the United States, ND is frequently associated with self imposed dieting due to fear of obesity and/or hypercholesterolemia, extreme exercise and eating disorders, among many other reasons [2-4]. In contrast to more severe forms of malnutrition observed worldwide, ND is manifested by a height-for-age deficit while both weight-for-height and standard biochemical nutritional indexes remain within normal limits [1]. Several reports have also demonstrated reduced Insulin-like Growth Factor-1 (IGF-I) level and decreased erythrocyte Na+K+-ATPase activity during suboptimal energy intake [5-9]. The degree of energy restriction portrays one of the most important factors related with growth hormone (GH) action. Consequently, elevated GH level along with reduced IGF-I, Insulin-like Growth Factor Binding Protein-3 (IGFBP-3) and Insulin levels have been described in both human and animal research as a consequence of starvation [10-18]. Furthermore, attenuated spontaneous GH secretion has also been reported in patients with ND during puberty [19]. However, utilizing an animal model of suboptimal nutrition, it has been reported that GH effects are feasible during mild energy restriction by administering exogenous GHi [20,21].
Although adequate energy intake plays a major role affecting growth, micronutrients like Zn have also demonstrated to be essential regulators of growth and GH action [22-24]. Profound Zn deficiency with adequate energy intake has been frequently associated with growth failure and delayed sexual maturation, affecting cell division, DNA, RNA and protein synthesis [25,26]. Zinc deficiency also reduces the level of liver GH receptors, serum IGF-I, GH binding proteins and both mRNA's for GH receptors and IGF-I [24,26]. Furthermore, binding of Zn in cultured rat hepatocytes with a Zn chelator DTPA (diethylenetriaminepenta-acetic acid) did not reduce mRNAs for IGF-I, GH receptors or GH binding proteins while metallothionein gene expression was strongly inhibited. Therefore, the decline in IGF-I associated with in vivo Zn deficiency did not appear to be due to changes in extracellular Zn concentration at the hepatocyte level [27]. Often Zn deficiency is associated with reduced energy intake [28,29] and the effects of each one of these nutritional alterations alone, or in combination with each other is not known. Furthermore, the effects of GHi administration in rats with experimentally induced suboptimal energy with and without decreased Zn intakes have not been elucidated.
In this paper we report the effects of exogenous GHi administration in an animal model of suboptimal nutrition where energy and/or Zn intake was decreased by 30% of requirements [5,20,21]. Our goal was to determine if exogenous GHi administration would attenuate the detrimental effects of mild restrictions of energy and Zn individually or in combination with each other. Furthermore, we hypothesized that energy expenditure and physical activity would be reduced with mild nutrient restrictions. This was determined utilizing a new Enhanced Metabolic Testing Activity Chamber.
Results
Effects of GHi on weight gain
All rats were healthy throughout the experiment without any apparent illness or weight loss. The effects of GHi administration (solid lines) versus NSS administration (dotted lines) for each dietary group are shown in Figures 3, 4, 5, 6.
Figure 3 Weekly weight gain for both growth hormone treated (filled dots, continued line) and normal saline control rats (filled squares, interrupted line) fed a 1:1 carbohydrate to fat purified diet with 100% of the energy and 100% of the zinc requirement for rats. * = significant difference (p < 0.05) for that particular week.
Figure 4 Weekly weight gain for both growth hormone treated (filled dots, continued line) and normal saline control rats (filled squares, interrupted line) fed a 1:1 carbohydrate to fat purified diet with 70% of the energy and 100% of the zinc requirement for rats. * = significant difference (p < 0.05) for that particular week.
Figure 5 Weekly weight gain for both growth hormone treated (filled dots, continued line) and normal saline control rats (filled squares, interrupted line) fed a 1:1 carbohydrate to fat purified diet with 100% of the energy and 70% of the zinc requirement for rats. * = significant difference (p < 0.05) for that particular week.
Figure 6 Weekly weight gain for both growth hormone treated (filled dots, continued line) and normal saline control rats (filled squares, interrupted line) fed a 1:1 carbohydrate to fat purified diet with 70% of the energy and 70% of the zinc requirement for rats. * = significant difference (p < 0.05) for that particular week.
No significant differences were found in body weight gain between the GHi treated and the NSS control group rats fed 100% energy/100% Zn (Diet A) (Figure 3). However, rats fed 70% energy/100% Zn (Diet B) and treated with GHi showed significantly (p < 0.05) increased weight gain over the four week experimental period as compared with their respective NSS controls. The differences in body weight gain became apparent after the first week of the study (Figure 4). However, the rats in this group fed restricted energy and full Zn gained less weight than the rats fed energy and Zn ad-libitum (Diet A) regardless of GHi treatment. Similarly growth hormone administration produced a response in the group fed 100% energy/ 70% Zn (Diet C) were GHi treated rats gained significantly (p < 0.05) more weight than their respective NSS controls (Figure 5). The differences became apparent after the second week of the study. However, the overall weight gain of the rats in this group fed restricted amounts of Zn and a full energy complement was similar to those in the group fed energy and Zn ad-libitum (100% energy/100% Zn; Diet A) regardless of GHi treatment. Finally, when both energy and Zn intake were reduced to 70% energy/70% Zn (Diet D), rats treated with GHi gained significantly (p < 0.05) more weight than their respective NSS controls (Figure 6). Weight gain differences became apparent after the first week of the study. The weight gain of the rats in this group was similar to that found for rats restricted to 70% of the ad-libitum energy intake with full Zn intake (Diet B).
Effects of GHi on IGF-I, IGFBP-3 and insulin levels
The hormonal pattern at the end of the study is shown in Table 3. When rats were fed 100% energy/100% Zn (Diet A) and treated with GHi, IGFBP-3 increased significantly (p < 0.05) versus those rats given NSS. In comparison to the NSS treated rats, treatment with GHi increased serum IGF-I in both groups of rats fed 70% ad-libitum energy with or without Zn restriction (Diets B and D). Furthermore, serum IGFBP-3 increased in the GHi-treated group when both energy and Zn intake were reduced to 70%. However, serum IGFBP-3 did not differ among GHi-treated rats fed at 70% of energy intake and 100% of Zn. Additionally, no significant differences were found in serum IGFBP-3 when energy was maintained at 100%. Serum insulin did not change significantly in GHi treated rats as compared to controls.
Table 3 Hormonal profile after mild energy and Zn restriction in rhGH treated rats
Energy/zinc (%) 100/100 70/100 100/70 70/70
rhGH1 NSS2 rhGH NSS rhGH NSS rhGH NSS
IGF-1 (ng/ml) 1012.5 ± 82.7 784.8 ± 231.3 568.4 ± 90.2* 420.3 ± 84.5 1089.5 ± 133.9 1074.1 ± 220.1 657.0 ± 114.1* 423.0 ± 139.9
IGFBP-3 (ng/ml) 214.5 ± 23.3* 184.7 ± 17.2 107.9 ± 35.5 96.9 ± 9.3 231.0 ± 36.0 201.0 ± 21.3 322.0 ± 28.4* 93.4 ± 28.0
Insulin (mIU/ml) 21.3 ± 7.6 52.7 ± 28.6 22.2 ± 12.6 39.5 ± 20.8 50.9 ± 14.4 71.4 ± 17.3 26.3 ± 6.6 17.1 ± 6.5
* = P < 0.05 between treatment groups
1 = Recombinant human growth hormone
2 = Normal saline solution
Metabolic changes associated with caloric restriction
In comparison to rats administered NSS, GHi administration had no effects on 24-hour energy expenditure and the index of physical activity within any of the dietary treatments (Table 4). However, the RQ was lower in GHi treated rats fed 70% of ad-libitum energy intake. Energy expenditure (p < 0.05) and physical activity index (p < 0.05) were reduced in all restricted groups as compared with all those fed a full complement of energy. The reduction in energy expenditure was from 19.5 ± 2.9 to 16.9 ± 1.9 kcal/100 g body weight while the reduction in the physical activity index was from 2.5 ± 0.5 to 2.1 ± 0.3 oscillations/min/kg body weight. Mild Zn restriction did not further reduce these parameters below those found for the energy restricted groups. Similarly, GHi administration did not overcome the reduced energy expenditure and physical activity index found in the energy restricted rats.
Table 4 Metabolic profile of rats after mild energy and zinc restriction
Energy/zinc (%) 100/100 70/100 100/70 70/70
Dietary treatment rhGH1 NSS2 rhGH NSS rhGH NSS rhGH NSS
24-EE2 (Kcal/100 g BW) 21.2 ± 2.4 19.6 ± 3.3 17.2 ± 1.5 16.4 ± 0.7 17.4 ± 3.4 19.8 ± 2.0 17.5 ± 2.8 16.5 ± 2.4
RQ3 (VCO2/VO2) 0.82 ± .02 0.89 ± .03 0.82 ± .08 0.89 ± .07 0.90 ± .09 0.81 ± .01 0.82 ± 0.1 0.91 ± 0.01
Physical activity4 2.6 ± 0.5 2.2 ± 0.2 2.0 ± 0.5 2.2 ± 0.1 2.3 ± 0.4 2.7 ± 0.8 2.3 ± 0.2 2.1 ± 0.3
1 = Recombinant human growth hormone
2 = Normal saline solution
3 = Twenty-four hour energy expenditure
4 = Respiratory quotient
5 = Oscillations/minute/kg body weight
Effects of GHi on body composition
Body composition analysis is shown in Table 5. There were no significant differences in FFM and BF between the GHi and NSS groups regardless of dietary treatment. However, compared to their respective NSS controls, rats treated with GHi and fed either energy restricted diet, with our without Zn deficiency (Diets B and D), showed significant (p < 0.05) reductions in TBW.
Table 5 Body composition by carcass analysis after mild energy and zinc restriction.
Energy/zinc (%) 100/100 70/100 100/70 70/70
Dietary treatment GHi1 NSS2 GHi NSS GHi NSS GHi NSS
Body fat (%) 12.0 ± 1.3 12.8 ± 2.1 8.4 ± 2.6 8.3 ± 1.0 12.3 ± 1.4 12.6 ± 2.1 7.2 ± 1.0 8.8 ± 3.0
Fat-free mass (%) 88.0 ± 1.3 87.2 ± 2.2 91.6 ± 2.6 91.7 ± 1.0 88.1 ± 1.4 87.6 ± 0.9 92.8 ± 1.0 91.2 ± 3.0
Total Body water (%) 57.7 ± 3.3 60.3 ± 2.2 61.1 ± 1.6* 65.8 ± 2.1 57.3 ± 2.9 59.9 ± 1.6 58.8 ± 7.2* 66.6 ± 2.1
* = P < 0.05 between treatment groups
1 Immunoreactive recombinant human growth hormone
2 Normal saline solution
Discussion
The experimental animal model of suboptimal nutrition allowed the independent assessment of suboptimal intake of specific nutrients and the effects of administered exogenous GHi. At the same time we were able to determine the effects of administered GHi under these conditions. In this study different conditions of suboptimal nutrition in rats were assessed by limiting dietary energy and/or zinc intakes to 70% of ad-libitum. This was possible by utilizing purified diets which allowed us to provide a precise amount of all nutrients, including Zn, which might not have been possible with commercial rat chows. Furthermore, this also permits determination of the effects of rhGH administration on 24-hour energy expenditure and on the index of physical activity utilizing an established rodent model of suboptimal nutrition, which has never been studied before. The data showed that exogenous recombinant human growth hormone enhanced body weight gain and increased serum IGF-I levels in energy restricted rats. Furthermore, serum IGFBP-3 levels were also increased during simultaneous suboptimal energy and zinc restriction. However, suboptimal zinc restriction alone did not interfere with body weight gain or exacerbate any of the detrimental effects of specific suboptimal energy restriction. Moreover, GHi treatment reduced the amount of total body water in energy restricted rats. Finally, a reduction in energy expenditure and the index of physical activity were shown during suboptimal energy intake and Zn restriction however, GHi treatment did not enhance these parameters. The rat's physical activity was not restricted in any way. The slightly smaller space provided by the Nalgene metabolic cage might have potentially reduced the rat's spontaneous physical activity, however this appeared not to affect the results. All rats had metabolic measurements under the same conditions.
Decreased body gain was previously reported during mild suboptimal energy intake in rats fed at 60 and 80% of ad-libitum energy intake [5]. These findings showed that a 40% reduction in energy intake reduced weight gain by 61% of control levels over a four week period. When considering all of the rats fed the reduced energy diet in the present study, regardless of treatment or Zn levels, a 30% reduction in energy intake resulted in a 40% reduction in body weight gain in comparison to all control animals.
It was previously found that GHi given to rats fed a restricted diet (60% energy) resulted in increased cumulative weight gain when compared to their NSS controls, while IGF-I and IGFBP-3 levels were elevated [20]. Our results confirm that GHi enhanced body weight gain by >18% when rats were fed suboptimally (reduction by 30% in dietary energy). Body weight differences were mainly associated to energy restriction. No body weight differences were evident in Zn restricted rats as long as the energy intake was appropriate.
Increments in IGF-I and IGFBP-3 serum concentrations were clearly linked to the effects of GHi in energy restricted rats. However, our data demonstrated no differences in serum IGF-I and IGFBP-3 concentrations with mild isolated Zn restriction. Thus, reduced Zn intake influenced the action of GHi in IGF-I and IGFBP-3 serum concentrations only when the energy intake was simultaneously restricted, suggesting no role for the suboptimal intake of Zn. It is possible that reducing Zn to only 70% of requirements, with adequate energy, is not enough to cause any significant effects on the hormonal parameters studied. It may be necessary to further reduce dietary Zn before an effect on IGF-I or IGFBP-3 is observed. For example, Zn deficiency in rats fed adequate energy grew poorly and showed reduced serum IGF-I, GH receptor numbers and GH binding proteins [26]. Furthermore, gene expression for both IGF-I and GH receptors in rats were reduced when fed a Zn deficient, adequate energy diet [24].
Altogether, recombinant human growth hormone therapy reduces carbohydrate utilization shifting all metabolic loads to lipid metabolism [30,31]. Moreover, physical activity, normal growth, dietary intake [31] and GH itself have an essential role in body composition [32-35]. For instance, the administration of GHi, promoting lipid utilization, reduces total body fat. In our study our partially energy restricted animals showed a slight increase in body weight gain with GHi treatment. Moreover, we found a slight reduction in the respiratory quotient with GHi administration in those rats on energy restricted diets. This suggests that GHi treatment shifts the metabolic load to lipogenesis, as has been found in a previous study [31,37]. Furthermore, GHi action was not affected by restricted Zn intake as long as the energy intake was not restricted.
The amount of GH administered may have variable effects on body composition in rats. For example, investigators in two previous studies found minimal changes in body fat and fat-free mass in rats fed ad-libitum and administered 0.05 or 0.10 mg/100 g body weight of GHi daily [20,21]. This is suggestive that under ad-libitum or mild energy restriction GHi treatment at or below 0.10 mg/100 g body weight will not change the amount of fat-free mass relative to body fat. It is possible that a higher dosage of GHi will elicit these effects. For example, body fat decreased significantly in similarly fed rats when administered 0.35 mg/100 g body weight of GHi daily [37]. Moreover, the whole body carcass analysis methodology may not be sensitive enough to detect subtle changes in body fat in rats at the lower dosages of GHi used in this and in similar studies [20,21]. This is further evidenced by the lack of changes in energy expenditure and physical activity index observed in rats fed either ad-libitum or energy restricted diets and treated with 0.1 mg/100 g body weight of GHi.
In this study energy restriction resulted in a reduction of 24-hour energy expenditure and physical activity index in rats. Furthermore, these changes were not affected by GHi treatment. This is suggestive that metabolic adaptations occurred due to suboptimal energy intake. Others have found similar metabolic and biochemical changes associated with suboptimal energy intake. For example, erythrocyte NaK-ATPase was reduced in children diagnosed with nutritional dwarfing, a form of suboptimal nutrition that exists for a prolonged period of time [8]. Similar reductions in erythrocyte NaK-ATPase were also found in rats fed energy restricted diets [5]. Moreover, metabolic rate and physical activity were reduced when rats were fed energy restricted diets [43]. The results of all these studies [5,8,20,21,37] suggests that metabolic adaptations occur beginning with mild energy restriction.
Growth deceleration, subsequent growth failure and short stature are the most remarkable consequences of persistent suboptimal nutrition [1-5]. Although adequate energy intake has a preponderant role affecting growth, many other micronutrients play a meaningful role in growth regulation and growth hormone action. Furthermore, extensive animal and human research have already established the association between energy restriction, severe Zn deficiency and growth failure. Therefore, elevated growth hormone levels, along with reduced IGF-I, IGFBP-3 and insulin levels, have been described in severe energy and Zn restrictions [10-18,22-26].
Malnourished humans have increased, whereas rats have decreased serum levels of growth hormone [38]. Decreased growth-hormone levels in starved rats may be caused by increased somatotropin-releasing inhibitory hormone or by reduced stimulation of hypothalamic somatotroph cells by growth-hormone-releasing hormone [39]. Furthermore, serum leptin has been found to play a role in the decline of growth hormone in rats [40]. However, serum insulin growth factor-1 (IGF-I) production in starved rats remains sensitive to growth hormone levels [41]. In previous studies we have validated the rat model for testing various treatments during suboptimal nutrition [5,20,21].
Growth retardation is the most predominant finding in many of the conventional Zn deficiency studies [24-29]. However, limited information exists concerning the effects of suboptimal Zn intake [42] and no one has investigated the effects of GHi administration during experimental suboptimal Zn restriction. Thus, our study showed GHi therapy promoting body weight gain despite suboptimal Zn intake. Consequently, in contrast to most of these studies involving severe Zn deficiency, our data suggest a secondary role of Zn when Zn intake was only restricted to 70% of daily requirements while maintaining adequate energy intake. In addition to weight gain, other GHi metabolic effects were also obtained, such as increases in both IGF-I and IGFBP-3 levels, during Zn restriction if associated with appropriate energy intake.
Our study results give us an idea of the role of Zn in growth during mild malnutrition, which is not clearly understood. The assessment of Zn status is hampered by the lack of a single sensitive and specific biochemical factor [43]. Currently, the clinical method to assess Zn status in children is to measure increase growth velocity in response to Zn supplementation. Controversy exists between studies, where positive effects of Zn supplementation have been found against no effects of supplementation. No effects were found in adolescents with ND and in sub Saharan infants and young children [5,6,12-15] while malnourished infants and children showed increased growth [6-8]. Researchers have suggested that these differences may be attributed to the severity of Zn deficiency, to the dose, frequency of dosing and age of the infants. It has also been observed that in non-stunted infants, Zn supplementation results in increased growth compared with the control group, but with a less pronounced effect. Therefore, researchers have concluded that beneficial effects of Zn supplementation on growth are related to the degree of stunting and of Zn deficiency. Furthermore, it was suggested that Zn supplementation halted the stunting process, due to improved appetite and decreased episodes of gastrointestinal and respiratory disease, in stunted infants in rural Ethiopia [44].
The amount and method of administration of growth hormone, along with the type of feeding regime might cause differences in anabolic responses. For example, healthy normally fed rats injected with 20 micrograms daily of recombinant human growth hormone at the sight of an induced fracture for 10 days showed reduced healing times without changes in bodyweight [45]. In another study, 15 day slow release recombinant human growth hormone laminar implants elicited the same anabolic effects in rats as the daily injectable form but without the associated problems such as renal toxicity and stress at the injection site [46]. Furthermore, a single injection of microencapsulated slow release recombinant human growth hormone in immunosuppressed Hpx rats showed greater long term pharmacological effects such as increased growth rate and IGF-I levels in comparison with daily injections for 35 days [47]. Moreover, there were no differences in the anabolic effects of either rodent or human growth hormone administration in rats [48]. In our study we utilized GHi was injected daily just under the skin (SC) at the back of the neck. The half-life of this type of growth hormone is 124 minutes when utilized in rodents [49]. We have conducted three prior studies of suboptimal nutrition in our laboratory [5,20,21] utilizing the same type of growth hormone. We always observed increased growth and body weight gain without any apparent health problems. It is possible that the short half life of this product prevented any toxic effects in our rats.
Conclusion
Our results demonstrate that beneficial effects of GHi are obtained during mild energy restriction. Moreover, mild Zn restriction does not have negative effects on body weight gain. Growth hormone therapy does promote body weight gain despite mild energy and Zn restrictions without any apparent effects on metabolic rate and physical activity. Finally, no additive effects between Zn and energy are observed during mild combined restriction of energy and Zn intakes.
Materials and methods
Forty pre-pubertal two week old male Sprague-Dawley rats were studied over a four-week period at the Miami Children's Hospital Research Institute in Miami Florida. The animals were individually housed in wire-bottom stainless steel cages avoiding coprophagia. The experimental design is depicted in Figure 1. Rats were maintained on a 12 hour light/dark cycle beginning at 7:00 AM. Four groups of 10 Sprague-Dawley male rats were fed four different balanced purified 1:1 carbohydrate to fat diets (Purina Mills Test Diets, Richmond, IN) adjusted for energy and Zn according to the following [50]:
Figure 1 Experimental design and assays
1) Diet A – 100% of all nutrients, fed ad-libitum
2) Diet B – 70% energy and 100% of all the nutrients, pair-fed with rats in group A
3) Diet C – 100% energy and all the nutrients but 70% Zn, fed ad-libitum
4) Diet D – 70% of both energy and Zn, and 100% of the rest of the nutrients, pair-fed with rats in group C
Purified diets were used in order to precisely control the amount of nutrients and to eliminate the variability often associated with commercial rat chows. The composition of the experimental diets is shown in Table 1. The control diet (Diet A) was formulated to contain all nutrients, including vitamins and minerals, necessary for normal growth in rats as defined by the National Research Council [50]. The dietary guidelines for rodents were based on rats consuming approximately 60 calories (15 g) of commercial rat chow per day. The formulation of each diet and the related modifications to the energy and Zn levels for the restricted diets (Diets B and D) were based on consumption of similar diets in three previous studies of suboptimal nutrition [5,20,21].
Table 1 Composition of experimental diets
Energy/Zinc (%)1 100/100 70/100 100/70 70/70
Casein (92% purity) 24.8 24.8 24.8 24.8
DL-Methionine 0.22 0.34 0.22 0.34
Sucrose 15.00 15.00 15.00 15.00
Dextrin 22.70 22.70 22.70 22.70
Lard 8.37 8.37 8.37 8.37
Corn Oil 8.37 8.37 8.37 8.37
Choline bitartrate 0.15 0.22 0.15 0.22
Vitamin mix 0.74 1.12 0.74 1.12
Mineral mix 2.60 3.79 2.60 3.79
Alpha cellulose 17.09 15.33 17.09 15.33
Zinc Carbonate (52% purity) 0.0023 0.0023 0.00162 0.00162
1 = The amount of each ingredient is the percentage of the total diet
The amount of both restricted diets (Diets B and D) fed were based on the amount of control diets (Diets A and C) consumed by pair-fed control rats. For example, the amount of the restricted diet (Diet B) fed was based on the consumption of the rats pair-fed the control diet (Diet A) from the previous day. All nutrients, except energy, were concentrated to compensate for the 30% reduction in food intake. For the rats fed the diet containing adequate energy but with reduced levels of dietary Zn (Diet C), the Zn level was reduced to 8.1 mg/kg of diet, based on the Zn requirements for adult rats (12 mg Zn/kg diet) as set by the National Research Council [50,30]. Both adequate Zn diets (Diets A and B) were formulated to contain 12 mg Zn/kg of diet. The amount of Diet D fed, were both energy and Zn were formulated at 70% of requirements, was based on the previous days intake of rats pair-fed adequate energy but with 70% of Zn requirements (Diet C).
One half of each group were administered a daily dose of 0.1 mg of GHi per 100 grams of body weight [20] (Somatropin rDNA Origin, Humatrope, Eli Lilly & Co. Indianapolis, IN) subcutaneously before 10:00 AM in the morning, while the controls received a similar amount of normal saline solution (NSS), the diluent for GHi.
The percentages of metabolisable energy from fat, carbohydrate and protein were adapted from McCarger et al [51] and approximated those of typical American diets. The proximate analysis and macro-nutrient distribution is shown in Table 2. Because the rats were provided excess protein over the minimal requirement (15% for growth, maintenance and breeding) in the ad-libitum diet (Table 2), it was not necessary to adjust crude protein content for formulation of the restricted diets. All control and restricted rats were consuming approximately 23 and 16% protein, respectively. However, micronutrient levels were maintained at adequate levels by increasing their concentration in the restrictive diets. The percentage of alpha cellulose was reduced to accommodate the increased concentrations of micronutrient mixes (Table 1).
Table 2 Proximate analysis of experimental diets
Energy/Zn(%) 100/100 70/100 100/70 70/70
Percentage macronutrient composition
Protein 22.78 15.95 22.78 15.95
Fat 38.38 38.38 38.38 38.38
Carbohydrate 37.70 26.39 37.70 26.39
Percentage contribution to energy
Protein 23.21 23.21 23.21 23.21
Fat 38.38 38.38 38.38 38.38
Carbohydrate 38.41 38.41 38.41 38.41
Percentage contribution to total intake
Energy 100 70 100 70
Zinc 100 100 70 70
Rats consumed deionized water ad-libitum. Furthermore, this water was used for food preparation. The deionized water was found to be zinc free by the Miami-Dade Department of Public Health utilizing graphic furnace atomic absorption spectroscopy. The lower limit of detection was 0.05 ug/l (Trace Analytical Laboratories, Muskegon MI). Weight and estimate of food intake were obtained for each rat daily in the morning prior to GHi and NSS injections. Food fed to the restricted rats was calculated based on the amount of food consumed by the respective pair mates in the ad-libitum fed groups. The amount of food given to the rats paired at 70% of energy restriction was calculated as follows: [(food consumed by the ad-libitum rat during the previous day/weight of this rat in the previous day) × (0.7) × (Current weight of the rat for which the food is estimated)]. Weight and food intake were recorded daily. After four weeks of the experimental period all rats were anesthetized with Nembutal (30–50 g/kg BW) and sacrificed by cardiac puncture in the morning at the same time of recording of body weight and GHi and NSS injections. Furthermore, all rats were fasted that same morning to prevent any effects of absorbed nutrients on hormonal assays. The last injection of GHi and NSS occurred the prior morning thus eliminating any effect of these injections on the results obtained from blood samples. Blood samples for serum IGF-I, IGFBP-3 and Insulin were drawn and carcasses were frozen at -20 degrees C for future measurements of body composition using Folch's analysis method for fat extraction [52]. Samples were not frozen more than one week prior to analysis.
All experimental protocols were reviewed and approved by the Miami Children's Hospital Institutional Animal Care and Use Committee.
Hormonal assays
Blood samples were spun and corresponding serum samples fractionated in aliquots. All serum samples were frozen at -20 degrees C for no more than one week for subsequent analysis. Each hormonal assay was run simultaneously on all samples.
Serum IGF-I concentration was measured by double antibody radioimmunoassay (RIA). Samples were prepared by acid-ethanol extraction followed by cryoprecipitation using a kit from Nichols Institute Diagnostics (San Juan Capistrano, CA). This procedure minimized the interference of IGF-I binding proteins at the time of the assay. The standard curve was linear between 5 and 600 ng/ml with a Coefficient of Variation of 3%. The sensitivity of this radioimmunoassay was equal to 0.06 ng/ml. The intra-assay coefficient of variation was equal to 2.4 and 3% at concentrations of 0.5 and 0.9 ng/ml, respectively. The inter-assay variance at 0.5 ng/ml and 0.8 ng/ml were equal to 5.2 and 8.4%, respectively.
Serum IGFBP-3 level was measured by double antibody RIA using a kit from Diagnostic Systems Laboratories, Inc. (Webster, TX). The standard curve was linear between 2.5 and 100 ng/ml with a Coefficient of Variation of 3.2%. The minimum detection limit was 0.5 ng/ml. The intra-assay precision at 82.7, 27.5 and 7.3 ng/ml was 1.8, 3.2 and 4%, respectively. The inter-assay precision at 80, 21, and 8 ng/ml was 1.9, 0.5 and 0.6%, respectively.
Serum Insulin level was measured by double antibody RIA using a kit from ICN Pharmaceuticals, Inc. (Costa Mesa, CA). The standard curve was linear between 5.5 and 310 mIU/ml with a Coefficient of Variation of 8%. The sensitivity was 1.5 mIU/ml. The intra-assay variation were determined for the following concentrations at 18.2, 36.5 and 91.5 mIU/ml and were 8.2, 4.2 and 5.4%, respectively.
Metabolic assessment
Each rat was weighed and had 24-hour energy expenditure (24-h EE; kcal/100 g BW), respiratory quotient (RQ;VCO2/VO2) and an index of physical activity (PA; oscillations/min/kg BW) measured in the Enhanced Metabolic Testing Activity Chamber (EMTAC). The main analytical unit of this instrument was developed to be suitable for various applications in both humans and animals for comprehensive measurements of energy expenditure and physical activity [53,54]. For this study, the EMTAC was retrofitted with a 72 liter plexiglass rodent enclosure (Figure 2). This maintained the rodent's environment thus eliminating the need to acclimate the rat to a different metabolic chamber. Measurements of 24-h energy expenditure and the index of physical activity consisted of first placing the rat, along with the appropriate experimental diet and deionized water, in a Nalgene metabolic cage. This cage was inserted into the EMTAC rodent enclosure at 10:00 AM and energy expenditure determinations were done by measurement of oxygen and carbon dioxide exchange within the enclosure. The same light/dark cycle used in the rodent facility was maintained. The physical activity index was obtained by placing the entire rodent enclosure on a balance that was connected to the EMTAC unit. The oscillations in weight (g), generated by movements of the rat, were read from the balance and utilized to calculate an index of physical activity expressed as oscillations in weight (g)/minute/kg body weight of the rat. For these calculations the software calculated the body weight of the rat in kilograms. The formulas used to calculate energy expenditure and physical activity index using the EMTAC have been described previously [53,54,34]. Energy expenditure was expressed per 100 g body weight to factor out the effects of body weight changes on metabolic rate.
Figure 2 Rodent EMTAC enclosure that was used to obtain measurements of 24-hour energy expenditure and the index of physical activity. Note the placement of the enclosure on the balance for measurement of the index of physical activity.
Body composition analysis
Body composition analysis included total body water content (TBW), fat-free mass (FFM) and body fat (BF) determinations. This procedure comprised three different steps: 1) carcass homogenization, 2) water content and dry weight determinations and 3) fat content determination by fat extraction.
The first step consisted of autoclaving each animal carcass for one hour at 15.3 psi and 120 degrees C in order to facilitate carcass homogenization. Each carcass was individually placed in a large beaker with covered tops with a known amount of distilled water. Each carcass was allowed to cool off overnight and then homogenized using a PowerGen700 blender. Triplicate aliquots of the homogenate were frozen at -20 degrees C for subsequent analysis. The chemical analysis of each homogenized carcass was carried out in triplicate and mean values of these triplicate samples were taken as ultimate values of TBW and BF content. The second step consisted of drying samples overnight at -380 mm Hg at 40 degrees C in a vacuum oven to determine water content. The difference between petri dish weight before and after overnight water extraction was considered as the dry weight of the carcass sample.
The third step consisted of determining BF content by a modified Folch's method [52] for fat extraction. All samples were analyzed following this technique's protocol which comprised two different procedures. In the first procedure, lipids were extracted from the homogenate by adding a 2:1 methanol-chloroform mixture. Each sample was separately filtered and a 5-fold volume of distilled water was added to separate lipids from non-lipid substances. This mixture was centrifuged for 15 minutes at 3000 rpm at three degrees C, producing three separated layers. The clear upper layer contained a mixture of methanol and water. The fluffy middle layer contained non-lipid substances and the clear lower layer contained a mixture of tissue lipids and chloroform. This bottom layer containing lipids and chloroform was isolated by removing the upper and the middle layers by vacuum aspiration. During the second procedure, all samples containing only the remaining bottom layer were dried overnight at -380 mm Hg at 40 degrees C in a vacuum oven, thus allowing chloroform evaporation and subsequent lipid separation. The difference between the tubes weight before and after fat extraction was taken as grams of fat content [55].
Fat-free mass was calculated by subtracting BF percentage from total body mass and expressed as percentage of total body mass.
Statistical analysis
The effects of GHi on body weight gain, FFM, BF, TBW, Insulin, IGF-I and IGPBP-3 within each dietary treatment group (GHi vs. NSS) were determined by Independent t-test. A similar analysis was utilized to compare changes in 24-hour energy expenditure and physical activity between all ad-libitum and energy restricted (70%) rats. The interaction between diet and hormone treatments across all diet groups was determined by 2-way ANOVA with least significant difference (LSD). Simple size was based on the results of three prior studies of suboptimal nutrition in rats [5,20,21]. In these studies five rats per individual treatment was enough to detect significant differences in body weight gain, the major parameter studied, at p < 0.05. The rest of the parameters were measured in order to explain the changes of body weight gain for each treatment group. All data are presented as mean ± standard deviation (SD) unless otherwise noted.
Abbreviations
ANOVA = Analysis of Variance
BF = Body Fat
FFM = Fat-free mass
GH = Growth hormone
Hg = Mercury
IGF-I = Insulin-like Growth Factor-1
IGFBP-3 = Insulin-like Growth Factor Binding Protein-3
NSS = Normal Saline Solution
ND = Nutritional dwarfing
PSI = Pounds per square inch
rhGH = recombinant human growth hormone
GHi = exogenous immunoreactive growth hormone
Na+K+-ATPase = Sodium-potassium ATP transporter enzyme
TBW = Total body water
Zn = Zinc
BW = Body weight
SD = Standard deviation
24-h EE = Twenty-four hour energy expenditure
RQ = Respiratory quotient
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
Dr. Russell Rising has contributed to the design of the experiment and conducted the data analysis. Furthermore, he either participated in some of the actual data acquisition or supervised pediatric research fellows in this regard. He also assisted in the preparation of the small grants necessary for funding of this project. Finally, he also assisted in the writing and editing of this manuscript.
Dr. Fima Lifshitz contributed to the preparation of the manuscript and assisted with data analysis. He also edited some of the grant proposals necessary for the financial support of this study. Both authors were involved in the final editing of this manuscript.
Acknowledgements
This work was supported in part by NIH grant (#1R43HD/DK38180-01A2) and by Pediatric Sunshine Academics.
==== Refs
Lifshitz F Tarim O Smith M Lifshitz F Nutritional growth retardation "Pediatric Endocrinology: A clinical guide" 1996 3 New York: Marcel Dekker, Inc
Pugliese MT Lifshitz F Grad G Fort P Marks-Katz M Fear of obesity: a cause of short stature and delayed puberty N Engl J Med 1983 309 513 518 6877321
Lifshitz F Moses N Growth failure: a complication of dietary treatment of hypercholesterolemia Am J Dis Child 1989 143 537 542 2718985
Smith NJ Excessive weight loss and food aversion in athletes simulating anorexia nervosa Pediatrics 1980 66 139 142 6931346
Tarim O Chasalow F Murphy J Rising R Carrillo A Lifshitz F Evaluation of differential effects of carbohydrate and fat intake on weight gain, IGF-I and erythrocyte Na+K+ATPase activity in suboptimal nutrition in rats J Am Coll Nutr 1997 16 159 165 9100217
David PJ Bernardis LL Changes in cell membrane Na+K+ATPase activity associated with induction of dietary obesity in the rat Metabolism 1984 33 591 595 6330493 10.1016/0026-0495(84)90054-4
Zhao M Willis JS Reduced ion transport in erythrocytes of male Sprague-Dawley rats during starvation J Nutr 1988 118 1120 1127 3418420
Lifshitz F Friedman S Smith M Cervantes C Recker B O'Connor M Nutritional dwarfing: a growth abnormality associated with reduced erythrocyte Na+, K+-ATPase activity Am J Clin Nutr 1991 54 997 1004 1659781
Jackson Smith W Underwood LE Clemmons DR Effects of caloric or protein restriction on insulin-like growth factor-1 (IGF-I) and IGF-binding proteins in children and adults J Clin Endocrinol Metab 1995 80 443 449 7531712 10.1210/jc.80.2.443
Le Roith D Insulin-like growth factors N Engl J Med 1997 336 633 640 9032050 10.1056/NEJM199702273360907
Zamboni G Dufillot D Antoniazzi F Valentini R Gendrek D Tat L Growth hormone-binding proteins and insulin-like growth factor-binding proteins in protein- energy malnutrition, before and after nutritional rehabilitation Pediatr Res 1996 39 410 414 8929859
Clemmons DR Underwood LE Nutritional regulation of IGF-I and IGF binding proteins Ann Rev Nutr 1991 11 393 412 1716449 10.1146/annurev.nu.11.070191.002141
Thissen JP Underwood LE Maiter D Maes M Clemmons DR Ketelslegers JM Failure of IGF-I infusion to promote growth in protein-restricted rats despite normalization of serum IGF-I concentrations Endocrin 1991 128 885 890
Snyder DK Clemmons DR Underwood LE Dietary carbohydrate content determines responsiveness to growth hormone in energy-restricted humans J Clin Endocrinol Metab 1989 69 745 752 2778033
Clemmons DR Klibanski A Underwood LE McArthur JW Ridgeway EC Beitins IZ Van Wyk JJ Reduction of plasma immunoreactive somatomedin-C during fasting in humans J Clin Endocrinol Metab 1981 53 1247 1250 7197688
Isley WL Underwood LE Clemmons DR Changes in plasma somatomedin-C in response to ingestion of diets with variable protein and energy content JPEN 1984 8 407 411
Phillips LS Unterman TG Somatomedin activity in disorders of nutrition and metabolism J Clin Endocrinol Metab 1984 13 145 189 10.1016/S0300-595X(84)80012-2
Underwood LE Thissen JP Lemozy S Ketelslegers JM Clemmons DR Hormonal and nutritional regulation of IGF-I and its binding proteins Horm Res 1994 42 145 151 7532613
Abdennur J Pugliese MT Cervantes C Fort P Lifshitz F Alterations in spontaneous growth hormone(GH) secretion and the response to GH-releasing hormone in children with nonorganic nutritional dwarfing J Clin Endocrinol Metab 1992 75 930 934 1517388 10.1210/jc.75.3.930
Carrillo A Rising R Tverskaya R Lifshitz F Effects of exogenous recombinant human growth hormone on an animal model of suboptimal nutrition J Am Coll Nutr 1998 17 276 281 9627915
Carrillo A Rising R Cole C Tverskaya R Lifshitz F Low dosages of exogenous growth hormone and its effects on growth in an animal model of suboptimal nutrition Nut 2000 16 1074 1078 10.1016/S0899-9007(00)00435-4
Harel Z Shaffer Tannenbaum G Long-term alterations in growth hormone and insulin secretion after temporary dietary protein restriction in early life in the rat Pediatr Res 1995 38 747 753 8552444
Ninh NX Thissen JP Collette L Gerard G Khoi HH Ketelslegers JM Zinc supplementation increases growth and circulating IGF-I in growth-retarded Vietnamese children Am J Clin Nutr 1996 63 514 519 8599314
McNall A Etherton T Fosmire G The impaired growth induced by zinc deficiency in rats is associated with decreased expression of the hepatic IGF-I and the growth hormone receptor genes J Nutr 1995 125 874 879 7722689
Roth HP Kirchgener M Influence of alimentary zinc deficiency on the concentration of growth hormone, IGF-I and insulin in the serum of force-fed rats Horm Metab Res 1994 26 404 408 7835822
Ninh NX Thissen JP Maiter D Adam E Mulumba N Ketelslegers JM Reduced liver IGF-I gene expression in young zinc-deprived rats is associated with a decrease in liver growth hormone (GH) receptors and serum GH-binding protein J Endocrin 1995 144 449 456
Lefebvre D Beckers F Ketelslegers JM Thissen JP Zinc regulation of insulin-like growth factor-1 (IGF-I), growth hormone receptor (GHR) and binding protein (GHBP) gene expression in rat cultured hepatocytes Mol Cell Endocrinol 1998 138 127 136 9685221 10.1016/S0303-7207(98)00012-4
MacDonald RS The role of zinc in growth and cell proliferation J Nutr 2000 130 1500S 1508S 10801966
Hosea HJ Taylor CG Wood T Mollard R Weiler HA Zinc-deficient rats have more limited bone recovery during repletion than diet-restricted rats Exp Biol Med 2004 229 303 311
Leveille GA Cloutier PF Isocaloric diets: effects of dietary changes Am J Clin Nutr 1987 45 158 163 3799509
Snyder DK Underwood LE Clemmons DR Persistent lipolytic effect of exogenous growth hormone during caloric restriction Am J Med 1995 98 129 134 7847429 10.1016/S0002-9343(99)80396-9
Snyder DK Underwood LE Clemmons DR Anabolic effects of growth hormone in obese diet-restricted subjects are dose dependent Am J Clin Nutr 1990 52 431 437 2203249
Clemmons DR Snyder DK Williams R Underwood LE Growth hormone administration conserves lean body mass during restriction in obese subjects J Clin Endocrinol Metab 1987 64 878 883 3558728
Salomon F Cuneo RC Hesp R Snksen H The effects of treatment with recombinant human growth hormone on body composition and metabolism in adults with growth hormone deficiency N Engl J Med 1989 321 1797 1803 2687691
Wilmore DW Moylan JA JrBristow BF Mason AD JrPruitt BA Jr Anabolic effects of human growth hormone and high caloric feeding following thermal injury Surg Gynecol Obstet 1974 138 875 884 4827031
Sevette A Kee AJ Carlsson AR Baxter RC Smith RC Parenteral nutrition with lipid or glucose suppresses liver growth and response to GH in adolescent male rats Am J Physiol Endocrinol Metab 2001 281 E1063 1072 11595664
Evans SA Overton JM Alshingiti A Levenson CW Regulation of metabolic rate and substrate utilization by zinc deficiency Metabolism 2004 53 727 232 15164319 10.1016/j.metabol.2004.01.009
Thissen JP Underwood LE Ketelsleger JM Regulation of insulin-like growth factor-1 in starvation and injury Nutr Rev 1999 57 167 176 10439629
Tannenbaum GS Epelbaum J Colle E Brazeau P Martin JB Anti serum to somatostatin reverses starvation-induced inhibition of growth hormone but not insulin secretion Endocrinology 1978 102 1909 1914 744057
Vuagnat BAM Pierroz DD Lalaoui M Englaro P Pralong FP Blum WF Aubert ML Evidence for a leptin-neuropeptide Y axis for the regulation of growth hormone secretion in the rat Neuroendocrinology 1998 67 291 300 9641610 10.1159/000054326
Oster MH Fielder PJ Levin N Cronin MJ Adaptation of the growth hormone and insulin-like growth factor-1 axis to chronic and severe caloric or protein malnutrition J Clin Invest 1995 95 2258 2265 7537760
Nishi Y Hatano S Aihara K Fujie A Kihara M Transient partial growth hormone deficiency due to zinc deficiency J Am Coll Nutr 1989 8 93 97 2708733
Prasad A Zinc deficiency in humans: a neglected problem J Am Col Nutr 1998 17 542 543
Umeta M West CE Haidar J Deurenberg P Hautvast JG Zinc supplementation and stunted infants in Ethiopia: a randomized controlled trial Lancet 2000 355 2008 2009 10885346 10.1016/S0140-6736(00)02348-5
Andreassen TT Oxlund H Local anabolic effects of growth hormone on intakct bone and healing fractures in rats Calcif Tissue Int 2003 73 258 264 14667139 10.1007/s00223-002-2074-6
Garcia J Dorta MJ Munguia O Llabres M Farina JB Biodegradable laminar implants for sustained release of recombinant human growth hormone Biomaterials 2002 24 4759 4764 12361614 10.1016/S0142-9612(02)00225-9
Takada S Kurokawa T Misaki M Taira K Yamagata Y A new animal model for evaluation of long-term growth rate over one month by rhGH/PLGA microcapsule formulations J Pharm Pharmacol 2003 55 951 961 12906752 10.1211/0022357021323
Malmlof K Din N Johansen T Pedersen SB Growth hormone affects both adiposity and voluntary food intake in old and obese female rats Eur J Endocrinol 2002 146 121 128 11751077 10.1530/eje.0.1460121
Varhaeghe J Van Bree R Van Herck E Thomas H Skottner A Dequeker J Mosekilde L Einhorn TA Bouillon R Effects of recombinant human growth hormone and insulin-like growth factor-I, with or without 17 beta-estradiol, on bone and mineral homeostasis of aged overiectomized rats J Bone Miner Res 1996 11 1723 1735 8915780
National research Council Nutrient requirements of laboratory animals 1995 Forth revised Washington DC. National Academy Press
McCargar LJ Baracos VE Clandinin T Influence of dietary carbohydrate-to-fat ratio on whole body nitrogen retention and body composition in adult rats J Nutr 1989 119 1240 1245 2795238
Folch J Lees M Sloane SGH A simple method for the isolation and purification of total lipids from animal tissues J Biol Chem 1957 226 497 509 13428781
Cole C Rising R Mehta R Hakim A Choudhury S Sundaresh M Lifshitz F Comprehensive assessment of the components of energy expenditure in infants using a new infant respiratory chamber J Am Coll Nutr 1999 18 233 41 10376779
Rising R Duro D Cedillo M Valois S Lifshitz F Daily metabolic rate in healthy infants J Peds 2003 143 180 185 10.1067/S0022-3476(03)00362-7
Boozer C Schoenbach G Atkinson R Dietary fat and adiposity: a dose- response relationship in adult male rats fed isocalorically Am J Physiol Endocrinol Metab 1995 268 E546 E550
| 15817131 | PMC1087503 | CC BY | 2021-01-04 16:37:46 | no | Nutr Metab (Lond). 2005 Apr 7; 2:10 | utf-8 | Nutr Metab (Lond) | 2,005 | 10.1186/1743-7075-2-10 | oa_comm |
==== Front
Reprod Biol EndocrinolReproductive biology and endocrinology : RB&E1477-7827BioMed Central London 1477-7827-3-111580436610.1186/1477-7827-3-11HypothesisMitochondrial 3 beta-hydroxysteroid dehydrogenase (HSD) is essential for the synthesis of progesterone by corpora lutea: An hypothesis Chapman John C [email protected] Jose R [email protected] Soohong [email protected] Sandra D [email protected] Department of Biological Sciences, Binghamton University, Binghamton, NY 13902-6000, USA2 Notre Dame Ambulatory Care Center, Medical Director, 1000 Broad Street, Central Falls, RI 02863, USA2005 3 4 2005 3 11 11 12 1 2005 3 4 2005 Copyright © 2005 Chapman et al; licensee BioMed Central Ltd.2005Chapman 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.
In mouse ovaries, the enzyme 3 beta-hydroxysteroid dehydrogenase (HSD) is distributed between microsomes and mitochondria. Throughout the follicular phase of the estrous cycle, the HSD activity in microsomes is predominant; whereas, after LH stimulation, HSD activity during the luteal phase is highest in the mitochondria. The current study examined whether or not LH stimulation always results in an increase in mitochondrial HSD activity. This was accomplished by measuring the HSD activity in microsomal and mitochondrial fractions from ovaries of pregnant mice. These animals have two peaks of LH during gestation, and one peak of LH after parturition. It was found that mitochondrial HSD activity was highest after each peak of LH. It is proposed that mitochondrial HSD is essential for the synthesis of high levels of progesterone. The increase in HSD activity in mitochondria after LH stimulation occurs because: 1) LH initiates the simultaneous synthesis of HSD and the cholesterol side-chain cleavage enzyme (P450scc); and, 2) HSD and P450scc bind together to form a complex, which becomes inserted into the inner membrane of the mitochondria. High levels of progesterone are synthesized by mitochondrial HSD because: 1) the requisite NAD+ cofactor for progesterone synthesis is provided directly by the mitochondria, rather than indirectly via the rate limiting malate-aspartate shuttle; and, 2) the end-product inhibition of P450scc by pregnenolone is eliminated because pregnenolone is converted to progesterone.
==== Body
Background
With the exception of 3β-hydroxysteroid dehydrogenase (HSD), the enzymes involved in the conversion of cholesterol to steroid hormones are located in either the mitochondria or the endoplasmic reticulum. HSD is unique in that it is located in both subcellular organelles. In either location, HSD converts pregnenolone and dehydroepiandosterone (DHEA) to progesterone and androstenedione, respectively, using NAD+ as cofactor. The reason for two separate sites for this enzyme is not known.
Establishing the existence of two separate locations for HSD has been a lengthy process. In 1956, Beyer and Samuels reported that the microsomal (endoplasmic reticulum) and mitochondrial fractions from the homogenate of bovine adrenal cortex contained HSD activity [1]. However, the HSD activity found in the mitochondrial fraction was attributed to microsomal contamination and the result of the homogenizion process. While this study established the legitimacy of microsomal HSD, it tended to preclude further research on mitochondrial HSD. For mitochondrial HSD to be considered a distinct and separate entity, additional research over a number of years would be required. Starting in 1965, investigators began to report a dual location for HSD in ovaries [2-4], testes [5,6], human term placenta [7-10], and rat adrenal cortex [11-14]. In toto, these studies suggested that mitochondrial HSD was indeed a separate entity. Other investigators, however, still considered mitochondrial HSD activity to be due to microsomal contamination [15,16], and the result of a redistribution artifact [17].
In 1979, we reported the results of an intracellular enzyme distribution study of HSD, cytochrome c oxidase (mitochondrial marker), and steroid 21 hydroxylase (microsomal marker) in rat adrenal cortex [18]. We found that exhaustively washed mitochondria retained 26 % of total HSD activity. In retrospect, this percentage appears to be on the low side. For example, when the specific activity of microsomal HSD is determined, and its contribution to the HSD activity in the remaining cell fractions (nuclear/unbroken cell, mitochondrial, and mitochondrial wash fractions) ascertained, then a maximum of 60 % of total homogenate HSD activity can be attributed to microsomal HSD. This indicates that mitochondrial HSD constitutes 40 % of total homogenate HSD activity, rather than 26 % as we initially reported. In rhesus monkey placenta, HSD activity is equally distributed between mitochondria and microsomes [19], and in bovine adrenal cortex, mitochondrial HSD comprises 30 % of total HSD activity [20,21].
Our study also found that mitochondrial HSD utilizes matrix space NAD+ as cofactor, indicating that the enzyme is located in the inner mitochondrial membrane [18]. This location for mitochondrial HSD has been established in bovine adrenal cortex [20,21], and in rat testis [5]. The combined techniques of immuno-cytochemistry and electron microscopy have identified immune reactive HSD in mitochondria of human ovary [22], and in the mitochondria of rat ovary, testis, and adrenal cortex [23]. Mitochondrial HSD has now been isolated from bovine adrenal cortex [21,24], and from human term placenta [25,26], and purified to homogeneity. It is now known that microsomal HSD and mitochondrial HSD are identical proteins [25-28]. The reason for two locations for the same enzyme has yet to be determined.
In a study of the intracellular distribution of mitochondrial HSD and microsomal HSD in mouse ovaries over the course of the estrous cycle [29], we reported that during diestrus (luteal phase), the specific activity of mitochondrial HSD was 80 % higher than that of microsomal HSD. This is in sharp contrast to the other three stages of estrus where microsomal HSD had the highest specific activity. In a study in which the cDNA of human placental HSD was transfected into Sf9 cells, the resultant HSD enzyme was distributed between mitochondria and microsomes [27]. In the luteal cells of mouse ovary, the distribution of the enzyme is skewed in favor of the mitochondria. During the luteinization process, luteal cells express high levels of mRNA for the cholesterol side-chain cleavage enzyme (P450scc) and for HSD [30,31]. The simultaneous synthesis of these two enzymes is very likely the reason for the increase in mitochondrial HSD activity, as will be discussed later.
The results of our previous study [29] tentatively suggested that the increase in mitochondrial HSD activity in mouse ovary during diestrus is due to LH. The pregnant mouse has two peaks of LH during gestation, and one peak of LH after parturition [32]. After each peak of LH, the levels of circulating progesterone increase [33]. The present study examined the distribution of HSD activity in pregnant mouse ovaries to determine whether or not mitochondrial HSD activity also increased after each peak of LH. We found that it did, which led to our proposed explanation for the reason for two separate locations for HSD in corpora lutea.
Methods
Animals
Female mice of the (C3H/HeJ × 129/J)F1 (C31) hybrid were used in the study. Parental stocks were purchased from Jackson Laboratories, Bar Harbor, ME. The mice were housed in an animal room kept at 24°C, with controlled lighting of 14L:10D (lights on at 0600 hr and off at 2000 hr). Purina lab chow (Ralston-Purina, St. Louis, MO) and water were provided ad libitum. The C31 offspring were weaned at 23–28 days of age. Female siblings were housed two per cage. For pregnancy experiments, C31 females were mated with C31 males. Only those with regular estrous cycles (4–5 days) were used for the study.
Determining Stage of the Estrous Cycle and mating
In order to promote a regular estrous cycle in the C31 females, cage shavings were taken from cages containing C31 males, and placed in cages containing the females. Also, male cages surrounded female cages. Vaginal smears were taken daily, usually in the afternoon. The smears were spread on a glass slide in a drop of physiological saline, stained using hematoxylin and eosin Y, and the stage of estrous determined by the method of Rugh [34].
In all experiments, each study group consisted of six females that were between 80 and 190 days of age. Two of the study groups were comprised of animals that were in diestrus and proestrus. The remaining groups all consisted of pregnant animals. These groups were formed as follows: females that were in either proestrus or estrus were mated late in the afternoon or about 2 hr prior to lights off. At lights on the following morning, and for subsequent days if necessary, all females were inspected for the presence of vaginal plugs. The day of vaginal plug was considered as day 0 of pregnancy. Over the gestation period, pregnant females were sacrificed on days 5, 10, 15, and 20. An additional group of females was sacrificed on day 5 postpartum.
Tissue Collection and Processing
Animals were lightly etherized, weighed, and then decapitated. The ovaries were removed, trimmed of fat, weighed in pairs, and placed on an ice-filled petri dish in a few drops of homogenizing buffer (0.3 % BSA, 1 mM EDTA, 0.25 M Sucrose, and 30 mM Tris-HCL [pH 7.4], 4°C), as per Chapman and Sauer [18]. The twelve ovaries were finely minced, then transferred to an ice-cold 5 ml Potter-Elvehjem glass homogenizer. The volume of the sample was brought up to 4 ml with additional homogenization medium, and the ovaries homogenized in the cold room (4°C) with 8 complete strokes of a motorized Teflon pestle. The homogenate was then transferred to a 10 ml centrifuge tube, and the glass homogenizer rinsed with 1 ml of homogenizing medium. The rinse was combined with the homogenate, and a 1 ml sample of the total homogenate removed and saved for later assay.
The uteri were removed from 10 day, 15 day, and 20 day pregnant mice. Trophoblasts were excised from the middle section of each uterine horn. After weighing, the trophoblasts were minced, and processed as described for the ovaries, except that the mince was homogenized in a Thomas glass-glass homogenizer.
Differential Centrifugation
The ovarian homogenate, contained in a 10 ml centrifuge tube, was centrifuged at 700 × g in a Sorvall RC-5 refrigerated centrifuge for 10 min. The supernatant was removed and spun at 10,000 × g for 20 minutes in the same centrifuge. The low-speed pellet, containing nuclei and unbroken cells, was resuspended in 1.5 ml homogenizing buffer. Following the 10,000 × g centrifugation, the resultant mitochondrial pellet was resuspended in 1.5 ml homogenizing buffer. The postmitochondrial supernatant was centrifuged in a Beckman L65 Ultracentrifuge (Beckman Co., Fullerton, CA) for 1 h at 105,000 × g. This yielded cytosol and a pellet of microsomes. Microsomes were resuspended in 1.5 ml homogenizing buffer using a 2 ml Potter-Elvehjem glass homogenizer with Teflon pestle. All fractions, including the total homogenate, were divided into aliquots in 12 × 75-mm borosilicate glass tubes, covered with Parafilm (American Can Co., Greenwich, CT), and frozen at -10°C. Enzymatic analyses were scheduled so that the frozen sub-cellular fractions were thawed only once prior to assay.
Enzymatic Analyses and Other Assays
The total homogenate, low speed pellet (nuclei/unbroken cells), mitochondrial, and microsomal fractions were assayed for HSD activity by measuring the conversion of pregnenolone to progesterone. Duplicate tissue samples of 50 μl and 100 μl were added to 15-ml glass test tubes containing 1 ml of incubation medium (50 mM sucrose, 20 mM KC1, 1 mM EDTA, 30 mM Tris-HCl [pH 7.4], 0.3% BSA), and 0.5 mM NAD+, as per Chapman et al. [29]. The incubates were then placed in a 37°C water bath and allowed to equilibrate for 5 min. The reaction was started by the addition of 100 nmol of pregnenolone in 10 μl of ethanol. After 15 min of incubation, the progesterone product was extracted into 1 ml of spectral grade heptane. The absorbency of progesterone was measured in a Gilford Response spectrophotometer (Gilford Systems, Oberlin, OH). Progesterone has an absorbency peak at 233 nm in heptane and a molar extinction coefficient of 17,000 [18]. In order to access the extraction efficiency of 1 ml of heptane, known concentrations of progesterone standards were run concurrently with the tissue samples. After extraction into heptane, the absorbancy of the standards was compared to their absorbance measured directly in heptane.
Cytochrome c oxidase, an inner mitochondrial membrane marker, was assayed by the procedure of Wharton and Tzalgaloff [35]. Enzymatic activity was determined by the rate of decrease in absorbancy at 550 nm. Protein content was measured by using the method of Bradford [36]. All assays were in duplicate. Replicate data were analyzed for significant differences using ANOVA (Dunnett, and Scheffe' F-test).
Results
The distribution of HSD activity between mitochondria and microsomes undergoes a unique shift in the transition from proestrus to diestrus. At proestrus, for example, the highest HSD activity is in the microsomes. At diestrus, in contrast, mitochondrial HSD activity is almost double that of microsomal HSD [29]. The present study re-examined this phenomenon; and, as Figure 1 shows, the activity of mitochondria HSD increases significantly at diestrus. The activity of the mitochondrial inner membrane enzyme, cytochrome c oxidase, also increases at diestrus. Total ovarian protein, in contrast, decreases.
Figure 1 Protein content, cytochrome c oxidase activity, and HSD activity in total homogenates, and in mitochondrial and microsomal fractions from ovaries of C31 mice that were in either proestrus or diestrus. Each experimental group consisted of 6 mice. Mitochondrial and microsomal fractions were isolated by differential centrifugation. Results are expressed per individual ovary.
Figure 2 contains the results of two separate experiments in which HSD activity was measured in mitochondrial and microsomal fractions during pregnancy, and 5 days after parturition. As indicated, the activities of mitochondrial HSD and microsomal HSD both increased over the course of the gestation period. However, the increase in HSD activity in the two organelles was inconsistent. For example, at 15 days and 20 days of gestation, the highest HSD activity was in the microsomal fraction. In contrast, at 5 days and 10 days of gestation, and at 5 days postpartum, mitochondrial HSD activity was greater than that of microsomal HSD. These three time points directly follow the peaks of LH [32]. Cytochrome c oxidase activity also increased during pregnancy. At day 20, cytochrome c oxidase activity was more than double the activity measured at 5 days. Total ovarian protein was inversely correlated with the peaks of LH. For example, at 15 days and 20 days of gestation, each ovary contained 1190 μg protein and 1025 μg protein, respectively. In contrast, at 5 days and 10 days of gestation, and at 5 days postpartum, each ovary contained 890 μg protein, 825 μg protein, and 810 μg protein, respectively. This relationship between LH and total ovarian protein also occurs during the estrous cycle [29]. For example, as shown in Figure 1, each ovary at proestrus contained 1030 μg protein; whereas at diestrus, each ovary contained 838 μg protein.
Figure 2 Protein content, cytochrome c oxidase activity, and HSD activity in total homogenates, and in mitochondrial and microsomal fractions from ovaries of C31 mice that were pregnant for 5 days, 10 days, 15 days, and 20 days, or were 5 days postpartum. Each experimental group consisted of 6 mice. Mitochondrial and microsomal fractions were isolated by differential centrifugation. Results of two separate experiments are expressed per individual ovary. Statistical analyses of the enzymatic activities of cytochrome c oxidase and HSD in total homogenates at each time-point showed a significant difference @ P < .05, when compared to animals in diestrus. In addition, the HSD activities in all mitochondrial and microsomal fractions of pregnant mice were significantly different @ P < .05, when compared to animals in diestrus.
Figure 3 contains the results of the measurement of HSD activity in mitochondrial and microsomal fractions of trophoblasts from 10 day, 15 day, and 20 day pregnant mice. As indicated, HSD activity was not detected (N.D.) in trophoblasts from 10 day pregnant mice. However, trophoblasts from 15 day and 20 day pregnant mice were found to produce 0.4 nmol progesterone/min/trophoblast and 0.6 nmol progesterone/min/trophoblast, respectively. Note that the highest HSD activity in trophoblasts is in the microsomal fraction.
Figure 3 Protein content, cytochrome c oxidase activity, and HSD activity in total homogenates, and in mitochondrial and microsomal fractions from trophoblasts of C31 mice that were pregnant for 10 days, 15 days, and 20 days. Each experimental group in the 10 day and 20 day pregnant animals consisted of 6 mice. N = 12 for the 15 day pregnant group. Mitochondrial and microsomal fractions were isolated by differential centrifugation. Results are expressed per individual trophoblast. N.D. = Not detected.
Discussion
The results of this and the previous study [29] leave little doubt that mitochondrial HSD activity increases after LH stimulation. How the increase in mitochondrial HSD activity is achieved and for what purpose, are the topics of this discussion.
When mitochondrial HSD was initially purified from the bovine adrenal cortex, the enzyme was found to have a close association with P450scc [24]. This association was of such a high degree that mitochondrial HSD actually copurified with P450scc. Antibodies against mitochondrial HSD precipitated both HSD and P450scc; and, conversely, antibodies against P450scc precipitated both P450scc and mitochondrial HSD. The degree of association between the two enzymes was measured, and a binding constant (KD) of 0.12 μM was determined. As would be expected, P450scc also bound to purified microsomal HSD [24]. HSD is insoluble and inactive in an aqueous medium, due to a segment of the protein, referred to as the "membrane-spanning domain" [27]. Delete this segment and HSD becomes soluble. With the segment in place, HSD is inserted into the membranes of microsomes and mitochondria [27]. The observation that mitochondrial HSD activity increases after LH stimulation suggests that HSD is preferentially inserted into the mitochondrial membrane. The fact that HSD binds to P450scc is very likely the mechanism for its insertion. This suggests that HSD and P450scc bind together, either during, or directly after their synthesis, since it is unlikely that HSD could bind to P450scc, already in place. The mRNAs for HSD and P450scc are expressed concurrently in luteal cells of the rat [37-40], cow [41,42], sheep [42-44], horse [45], macaque monkey [46], and human [38,47-49].
Figure 4 is a representative diagram of the proposed effect of the concurrent synthesis of HSD and P450scc on the intracellular distribution of HSD in luteal cells. Initiated by LH; the expression of HSD mRNA and P450scc mRNA results in the simultaneous synthesis of HSD and P450scc. The two enzymes bind to each other to form a complex, which is then inserted into the inner mitochondrial membrane. Molecules of HSD that do not bind to P450scc are inserted into the membrane of the endoplasmic reticulum.
Figure 4 Representative diagram of the proposed effect of the concurrent synthesis of HSD and cytochrome P450scc on the intracellular distribution of HSD in luteal cells. Initiated by LH; the transcription of HSD mRNA and P450scc mRNA results in the simultaneous production of HSD and P450scc. The two enzymes bind to each other to form a complex, which is then inserted into the inner mitochondrial membrane. Molecules of HSD that do not bind to P450scc are inserted into the membrane of the endoplasmic reticulum.
In the mouse there are peaks of estradiol-17β at proestrus (100 pg/ml) and metestrus (200 pg/ml) [29]. However, these levels are significantly less than the levels of progesterone produced throughout the luteal phase. During pregnancy the levels of circulating progesterone are even higher. One could argue that the increased levels of circulating progesterone in pregnant mice are due to the HSD activity in trophoblasts. This possibility was addressed in the current study. As shown in Figure 3, homogenates of individual trophoblasts from 15 day and 20 day pregnant mice are capable of producing 0.4 nmol progesterone/min/trophoblast and 0.6 nmol progesterone/min/trophoblast, respectively. This level of HSD activity in a single trophoblast is only 3 % of the HSD activity produced by the paired ovaries. However, large litters would increase the percentage. At day 10 there is no question that the ovaries are the major source of circulating progesterone, which averages 55 ng/ml [33]. This level of progesterone is between 275 fold and 550 fold of the peak levels of estradiol-17β produced during the follicular phase.
In order for luteal cells to synthesize high levels of progesterone, a number of events have to occur. First, higher levels of the two enzymes, P-450scc and HSD, have to be produced. This event is initiated when their respective mRNAs are expressed, as referenced above. Secondly, the increase in steroid synthesis requires an increased supply of cholesterol. This is achieved by removing cholesterol from cholesterol ester stores [50,51], and by initiating the de novo synthesis of cholesterol [52-54]. The latter event is quite likely in anticipation of fertilization of the ova and the need for a supply of cholesterol beyond diestrus. In the pregnant mouse, luteal cells synthesize high levels of progesterone for three weeks [33]. The de novo synthesis of cholesterol requires a carbon source, as well as ATP and NADPH. In luteal cells the carbon source is acetate; ATP is generated through glycolysis; and NADPH is produced during the oxidative decarboxylation of isocitrate and malate [54]. The enzymes that catalyze the latter reaction are NADP+-linked isocitrate dehydrogenase and NADP+-linked malate dehydrogenase. Both enzymes are in abundance in the cytoplasm of luteal cells [55].
The enzyme, P450scc is capable of producing 53 nmol pregnenolone/min/mg protein [56]. This requires an undiminished supply of NADPH, as well as the aforementioned cholesterol. The NADPH that is produced in the cytoplasm is of no direct use to the P450scc enzyme. However, reducing equivalents from NADPH can be transferred from the cytoplasm to mitochondria via the malate-aspartate shuttle [57,58]. Unfortunately, the mitochondria in luteal tissue lack an NADP+-linked malate dehydrogenase [59,60]. As a result, NADPH cannot be generated in the mitochondria by the oxidative decarboxylation of malate. This is in sharp contrast to adrenal cortex tissue, which does have a mitochondrial NADP+-linked malate dehydrogenase [61,62]. In luteal tissue the mitochondria and cytoplasm both contain an NAD+ -linked malate dehydrogenase, each enzyme having a high specific activity [55,59]. This indicates that there is ample capacity to transfer reducing equivalents from NADH into the mitochondria. If this transfer were to occur in tissues such as liver and heart, the NADH would be used to produce ATP. In steroidogenic tissues, the NADH can also be used to produce NADPH. For example, in mitochondria of luteal tissue, reducing equivalents are transferred from NADH to NADP+ by an energy-independent pyridine-nucleotide transhydrogenase [56]. Sonicates of luteal mitochondria are reported to catalyze the production of 60 nmol NADPH/min/mg protein from NADH [56]. Mitochondria in adrenal cortex tissue have a similar process, except that the transfer is energy-dependent [63]. The main source of NADPH in the mitochondria of luteal tissue is provided by NADP+-linked isocitrate dehydrogenase [56]. This enzyme is capable of reducing 253 nmol NADP+/min/mg protein [56].
In the synthesis of progesterone, NAD+ is reduced to NADH. In order to maintain a high rate of progesterone synthesis, NADH has to be continually oxidized to NAD+. With microsomal HSD, the oxidation of NADH would have to occur via either the α-glycerol phosphate shuttle or the malate-aspartate shuttle. Both shuttles transfer reducing equivalents from NADH into the mitochondria. However, the α-glycerol phosphate shuttle does not operate in the ovary [64], which leaves the transfer of reducing equivalents to the malate-aspartate shuttle. The ovary already heavily utilizes this shuttle. As discussed earlier, the malate-aspartate shuttle transfers reducing equivalents from NADH into the mitochondria for the P450scc enzyme. In addition, the shuttle is involved in the oxidation of the NADH produced during glycolysis. A high rate of glycolysis during the de novo synthesis of cholesterol, for example, generates high levels of NADH. If the levels of NADH exceed the carrying capacity of the shuttle system, the reducing equivalents are transferred to pyruvate via the enzyme, lactate dehydrogenase. A high level of lactate is a signal that the shuttle system is rate limiting.
The ovary produces appreciable amounts of lactate, even during the early follicular phase [65]. As follicular size increases, lactate levels also increase [66], coinciding with the start of antrum formation and detectable estradiol-17β secretion [66,67]. After the LH surge, the levels of lactate increase an additional 2.5 fold [66,68]. In luteal tissue, a high percentage of the glucose taken up is metabolized only as far as pyruvate and lactate [69]. Iodoacetate, an inhibitor of glycolysis, abolishes the effect of LH on lactate accumulation and significantly reduces LH-stimulated progesterone synthesis [68,70].
The fact that the malate-aspartate shuttle is rate-limiting could be the reason for a mitochondrial location for HSD. However, pregnenolone is an end-product inhibitor of the P450scc reaction [71,72], and a mitochondrial location for HSD would remove the steroid from its site of inhibition. Progesterone does not inhibit the P450scc reaction [71]. Evidence that mitochondrial HSD is involved in the production of high levels of progesterone is provided by the observation that mitochondria from thecal tissue convert only 6.4 % of total 14C-cholesterol to 14C-progesterone (1.2 %) and 14C-pregnenolone (5.2 %); whereas, in contrast, mitochondria from luteal tissue convert 16.1 % of total 14C-cholesterol to 14C-progesterone (13.5 %) and 14C-pregnenolone (2.6 %) [73].
In human adrenals and gonads, HSD is derived from the same gene and has been classified as type II, relative to the placental enzyme, which is classified as type I [27,28,74,75]. In the adrenals and gonads of the rat [74-76] and mouse [77,78], the enzyme is also derived from the same gene, which in these rodents is classified as type I. The fact that ovaries and adrenal cortex contain the same HSD enzyme indicates that their response to trophic hormone stimulation would also be the same. One would expect, therefore, that ACTH stimulation would cause an increase in mitochondrial HSD activity in adrenal cortex tissue. ACTH stimulation of male and female rats does cause an increase in HSD mRNA and HSD activity [79]. However, it is not known if ACTH stimulation causes a preferential increase in mitochondrial HSD activity.
Aside from the role that Steroidogenic Acute Regulatory protein (StAR) plays in controlling cholesterol access to mitochondria, the rate-limiting step in steroidogenesis is considered to be P450scc [80]. However, the fact that P450scc and HSD are bound together as a complex [24] suggests that the rate-limiting step, or steps, entails the conversion of cholesterol to progesterone. There is a decided advantage in these two enzymes functioning as a unit. Instead of shuttling steroid intermediates from organelle to organelle, cholesterol can be converted to progesterone in what could be described as a single step. The levels of progesterone can be increased even further if the reactions of both enzymes are coupled together, which appears to be the case. The discovery of the energy-independent NADH/NADP+ transhydrogenase initially led to the assumption that it existed to supply P450scc with NADPH [56]. However, the transhydrogenase is capable of producing less than one/half the NADPH needed to synthesize high levels of pregnenolone. If its function is to act as a principal supplier of NADPH, it is inadequate. However, if its main function is to ensure that HSD has an undiminished supply of NAD+, and is operating at Vmax, it is more than adequate. It is an ideal means of coupling HSD to P450scc. Unfortunately, the exchange of one molecule of NADH to produce one molecule of NADPH is insufficient for the overall conversion of progesterone to cholesterol. This is because the P450scc reaction utilizes 3 molecules of NADPH for the synthesis of one molecule of pregnenolone. The remainder of the NADPH for this reaction would have to be supplied by mitochondrial NADP+-linked isocitrate dehydrogenase [56].
The P450scc/HSD enzyme complex is well regulated. In addition to the end-product inhibition exerted by pregnenolone on the P450scc reaction [71,72], the HSD reaction is affected by the redox state of cytoplasmic pyridine nucleotides [81]. For example, extramitochondrial NAD+ increases mitochondrial HSD activity by up to 40 %; whereas, extramitochondrial NADH inhibits HSD activity by as much as 70 %. Figure 5 is a representative diagram of the proposed regulation of the conversion of cholesterol to progesterone. As indicated, pregnenolone acts as an end product inhibitor of the P450scc reaction, [······ (-)]; and, in the mitochondrial HSD reaction, extramitochondrial NAD+ operates as an allosteric activator [------(+)] and extramitochondrial NADH operates as an allosteric inhibitor [······ (-)]. With this level of control it is difficult to imagine how pregnenolone could ever leave the luteal cells without being converted to progesterone.
Figure 5 The proposed regulation of progesterone synthesis by pregnenolone and by extramitochondrial NAD+/NADH. In the overall conversion of cholesterol to progesterone, pregnenolone acts as an end product inhibitor in the P450scc reaction, [······ (-)]; and, in the mitochondrial HSD reaction, extramitochondrial NAD+ operates as an allosteric activator [------(+)] and extramitochondrial NADH operates as an allosteric inhibitor [······ (-)].
Finally; it was noted in the results section that total ovarian protein was inversely correlated with the peaks of LH. A decrease in total ovarian protein during diestrus could be due to ovulation. However, this does not explain the lower protein levels following LH stimulation that occurs during pregnancy and after parturition. A rapid and ongoing synthesis of the two enzymes, P450scc and HSD is critical to the production of high levels of progesterone. This necessitates a ready supply of amino acids. It is speculative of course, but the action of LH could include the initiation of proteolysis of protein stores from luteal tissue, which could explain the lower protein levels.
Conclusion
The ovary has two levels of steroid synthesis. One level occurs during the follicular phase, and a higher level of synthesis occurs throughout the luteal phase. We believe the higher level of synthesis is due to, and the reason for, mitochondrial HSD. To synthesize estradiol-17β during the follicular phase, steroid precursors are shuttled from cell type to cell type and from organelle to organelle. The synthesis of progesterone during the luteal phase involves one cell type and two enzymes. With HSD in the mitochondria, rather than in the microsomes, the shuttle of steroid precursors is unnecessary. It also allows for HSD and P450scc to function together as a unit, a decided advantage for producing high levels of progesterone. This is especially true if the two enzymes are coupled together by the NADH/NADP+ transhydrogenase. A mitochondrial location for HSD also solves the problem inherent with the rate-limiting malate-aspartate shuttle, and it removes the end-product inhibition of pregeneolone by converting it to progesterone. Finally, the fact that HSD and P450scc have a strong binding affinity for each other, and are synthesized simultaneously, tentatively suggests a means by which LH stimulation results in an increase in mitochondrial HSD activity.
==== Refs
Beyer KL Samuels LT Distribution of the steroid-3β-ol dehydrogenase in cellular structures of the adrenal gland J Biol Chem 1956 219 69 76 13295257
Armstrong DT Jackanicz Keyes PL Mckerns KW Regulation of steroidogenesis in the rabbit ovary The gonads 1969 Amsterdam, North-Holland Publishing Co 3 25
Sulimovici S Boyd GS The Δ5-3β-hydroxysteroid dehydrogenase of rat ovarian tissue: the effect of adenosine 3',5' cyclic-monophosphate acid Eur J Biochem 1969 7 549 558 4388369
Dimino MJ Campbell MD Progesterone synthesis by luteal mitochondria in vitro Proc Soc Exp Biol Med 1976 152 54 56 1265080
Sulimovici S Bartoov Lunenfeld B Localization of 3-β-hydroxysteroid dehydrogenase in the inner membrane subfraction of rat testis mitochondria Biochem Biophys Acta 1973 321 27 40 4356307
van der Vusse GJ Kalkman ML van der Molen HJ 3β-Hydroxysteroid dehydrogenase in rat testis tissue. Inter- and subcellular localization and inhibition by cyanoketone and nagarse Biochem Biophys Acta 1974 348 404 414 4847562
Koide SS Torres MT Distribution of 3β-hydroxysteroid dehydrogenase and Δ5-3-oxosteroid isomerase in homogenate fractions of human term placenta Biochem Biophys Acta 1965 105 115 120 5849107
Ferre F Breuiller M Cedard L Duchesne M-J Saintato M Descomps B Crastes de Paulet A Human placental Δ5-3β hydroxysteroid dehydrogenase activity: intracellular distribution, kinetic properties, retroinhibition and influence of membrane delipidation Steroids 1975 26 551 570 1879 10.1016/0039-128X(75)90050-1
Edwards DP O'Conner JL Bransome ED JrBraselton WE Jr Human placental 3β-hydroxysteroid dehydrogenase/Δ5-isomerase. Demonstration of an intermediate in the conversion of 3β-hydroxypregn-5-en-20-one to pregn-4-ene-3, 20-dione J Biol Chem 1976 251 1632 1638 1254588
Meigs RA Sheean LA Mitochondria from human term placenta. III. The role of respiration and energy generation in progesterone synthesis Biochim Biophys Acta 1977 489 225 235 922026
McCune RW Roberts S Young PL Competitive inhibition of adrenal Δ5-3β-hydroxysteroid dehydrogenase and Δ5-3-ketosteroid isomerase activities by adenosine 3', 5'-monophosphate J Biol Chem 1970 245 3859 3867 4321768
Basch RS Finegold MJ 3β-Hydroxysteroid dehydrogenase activity in the mitochondria of rat adrenal homogenates Biochem J 1971 125 983 989 4401302
Hochberg RB Ladany S Welch M Lieberman S Cholesterol and cholesterol sulfate as substrates for the adrenal side-chain cleavage enzyme Biochemistry 1974 13 1938 1945 4152127 10.1021/bi00706a025
Kream BE Sauer LA On the intracellular localization of the 3β-hydroxysteroid dehydrogenase/isomerase in the rat adrenal cortex Proc Soc Exp Biol Med 1976 152 459 462 181764
Moustafa AM Koritz SB Concerning the subcellular distribution of 3β-hydroxysteroid dehydrogenase/isomerase in the rat adrenal Proc Soc Exp Biol Med 1975 149 823 825 170624
Caffrey JL Nett TM Abel JH JrNiswender GD Activity of 3β-hydroxy-Δ5- steroid dehydrogenase/Δ5-Δ4-isomerase in the ovine corpus luteum Biol Reprod 1979 20 279 287 222364
Cowan RA Giles CA Grant JK The intramitochondrial distribution of 3β-hydroxysteroid dehydrogenase-oxosteroid isomerase. A probable redistribution artifact J Steroid Biochem 1974 5 607 608 4372491 10.1016/0022-4731(74)90113-7
Chapman JC Sauer LA Intracellular localization and properties of 3β-hydroxysteroid dehydrogenase/isomerase in the adrenal cortex J Biol Chem 1979 254 6624 6630 447739
Sholl SA 3β-Hydroxysteriod dehydrogenase/Δ5 -Δ4 isomerase activity in the rhesus monkey placenta and fetal adrenal, testis and ovary during late gestation Steroids 1983 41 757 768 6582658 10.1016/0039-128X(83)90050-8
Cherradi N Defaye G Chambaz EM Dual subcellular localization of the 3β-hydroxysteroid dehydrogenase isomerase: characterization of the mitochondrial enzyme in bovine adrenal cortex J Steroid Biochem Mol Biol 1993 46 773 779 8274411 10.1016/0960-0760(93)90318-Q
Cherradi N Defaye G Chambaz EM Characterization of the 3β-hydroxysteroid dehydrogenase activity associated with bovine adrenocortical mitochondria Endocrinology 1994 134 1358 1364 8119176 10.1210/en.134.3.1358
Hiura M Nogawa T Fujiwara A Electron microscopy of cytodifferentiation and its subcellular steroidogenic sites in the thecal cell of the human ovary Histochemistry 1981 71 269 277 6940854 10.1007/BF00507830
Pelletier G Li S Luu-The V Tremblay Y Belanger A Labrie F Immunoelectron microscopic localization of three key steroidogenic enzymes (cytochrome P450(scc), 3β-hydroxysteroid dehydrogenase and cytochrome P450(c17)) in rat adrenal cortex and gonads J Endocrinol 2001 171 373 383 11691658 10.1677/joe.0.1710373
Cherradi N Chambaz EM Defaye G Organization of 3β-hydroxysteroid dehydrogenase/isomerase and cytochrome P450scc into a catalytically active molecular complex in bovine adrenocortical mitochondria J Steroid Biochem Mol Biol 1995 55 507 514 8547175 10.1016/0960-0760(95)00199-9
Thomas JL Meyers RP Strickler RC Human placental 3β-hydroxy-5-ene-steroid dehydrogenase and steroid 5 → 4-ene-isomerase: purification from mitochondria and kinetic profiles, biophysical characterization of the purified mitochondrial and microsomal enzymes J Steroid Biochem 1989 33 209 217 2770297 10.1016/0022-4731(89)90296-3
Luu-The V Takahashi M Labrie F Purification of microsomal and mitochondrial 3β-hydroxysteroid dehydrogenase/Δ5 - Δ4 isomerase from human placenta Ann NY Acad Sci 1990 595 386 388
Thomas JL Evans BW Blanco G Mason JI Strickler RC Creation of a fully active, cytosolic form of human type I 3β-hydroxysteroid dehydrogenase/isomerase by the deletion of a membrane spanning domain J Mol Endocrinol 1999 23 231 239 10514560 10.1677/jme.0.0230231
Thomas JL Mason JI Blanco G Veisaga ML The engineered, cytosolic form of human type I β-hydroxysteroid dehydrogenase/isomerase: purification, characterization and crystallization J Mol Endocrinol 2001 27 77 83 11463578 10.1677/jme.0.0270077
Chapman JC Waterhouse TB Michael SD Changes in mitochondrial and microsomal 3β-hydroxysteroid dehydrogenase activity in mouse ovary over the course of the estrous cycle Biol Reprod 1992 47 992 997 1337280
Richards JS Jahnsen T Hedin L Lifka J Ratoosh S Durica JM Goldring NB Ovarian follicular development: from physiology to molecular biology Rec Prog Horm Res 1987 43 231 276 2819995
Richards JS Hedin I Molecular aspects of hormone action in ovarian follicular development, ovulation, and luteinization Ann Rev Physiol 1988 50 441 463 3288100 10.1146/annurev.ph.50.030188.002301
Murr SM Bradford GE Geschwind II Plasma luteinizing hormone, follicle-stimulating hormone and prolactin during pregnancy in the mouse Endocrinology 1974 94 112 116 4855563
Murr SM Stabenfeldt GH Bradford GE Geschwind II Plasma progesterone during pregnancy in the mouse Endocrinology 1974 94 1209 1211 4856431
Rugh R Reproductive systems of adult mice The Mouse, its Reproduction and Development 1990 Oxford, Oxford University Press 24 43
Wharton DC Tzagoloff A Cytochrome c oxidase from beef heart mitochondria Methods Enzymol 1967 10 245 250
Bradford MM A rapid and sensitive method for the quantitation of microgram quantities of protein utilizing the principle of protein-dye binding Anal Biochem 1976 72 248 254 942051
Goldring NB Durcia JM Lifka J Hedin L Ratoosh SL Miller WL Orly J Richards JS Cholesterol side-chain cleavage P450 messenger ribonucleic acid: evidence for hormonal regulation in rat ovarian follicles and constitutive expression in corpora lutea Endocrinology 1987 120 1942 1950 3106012
Mason JI Keeney DS Bird IM Rainey WE Morohashi K-I Leers-Sucheta S Melner MH The regulation of 3β-hydroxysteroid dehydrogenase expression Steroids 1997 62 164 168 9029732 10.1016/S0039-128X(96)00176-6
Juneau C Dupont E Luu-The V Labrie F Pelletier G Ontogenesis of 3β-hydroxysteroid dehydrogenase Δ5-Δ4 isomerase in the rat ovary as studied by immunocytochemistry and in situ hybridization Biol Reprod 1993 48 226 234 8439611
Zhao H-F Labrie C Simard J de Launoit Y Trudel C Martel C Rheaume E Dupont E Luu-The V Pelletier G Labrie F Characterization of rat 3β-hydroxysteroid dehydrogenase/Δ5-Δ4 isomerase cDNAs and differential tissue-specific expression of the corresponding mRNAs in steroidogenic and peripheral tissues J Biol Chem 1991 266 583 593 1985917
Couet J Martel C Dupont E Luu-The V Sirard M-A Zhao H-F Pelletier G Labrie F Changes in 3β-hydroxysteroid dehydrogenase/Δ5-Δ4-isomerase messenger ribonucleic acid, activity and protein levels during the estrous cycle in the bovine ovary Endocrinology 1990 127 2141 2148 2226306
Conley AJ Kaminski MA Dubowsky SA Jablonka-Shariff A Redmer DA Reynolds LP Immunohistochemical localization of 3β-hydroxysteroid dehydrogenase and P450 17α-hydroxylase during follicular development in pigs, sheep, and cows Biol Reprod 1995 52 1081 1094 7626708
Hawkins DE Belfiore CJ Kile JP Niswender GD Regulation of messenger ribonucleic acid encoding 3β-hydroxysteroid dehydrogenase/Δ5-Δ4 isomerase in the ovine corpus luteum Biol Reprod 1993 48 1185 1190 8481479
Juengel JL Guy MK Tandeski TR McGuire WJ Niswender GD Steady-state concentrations of messenger ribonucleic acid encoding cytochrome P450 side-chain cleavage and 3β-hydroxysteroid dehydrogenase/ Δ5-Δ4-isomerase in ovine corpora lutea during the estrous cycle Biol Reprod 1994 51 380 384 7803610
Boerboom D Sirois J Equine P450 cholesterol side-chain cleavage and 3β-hydroxysteroid dehydrogenase/Δ5-Δ4 isomerase: molecular cloning and regulation of their messenger ribonucleic acids in equine follicles during the ovulatory process Biol Reprod 2001 64 206 215 11133676
Bassett SG Little-Ihrig LL Mason JI Zeleznik AJ Expression of messenger ribonucleic acids that encode for 3β-hydroxysteroid dehydrogenase and cholesterol side-chain cleavage enzyme throughout the luteal phase of the macaque menstrual cycle J Clin Endocrinol Metab 1991 72 362 366 1991805
Doody KJ Lorence MC Mason JI Simpson ER Expression of messenger ribonucleic acid species encoding steroidogenic enzymes in human follicles and corpora lutea throughout the menstrual cycle J Clin Endocrinol Metab 1990 70 1041 1045 2180973
Dupont E Labrie F Luu-The V Pelletier G Immunocytochemical localization of 3β-hydroxysteroid dehydrogenase/Δ5-Δ4 isomerase in human ovary J Clin Endocrinol Metab 1992 74 994 998 1569177 10.1210/jc.74.5.994
Suzuki T Sasano H Tamura M Aoki H Fukaya T Yajima A Nagura H Mason JI Temporal and spatial localization of steroidogenic enzymes in premenopausal human ovaries: in situ hybridization and immunohistochemical study Mol Cell Endocrinol 1993 97 135 143 8143896 10.1016/0303-7207(93)90220-E
Herbst AL Response of rat ovarian cholesterol to gonadotropins and anterior pituitary hormones Endocrinology 1967 81 54 60 4290935
Bjersing L On the ultrastructure of granulosa lutein cells in porcine corpus luteum. With special reference to endoplasmic reticulum and steroid hormone synthesis Z Zellforsch Mikrosk Anat 1967 82 187 211 4881190
Savard K Marsh JM Rice BF Gonadotropins and ovarian steroidogenesis Rec Prog Hormon Res 1965 21 285 365
Armstrong DT Gonadotropins, ovarian metabolism, and steroid biosynthesis Rec Prog Horm Res 1968 24 255 319 4882326
Flint APF Denton RM Glucose metabolism in the superovulated rat ovary in vitro Biochem J 1969 112 243 254 4240707
Flint APF Denton RM The role of nicotinamide-adenine dinucleotide phosphate-dependent malate dehydrogenase and isocitrate dehydrogenase in the supply of reduced nicotinamide-adenine dinucleotide phosphate for steroidogenesis in the superovulated rat ovary Biochem J 1970 117 73 83 4393612
Robinson J Stevenson PM Electron flow and cholesterol-side-chain cleavage in ovarian mitochondria Eur J Biochem 1971 24 18 30 4400340 10.1111/j.1432-1033.1971.tb19651.x
Simpson ER Estabrook RW A possible mechanism for the transfer of cytosol-generated NADPH to the mitochondrial mixed-function oxidases in bovine adrenal cortex: a malate shuttle Arch Biochem Biophys 1968 126 977 978 4386814 10.1016/0003-9861(68)90501-8
Simpson ER Estabrook RW The "malate shuttle" and control of steroid hydroxylation in the adrenal cortex Adv Enzyme Regul 1969 7 259 279 4391642 10.1016/0065-2571(69)90022-3
Stevenson PM Forman E Malate and gonadotropin stimulation of rat ovary J Endocrinol 1971 49 22 23 5090892
Stevenson PM Taylor PL "Malic enzyme" and alkaline phosphatase in mitochondria from pig and rat ovaries FEBS Lett 1971 19 251 254 11946224 10.1016/0014-5793(71)80526-4
Simpson ER Estabrook RW Mitochondria malic enzyme: the source of reduced nicotinamide adenine dinucleotide phosphate for steroid hydroxylation in bovine adrenal cortex mitochondria Arch Biochem Biophys 1969 129 384 395 4178715 10.1016/0003-9861(69)90190-8
Sauer LA An NAD+- and NADP+-dependent malic enzyme with regulatory properties in rat liver and adrenal cortex mitochondrial fractions Biochem Biophys Res Commun 1973 50 524 531 4144006 10.1016/0006-291X(73)90871-1
Cammer W Estabrook RW Respiratory activity of adrenal cortex mitochondria during steroid hydroxylation Arch Biochem Biophys 1967 122 721 734 10.1016/0003-9861(67)90182-8
Knodel EL Doeg KA Transfer of reducing equivalents in the rat polycystic ovary J Endocrinol 1977 72 107 108 833536
Zeilmaker GH Verhamme CM Lactate concentrations in pre-ovulatory follicles of pro-oestrus rats before and after onset of oocyte maturation Acta Endocrinol 1977 86 380 383 578626
Harlow CR Winston RML Margara RA Hillier SG Gonadotrophic control of human granulosa cell glycolysis Human Reprod 1987 2 649 653
Boland NI Humpherson PG Leese HJ Gosden RG Pattern of lactate production and steroidogenesis during growth and maturation of mouse ovarian follicles in vitro Biol Reprod 1993 48 798 806 8485244
Lieberman ME Ahren K Tsafriri A Bauminger S Linder HR Relationship between glycolysis and steroidogenesis in cultured Graafian follicles stimulated by LH or prostaglandin E2 J Steroid Biochem 1975 6 1445 1449 1052845 10.1016/0022-4731(75)90083-7
Armstrong DT Stimulation of glycolytic activity of rat corpus luteum tissue by luteinizing hormone Endocrinology 1963 72 908 913
Linder HR Tsafriri A Lieberman E Zor U Koch Y Bauminger S Barnea A Gonadotropin action on cultured Graafian follicles: induction of maturation division of the mammalian oocyte and differentiation of the luteal cell Rec Prog Horm Res 1974 30 79 138 4601886
Koritz SB Hall PF End-product inhibition of the conversion of cholesterol to pregnenolone in an adrenal extract Biochemistry 1964 3 1298 1304 14231720 10.1021/bi00897a019
Koritz SB Hall PF Feedback inhibition by pregnenolone: A possible mechanism Biochim Biophys Acta 1964 93 215 217 14249162
Dimino MJ Differences in mitochondrial steroidogenesis between follicular and luteal tissues of porcine ovaries Endocrinology 1977 101 1844 1849 590196
Pelletier G Dupont E Simard J Luu-The V Belanger A Labrie F Ontogeny and subcellular localization of 3β-hydroxysteroid dehydrogenase (3β-HSD) in the human and rat adrenal, ovary and testis J Steroid Biochem Mol Biol 1992 43 451 467 1390295 10.1016/0960-0760(92)90084-V
Zhao H-F Labrie C Simard J de Launoit Y Trudel C Martel C Rheaume E Dupont E Luu-The V Pelletier G Labrie F Characterization of rat 3β-hydroxysteroid dehydrogenase/Δ5-Δ4 isomerase cDNAs and differential tissue-specific expression of the corresponding mRNAs in steroidogenic and peripheral tissues J Biol Chem 1991 266 583 593 1985917
Simard J de Launoit Y Labrie F Characterization of the structure-activity relationships of rat types I and II 3β-hydroxysteroid dehydrogenase/Δ5 - Δ4isomerase by site-directed mutagenesis and expression in HeLa cells J Biol Chem 1991 266 14842 14845 1831194
Bain PA Yoo M Clarke T Hammond SH Payne AH Multiple forms of mouse 3β-hydroxysteroid dehydrogenase/Δ5-Δ4 isomerase and differential expression in gonads, adrenal glands, liver, and kidneys of both sexes Proc Natl Acad Sci 1991 88 8870 8874 1924345
Abbaszade IG Arensburg J Park C-H Kasa-Vubu JZ Orly J Payne AH Isolation of a new mouse 3β-hydroxysteroid dehydrogenase isoform, 3β-HSD VI, expressed during early pregnancy Endocrinology 1997 138 1392 1399 9075693 10.1210/en.138.4.1392
Trudel C Couet J Martel C Labrie C Labrie F Regulation of adrenal 3β-hydroxysteroid dehydrogenase/Δ5-Δ4-isomerase expression and activity by adrenocorticotropin and cortisone in the rat Endocrinology 1991 129 2077 2084 1655393
Miller WL Mitochondrial specificity of the early steps in steroidogenesis J Steroid Biochem Mol Biol 1995 55 607 616 8547188 10.1016/0960-0760(95)00212-X
Sauer LA Chapman JC Dauchy RT Topology of β-hydroxy-5-ene-steroid dehydrogenase/Δ5-Δ4-isomerase in adrenal cortex mitochondria and microsomes Endocrinology 1994 134 751 759 8299570 10.1210/en.134.2.751
| 15804366 | PMC1087504 | CC BY | 2021-01-04 16:37:12 | no | Reprod Biol Endocrinol. 2005 Apr 3; 3:11 | utf-8 | Reprod Biol Endocrinol | 2,005 | 10.1186/1477-7827-3-11 | oa_comm |
==== Front
Reprod Biol EndocrinolReproductive biology and endocrinology : RB&E1477-7827BioMed Central London 1477-7827-3-121582900410.1186/1477-7827-3-12ResearchUltrasound image attributes of human ovarian dominant follicles during natural and oral contraceptive cycles Birtch Rebecca L [email protected] Angela R [email protected] Olufemi A [email protected] Roger A [email protected] Department of Obstetrics, Gynecology and Reproductive Sciences, College of Medicine, University of Saskatchewan, Saskatoon, Saskatchewan, Canada2005 13 4 2005 3 12 12 9 2 2005 13 4 2005 Copyright © 2005 Birtch et al; licensee BioMed Central Ltd.2005Birtch 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
Computer-assisted analyses were used to examine ultrasound image attributes of human dominant ovarian follicles that developed during natural and oral contraceptive (OC) cycles. We hypothesized that image attributes of natural cycle follicles would quantitatively differ from those in OC cycles and that OC cycle follicles would possess image attributes indicative of atresia.
Methods
Dominant ovarian follicles of 18 clinically normal women were compared using transvaginal ultrasonography for the 7 days before ovulation during a natural cycle (n = 9) or the 7 days before peak estradiol in women using OC (n = 11). Follicles were analyzed using region and line techniques designed to compare the image attributes numerical pixel value (NPV), pixel heterogeneity (PH) and area under the curve (AUC).
Results
NPV was higher in OC cycle follicles with region analysis and tended to be higher with line analysis (p = 0.005 and p = 0.06, respectively). No differences were observed in two other image attributes (AUC and PH), measured with either technique, between natural and OC cycle follicles.
Conclusion
The increased NPV value of OC cycle follicles and lack of differences in PH and AUC values between natural cycle and OC cycle follicles did not support the hypothesis that OC cycle follicles would show ultrasonographically detectable signs of atresia. Image attributes observed in OC cycle follicles were not clearly indicative of atresia nor were they large enough to preclude preovulatory physiologic status in OC cycle follicles.
==== Body
Background
Diagnostic gray-scale ultrasonography has revolutionized the study of ovarian biology in animals and humans because it allows researchers and clinicians to assess the development of individual follicles in a direct, non-invasive, and atraumatic manner without interruption or distortion of ovarian function. Prior to the introduction of ultrasonography, histological slices of ovarian tissue were used to elucidate ovarian follicular development; however, histologic investigation only provides information about a single time point and does not permit assessment of follicular function over time. Further, histology cannot be used for time-series studies in humans.
Animal models have been developed to elucidate the basic mechanisms of ovarian function in humans and to overcome ethical impossibilities of some aspects of research in humans. To date, however, no appropriate animal models are available to elucidate the physiological effect of oral contraceptives (OC) on human ovarian function due to species specific differences in the metabolism of the exogenous estrogen and progestins used to control reproductive function. Nor do non-invasive techniques exist that will allow for the determination of the physiological status of individual ovarian follicles with a single observation. However, new technologies involving computer assisted image analysis to elucidate a follicle's physiologic status show promise [1,2].
Quantitative changes in ultrasound image echotexture as indicators of physiological function of ovarian structures have been described in domestic animal models [3-8]. The validity of the image analysis technique has been verified through correlation of ultrasound image attributes with histologic attributes [3,4,8]. Similar studies in humans are ethically impossible; however, information generated in animal studies may be applied to human imaging based studies [3-6,9,10]. The overall goal of this line of research in our laboratory is to elucidate physiologic status of dominant follicles with non-invasive ultrasonography in humans. Imaging-based techniques that could be used to determine follicular health would obviate the ethical and logistical limitations associated with ovarian function research in women.
It has recently been discovered that women grow follicles in two or three follicular waves during each natural menstrual cycle [11,12]. The pattern of folliculogenesis is similar in women to those observed in several species of domestic animals (bovine, equine, caprine and ovine) [9,13-17]. A follicular wave is defined as a cohort of follicles that enter the antral growth phase synchronously. Growth of all follicles in the cohort continues until one follicle is physiologically selected as the dominant follicle. The dominant follicle continues its development to pre-ovulatory diameter while the remaining follicles in the cohort undergo atresia. The dominant follicle will ovulate if the appropriate hormone signals (i.e., mid-cycle luteinizing hormone surge) are provided. If the hormone signals which trigger ovulation are not provided, the dominant follicle enters a static phase and remains approximately the same diameter until it enters the regressing phase, when it decreases in diameter until it is no longer detectable.
Dominant follicles develop in women during compliant use of OC, with most dominant follicles initiating growth during the hormone-free interval [18-22]. Ovulatory follicles that develop during spontaneous natural cycles are presumed to be healthy because they typically ovulate. It is not known whether dominant follicles of ostensibly preovulatory diameter that arise during OC use have the same physiologic status and/or ovulatory capacity as natural preovulatory follicles.
The primary objective of the present study was to assess ultrasonographic image attributes of dominant ovarian follicles during the final stages of development in natural menstrual cycles and OC cycles. We hypothesized that image attributes of follicles from natural cycles would quantitatively differ from OC cycle follicles due to the effects of the exogenous hormones on OC cycle follicles and that the image attributes of OC cycle follicles would be consistent with those indicative of atresia.
Materials and Methods
This study was a retrospective, observational study designed to evaluate and compare ultrasound image attributes of dominant ovarian follicles (≥ 10 mm in diameter) during natural and OC cycles. Images of twenty (n = 20) dominant follicles were analyzed. The images were obtained in two previous studies designed to characterize ovarian follicular wave dynamics during spontaneous natural cycles [11,12] and OC cycles [22]. The inclusion and exclusion criteria of both studies were similar. Participants were assessed, by history and physical examination, to be healthy women of reproductive age (28.0 ± 0.14 years, natural cycles; 24.5 ± 0.02 years, OC cycles; mean ± SEM). All women in both studies had a history of normal menstrual cycles and could not have used OC for a minimum of three months before participating. In the OC trial, one of three different OC formulations were administered to thirty six women for three consecutive 28-day cycles; (i) 20 μg ethinyl estradiol (21)/100 μg levnogestrel (21) (n = 7), (ii) 30 μg ethinyl estradiol (21)/150 μg desogetrel (21) (n = 2), or (iii) 35 μg ethinyl estradiol (21)/180 μg norgestimate (7)/215 μg norgestimate (7)/250 μg norgestimate (7) (n = 2). Nine women grew 11 dominant follicles during compliant OC use. Two women grew dominant follicles during two separate cycles of OC-use; therefore, images from eleven dominant, ostensibly, ovulatory follicles were analyzed in the present study [22]. Images of nine dominant follicles from nine volunteers were randomly selected from natural cycle data to act as controls [11,12]. Blood samples were drawn every third day for women in the natural cycle trial and every second day once a follicle reached ≥ 14 mm diameter for women in the OC cycle trial.
A high resolution ultrasound instrument equipped with 5–9 MHz multi-frequency intravaginal convex array transducer (ATL Ultramark HDI 5000, Advanced Technologies Laboratories, Bothwell, WA, USA) was used. Settings of the ultrasound instrument that affect image attributes (beam focus, overall time-gain, near-field and far-field gain) were standardized to the same predetermined standards for both studies. Each image was digitally acquired and transferred into a customized computer database during the ultrasound examinations. Images of the dominant follicle acquired in largest cross-sectional diameter with the fewest image artifacts were selected for image analysis. All image analyses were performed by the same individual (RLB). Blinding was not possible because each image contained identifying study information.
Image attributes of dominant follicles were analyzed using a graphics workstation equipped with a customized software (SYNERGYNE 2©, Saskatoon, SK, Canada) integrating complex algorithms designed for ultrasonographic image analysis [1,2]. Each image was analyzed with two different techniques: i) region analysis and ii) line analysis designed to quantify gray-scale values of selected regions [1,2]. Three image attributes were quantified: i) numerical pixel value (NPV) defined as the mean pixel gray-scale value of the sampled pixels, ii) pixel heterogeneity (PH) defined as the standard deviation of the mean gray-scale values of the sampled pixels and, iii) area under the curve (AUC) defined as the area of the sampled region in pixels.
Briefly, region analyses involve overlaying a computer-generated grid onto a selected area of the 2-dimensional image of a follicle to generate a 3-dimensional framework representing processed pixel intensities. Placing a computer-generated opaque "film" or blanket over the wire framework representing pixel values comprising the ultrasound image yields a 3-dimensional contoured surface. This technique allows discriminatory examination of the surfaces and rapid visual assessment of ultrasonographic attributes associated with follicle health (state of viability or atresia); [2]. Region analyses were used to measure NPV, PH and AUC within a region of the follicle wall. A one-pixel-wide line was used to outline and isolate the largest continuous portion of follicular wall extending from the peripheral antrum to the stroma with the fewest image artifacts. The antrum-wall interface was defined as the last pixel along the line in which a sequential rise in gray-scale occurred [4].
Line analysis places a line across a specified section of the image of the follicle to produce a graph of the pixel intensities along the line displayed. The graph depicts the amplitude of the echoes located along the line [2]. Line analysis was used to measure NPV, PH and AUC of a single line across the follicle wall from the antrum-wall interface to the ovarian stroma using two straight three-pixel wide lines. The two linear measurements were taken between the 4 and 9 o'clock positions of the follicle image.
Ovulation was defined as the disappearance of a follicle greater than 15 mm in diameter identified the previous day and confirmed with visualization of a corpus luteum [11,12,23,24]. Ovulation was used as a reference point (Day 0) to standardize the data for natural cycle follicles. No follicles ovulated during OC cycles; therefore, OC cycle follicles were standardized to the day of peak estradiol concentration (Day 0) which was also defined as the last day of the growth phase. The day of peak estradiol concentration for each individual was determined by reviewing the estradiol profiles for each woman in the OC trial [22].
Ultrasound images were available for each of the seven days leading to ovulation for women in the natural cycle study and approximately every second day for seven days leading to peak estradiol levels for women participating in the OC trial. Data were standardized by aligning data from Day 0 as defined for each group. Image attributes on each of the seven days (Days -1 to -7) leading to ovulation during the natural cycle were compared to the corresponding days (Day -1 to -7) leading to peak estradiol concentrations during the OC cycle. Data were inadequate for statistical comparison of image attributes among the three different OC formulations, therefore data were combined.
Image attributes were compared using repeated measures analysis of variance (PROC MIXED, SAS/STAT, v8) for main effects of time and follicle type. Significance was set at p < 0.05. Results are expressed as the mean ± SEM.
Results
Peak estradiol-17β concentrations for OC cycle follicles were 170.1 ± 30.4 pg/mL (range, 35.1 – 364.5 pg/mL) [22]. Mean follicle diameter of OC cycle follicles at peak estradiol concentrations was 20.3 ± 2.0 mm. Mean estradiol concentrations for natural cycle follicles on Day -1 were 190.8 ± 19.6 pg/mL (range 20.0 – 340 pg/mL) [11]. Mean follicle diameter of natural cycle follicles on Day -1 was 21.1 ± 1.1 mm. Mean growth profiles for natural and OC cycle follicles are displayed in Figure 1. Ultrasonographic images of a natural cycle follicle one day before ovulation and an OC cycle follicle on the day of peak estradiol are shown (Figure 2,a and 2b, respectively).
Figure 1 Growth profiles of natural and OC cycle follicles
Figure 2 Ultrasonographic images of a natural cycle and OC cycle follicles. (a) Natural cycle follicle on Day 0, (b) OC cycle follicle on Day 0.
Numerical pixel values were higher (p = 0.005) for OC cycle follicles versus natural cycle follicles using region analyses and tended to be higher (p = 0.06) using line analysis for the 7 days prior to Day 0. Oral contraceptive cycle follicles had visually similar NPV for both region and line analyses; the values were initially high and then decreased to a nadir at Day -3 (region analysis) or Day -4 (line analysis) after which they increased until Day -1 and subsequently decreased on Day 0 (Figure 3a,b). Numerical pixel values for natural cycle follicles progressively increased from Day -7 (Region analysis; 31.64 ± 4.26, Line analysis; 41.66 ± 10.49) to Day 0 using line analysis (64.03 ± 6.37; p = 0.01) and tended to increase (43.80 ± 5.30; p = 0.08) with region analysis (Figure 3a,b).
Figure 3 Graphical representation of NPV, PH and AUC. Mean ± SEM (a, b) numerical pixel value (c,d) pixel heterogeneity and (e, f) area under the curve obtained by region analysis (a, c, e) and line analysis (b, d, f) of natural cycle and OC-cycle follicles.
Pixel heterogeneity values were not different between natural and OC cycle follicles for either region or line analyses (p = 0.113 and p = 0.515, respectively). No changes were observed over the seven days prior to Day 0.
Area under the curve evaluated using line analyses tended to be higher (p < 0.09) in OC cycle follicles; however, no differences were observed (p = 0.23) in follicle type using region analyses. Values for AUC for both natural and OC cycles increased as the interval to Day 0 decreased.
Nadirs in image attributes for the values representing AUC and PH were observed on Day -4 (Figure 3c–f).
Discussion
The study of ovarian follicular dynamics in women taking hormonal contraceptives is a relatively new research area with profound clinical relevance. Follicular development in up to 50% of new OC cycles and up to a 25% ovulation rate have been reported [18]. Most follicular development observed in OC cycles is initiated during the hormone free interval of the OC cycle [22]. Our objective was to assess the indicators of follicular physiological status in women by comparing ultrasound image attributes of human dominant ovarian follicles leading to ovulation in natural cycles or peak estradiol concentrations in OC cycles. Numerical pixel values representing the relative brightness of the elements comprising the ultrasound image were the only image attributes that differed between natural and OC cycle follicles. Morphologic changes indicative of atresia in animal models are typically represented by high values of NPV [4,6]. These morphologic changes include, but are not limited to, an increase in the number of multivesicular bodies, lipid droplets and vacuoles within granulosa cells, dilation of the smooth endoplasmic reticulum and golgi apparatus and thecal cell hypertrophy/degeneration; which, when taken together, appear to affect the tissue properties enough to change the image attributes [25,26]. Although NPV were greater in OC cycle follicles, the increased values were not high enough to indicate poor physiologic status as described in studies using animal models [4,6]. The hypotheses that image attributes of natural and OC cycle would quantitatively differ and that OC cycle follicles would possess image attributes characteristic of atresia were only partially supported.
An interesting common observation in all image attributes of OC follicles that we evaluated fell to a nadir at Day -4, suggesting that an important physiological event occurs approximately four days before peak estradiol concentrations were attained. The image attributes of the values reflected in NPV and AUC in both natural and OC cycle follicles tended to increase from Day -4 to Day 0. This observation may be interpreted to mean that the follicles in both natural and OC cycles have similar physiologic status and undergo similar changes during the final stages of development. Classically described, ovulatory changes include thickening of the follicle wall, development of lipid inclusion granulosa and theca cells, hypertrophy of theca cells and increased vascularization of the theca layer [27,28]. Taken together, these developmental changes may be reflected as increases in AUC as the follicle wall thickens and increases in NPV as the follicle wall develops to fulfill its biological functions. Further research is needed to determine the physiologic changes responsible for changes in image attributes.
No OC cycle follicles analyzed in the present study ovulated. However, we suggest that follicles which develop during compliant OC use may differ in ovulatory potential because ovulation has been documented in up to 50% of follicles which develop to ostensibly ovulatory diameter during compliant OC use [18]. Differences in OC formulations (i.e. estradial concentrations and progestin type and associated activity) may have different physiological effects on follicular development and subsequent ovulatory ability, and individual responses to the different estrogen levels and the various progestins used in OC formulations may affect ovulatory ability differently. We were unable to test this hypothesis directly in the present study. Variations in image attributes of OC cycle follicles may be due to differences in physiologic status and ovulatory ability. Further research to determine the physiological mechanisms responsible for differences in ovulatory ability of OC cycle follicles needs to be completed.
Studies in domestic animals have demonstrated that ultrasound image attributes of ovarian structures are related to their physiologic function [3,4,8]. The present study provides rationale for exploring the hypothesis that the same is true for humans. In order to have a complete understanding of the relationship between image attributes and follicular development in humans, we must assess all stages of follicular development throughout the menstrual cycle, prediction of dysfunctional follicular development (i.e., hemorrhagic anovulatory follicles) and the effects of different OC formulations on follicle development. Once the association between image attributes and various conditions of follicular development are determined, follicular status could be assessed with a single ultrasound examination, obviating many ethical constraints that currently prevent progress in human ovarian follicular research.
Ovarian follicles provide a unique endocrine environment fundamental to normal oocyte development and competence. Abnormal follicular development can lead to incompetent oocytes; therefore it follows that image attributes of the follicle wall may be related to oocyte competence. Currently, image attributes only provide us with information about the physiologic status of the follicular wall. Image analyses do not provide information about the health and viability of the oocyte; however, this area is under active investigation [7].
The primary objective of this research was to evaluate and compare ultrasound image attributes of dominant follicles that develop during natural cycles and OC cycles. We were not able to examine the effect of different OC formulations on image attributes due to constraints imposed by study design. It is hoped that future research will explore the effects of different OC formulations on the image attributes of dominant follicles. Research to evaluate physiologic status may well hinge upon follicular response to administration of ovulation inducing doses of recombinant human chorionic gonadotrophin (hCG) in women who develop dominant follicles during OC cycles in concert with development of an animal model to test the biological effects of varying doses of ethinyl estradiol or different progestins on follicle and oocyte status.
Conclusion
In conclusion, similarities in image attributes between natural and OC cycle follicles provides preliminary evidence that ultrasound image attributes of human follicles are associated with physiologic status during the growth phase and that follicles which develop during OC cycles have image attributes similar to those of natural cycle follicles. Only one of the three image attributes studied differed between natural and OC cycle follicles. The changes observed in OC cycle follicles were not clearly indicative of atresia nor were they large enough to preclude preovulatory physiologic status in OC cycle follicles.
Acknowledgements
The authors would like to thank the research volunteers who participated in the original research utilized in the current study. Appreciation is also expressed to John Deptuch, in the Department of Obstetrics, Gynecology and Reproductive Sciences at the University of Saskatchewan for his technical guidance and support. Funding for this project was provided by the Canadian Institutes of Health Research.
==== Refs
Singh J Adams G Pierson R Promise of New Imaging Technologies for Assessing Ovarian Function Animal Reproduction Science 2003 78 371 399 12818654 10.1016/S0378-4320(03)00100-3
Pierson R Adams G Computer-Assisted Image Analysis, Diagnostic Ultrasonography and Ovulation Induction: Strange Bedfellows Theriogenology 1995 43 105 112 10.1016/0093-691X(94)00014-L
Singh J Pierson R Adams G Ultrasound Image Attributes Of The Bovine Corpus Luteum: Endocrine And Functional Correlates Journal of Reproduction and Fertility 1997 109 35 44 9068411
Singh J Pierson R Adams G Ultrasound Image Attributes of Bovine Ovarian Follicles: Endocrine and Functional Correlates Journal of Reproduction and Fertility 1998 112 19 29 9538326
Tom J Pierson R Adams G Quantitative Echotexture Analysis of Bovine Corpora Lutea Theriogenology 1998 40 1345 1352 10732071 10.1016/S0093-691X(98)00081-8
Tom J Pierson R Adams G Quantitative Echotexture Analysis Of Bovine Ovarian Follicles Theriogenology 1998 50 339 346 10732129 10.1016/S0093-691X(98)00143-5
Vassena R Adams G Mapletoft R Pierson R Singh J Ultrasound Image Characteristics of Ovarian Follicles in Relation to Oocyte Competence and Follicular Status in Cattle Animal Reproduction Science 2003 76 25 41 12559718 10.1016/S0378-4320(02)00234-8
Duggvathi R Bartlewski P Pierson R Rawlings N Luteogenesis in Cyclic Ewes: Echotextural, Histological, and Functional Correlates Biology of Reproduction 2003 69 634 639 12724274 10.1095/biolreprod.103.016576
Ginther O Gastal E Bergfelt D Baerwald A Pierson R Comparative Study of the Dynamics of Follicular Waves in Mares and Women Biology of Reproduction 2004 71 1195 1201 15189824 10.1095/biolreprod.104.031054
Adams G Pierson R Bovine Model for Study of Ovarian Follicular Dynamics in Humans Theriogenology 1995 43 113 120 10.1016/0093-691X(94)00015-M
Baerwald A Adams G Pierson R Characterization of Ovarian Follicular Wave Dynamics in Women Biology of Reproduction 2003 69 1023 1031 12748128 10.1095/biolreprod.103.017772
Baerwald A Adams G Pierson R A New Model for Ovarian Follicular Development during the Human Menstrual Cycle Fertility and Sterility 2003 80 116 120 12849812 10.1016/S0015-0282(03)00544-2
Ginther O Kastelic J Knopf L Composition and Characteristics of Follicular Waves during the Bovine Oestrus Cycle Animal Reproduction Science 1989 20 187 200 10.1016/0378-4320(89)90084-5
Sirois J Fortune J Ovarian Follicular Dynamics during the Estrous Cycle in Heifers Monitored by Real-time Ultrasonography Biology of Reproduction 1988 139 308 317 3052602
Bister J Noel B Perrad B Mandiki S Mbayahaga J Paquay R Control of Ovarian Follicles Activity in the Ewe Domestic Animal Endocrinology 1999 17 315 328 10527133 10.1016/S0739-7240(99)00047-8
Donadeu F Ginther O Follicular Waves and Circulating Concentrations of Gonadotropins, Inhibin, and Oestradiol during the Anovulatory Season in Mares Reproduction 2002 124 875 885 12530925 10.1530/rep.0.1240875
Evans A Characteristics of Ovarian Follicle Development in Domestic Animals Reproduction in Domestic Animals 2003 38 240 246 12887563 10.1046/j.1439-0531.2003.00439.x
Pierson R Archer D Moreau M Shangold G Fisher A Creasy G Ortho Evra/ Evra versus Oral Contraceptives: Follicular Development and Ovulation in Normal Cycles and After and Intentional Dosing Error Fertility and Sterility 2003 80 34 32 12849799 10.1016/S0015-0282(03)00556-9
Hoogland J Skouby S Ultrasound Evaluation of Ovarian Activity under Oral Contraceptives Contraception 1993 47 583 590 8334892 10.1016/0010-7824(93)90025-3
Killick S Eyong E Elstein M Ovarian Follicular Development in Oral Contraceptive Cycles Fertility and Sterility 1987 48 409 413 3114009
Van Heusden A Fauser B Activity of the Pituitary-Ovarian Axis in the Pill-Free Interval during use of Low-Dose Combined Oral Contraceptives Contraception 1999 59 237 243 10457868 10.1016/S0010-7824(99)00025-6
Baerwald A Olatunbosun O Pierson R Ovarian Follicular Development Is Initiated During The Hormone-Free Interval Of Oral Contraceptive Use Contraception 2004 70 371 377 15504375 10.1016/j.contraception.2004.05.006
Hanna M Pierson R Ultrasonographic morphology of the human preovulatory follicle wall Ultrasound International 1999 5 5 13
Pierson R Martinuk S Chizen D Simpson C Evers J, Heineman M Ultrasonographic visualization of the human ovulation Proceeding of the 7th Reinier de Graaf Symposium 1990 Maastricht, The Netherlands: Elsevier Publishers (Biomedical Division) 73 79
Billig H Chun S-Y Eisenhauser K Hsueh A Gonadal Cell Apoptosis: Hormone-Regulated Cell Demise Human Reproduction Update 1996 2 103 117 9079407 10.1093/humupd/2.2.103
Hurwitz A Adashi E Adashi E, Leung P Ovarian Follicular Atresia as an Apoptotic Process The Ovary 1993 New York: Raven Press; 473 486
Gougeon A Some Aspects of the Dynamics of Ovarian Follicular Growth in the Human Acta Europaea Fertilitatis 1989 20 185 192 2519570
Gougeon A Dynamics of Follicular Growth in the Human: A Model from Preliminary results Human Reproduction 1986 1 81 87 3558758
| 15829004 | PMC1087505 | CC BY | 2021-01-04 16:37:12 | no | Reprod Biol Endocrinol. 2005 Apr 13; 3:12 | utf-8 | Reprod Biol Endocrinol | 2,005 | 10.1186/1477-7827-3-12 | oa_comm |
==== Front
Respir ResRespiratory Research1465-99211465-993XBioMed Central London 1465-9921-6-261580435610.1186/1465-9921-6-26ResearchAirway epithelial cell tolerance to Pseudomonas aeruginosa Wu Qi [email protected] Zhong [email protected] Margrith W [email protected] Scott H [email protected] Cystic Fibrosis/Pulmonary Research and Treatment Center, Department of Medicine, The University of North Carolina, Chapel Hill, NC 27599, USA2 Department of Cellular and Molecular Physiology, The University of North Carolina, Chapel Hill, NC 27599, USA2005 1 4 2005 6 1 26 26 21 12 2004 1 4 2005 Copyright © 2005 Wu et al; licensee BioMed Central Ltd.2005Wu 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 respiratory tract epithelium is a critical environmental interface that regulates inflammation. In chronic infectious airway diseases, pathogens may permanently colonize normally sterile luminal environments. Host-pathogen interactions determine the intensity of inflammation and thus, rates of tissue injury. Although many cells become refractory to stimulation by pathogen products, it is unknown whether the airway epithelium becomes either tolerant or hypersensitive in the setting of chronic infection. Our goals were to characterize the response of well-differentiated primary human tracheobronchial epithelial cells to Pseudomonas aeruginosa, to understand whether repeated exposure induced tolerance and, if so, to explore the mechanism(s).
Methods
The apical surface of well-differentiated primary human tracheobronchial epithelial cell cultures was repetitively challenged with Pseudomonas aeruginosa culture filtrates or the bacterial media control. Toxicity, cytokine production, signal transduction events and specific effects of dominant negative forms of signaling molecules were examined. Additional experiments included using IL-1β and TNFα as challenge agents, and performing comparative studies with a novel airway epithelial cell line.
Results
An initial challenge of the apical surface of polarized human airway epithelial cells with Pseudomonas aeruginosa culture filtrates induced phosphorylation of IRAK1, JNK, p38, and ERK, caused degradation of IκBα, generation of NF-κB and AP-1 transcription factor activity, and resulted in IL-8 secretion, consistent with activation of the Toll-like receptor signal transduction pathway. These responses were strongly attenuated following a second Pseudomonas aeruginosa, or IL-1β, but not TNFα, challenge. Tolerance was associated with decreased IRAK1 protein content and kinase activity and dominant negative IRAK1 inhibited Pseudomonas aeruginosa -stimulated NF-κB transcriptional activity.
Conclusion
The airway epithelial cell response to Pseudomonas aeruginosa entails adaptation and tolerance likely mediated, in part, by down-regulation of IRAK1.
==== Body
Background
The innate immune system suppresses pathogen attachment, colonization, growth, and invasion and co-ordinates adaptive immunity [1]. Innate immunity entails recognition of microbial signatures by the cellular repertoire of Toll-like receptors (TLRs [2,3]). TLR agonists initiate receptor-specific downstream signaling pathways, ultimately enhancing production of anti-microbial molecules and inflammatory mediators [4]. Much data has been derived from monocyte/macrophages and myelocytic cell lines, but the TLR pathway also functions in epithelial cells where receptor and co-receptor expression levels, and the activity of downstream signal transduction intermediates, likely determine cellular sensitivity to pathogen products [5-7]. In the human airway, polymorphisms in TLR4, the LPS receptor, modulated the response to inhaled LPS, and TLR4 was found on primary tracheobronchial epithelial (hTBE) cells [8]. RNA for TLR1-6 was present in hTBE cells in vitro, and high doses of commercial LPS activated NF-κB and induced the neutrophil chemotactic cytokine IL-8 and the anti-microbial peptide beta defensin 2 (hBD-2) [9]. More recent studies demonstrate TLR2-dependent IL-8 and hBD-2 production by hTBE cells [10]. Culture filtrates of both Gram-positive and -negative bacteria and a TLR2 agonist enhanced IL-8 secretion by hTBE cells 3–5 fold [11]. Hemophilus influenzae, an important respiratory tract pathogen, signals via TLR2 in epithelial cells [12]. Recent studies indicate a role for TLR2 and TLR5 in stimulation of airway epithelial cells by flagellin or live Gram-positive and -negative bacteria [13,14] and TLR2 was apparently recruited to lipid rafts at the apical epithelial cell surface [15]. TLR1-10 expression and positive responses to several TLR agonists were recently reported in airway epithelial cells[16]. Thus, the TLR signal transduction pathway is likely an important regulator of airway immunity and inflammation.
The regulation of TLR signaling is dynamic. Up-regulation of TLRs, for example by interferon [17] or virus [18], may enhance epithelial cell sensitivity to pathogen products. On the other hand, LPS exposure induces hypo-responsiveness to a second challenge, termed LPS tolerance (reviewed in [19]). Other molecules acting through the TLR pathway, including mycobacterial products [20] and lipoteichoic acid from Gram-positive bacteria [21], also induce tolerance. Tolerance is associated with decreased degradation of NF-κB inhibitory proteins, reduced MAP kinase phosphorylation, prevention of NF-κB and AP-1 activation, altered transcriptional responses and suppression of pro-inflammatory cytokine and chemokine production [22]. Decreased cell surface TLR4 protein [23] was associated with tolerance, but tolerance could not be attributed solely to receptor loss, since cells that did not decrease TLR4 still became tolerant, and over-expression of CD14, TLR4 and MD-2 in HEK293 cells did not prevent LPS tolerance [20,24]. However, a hypo-responsive state can be induced at the level of the plasma membrane by the expression of endogenous, functionally inactive members of the Toll-interleukin 1 receptor superfamily [25]. Many substances acting though the TLR signal transduction pathway induce cross tolerance, including LPS and IL-1β [22], LPS and mycobacterial products [20], or LPS and lipoteichoic acid [21]. Tolerance without down-regulation of surface receptors and cross-tolerance suggest negative regulation of common elements in the downstream signal transduction pathway. IRAK1 functions just distal to TLRs and their adaptor proteins [3], and tolerance is associated with decreased IRAK1 protein many cell types [26-31]. Alternatively, signaling through IRAK1 may be impaired due to decreased TLR4-MyD88 complex formation [32], lack of dissociation from the receptor complex [33], or increased function of inhibitory forms of IRAK such as IRAK-M [34]. Hypo-responsiveness may also be due to events closer to activation of transcription factors, for example, abrogation of IκBα polyubiqitination [35], over-expression of unique IκB inhibitory proteins [36], or production of NF-κB p50 homo-dimers [37], or other factors that block NF-κB DNA binding [38]. These diverse negative regulatory processes may be generally important to protect the host from overly exuberant, destructive inflammatory responses.
In cystic fibrosis (CF), the lack of functioning CFTR impairs mucociliary and cough clearance [39], forming a nidus for infection and allowing organisms such as Pseudomonas aeruginosa (Ps. a.) to evade host defenses. Inability to clear infected mucus results in continuous exposure of airway epithelial cells to bacteria and their products. Ongoing host-pathogen interactions determine the extent of the inflammatory response, and in turn, rates of tissue destruction and loss of pulmonary function. Strategically located between luminal bacterial masses and the host circulation, the airway epithelium is in a key position to regulate inflammation, but it remains unknown whether the airway epithelium becomes either hypersensitive or tolerant to the chronic presence of bacterial products. Our goals were to characterize the response of well-differentiated, primary hTBE cells to an apical surface challenge with Ps. a. products, to determine whether repeated exposure induced tolerance and, if so, to explore the mechanism(s). We show that Ps. a. products activate the TLR pathway and that hTBE cells become tolerant via a mechanism likely involving down-regulation of IRAK1.
Methods
Reagents
Antibodies against phosphorylated or total c-jun NH2-terminal kinases (JNK), p38, extracellular signal-regulated kinases (ERK), and total IκBα were from Cell Signaling Technology (Beverly, MA). Antibody against IRAK1 was from Upstate Biotechnology, Inc. (Lake Placid, NY). Anti-NF-κB p50 and p65 subunits were from Santa Cruz Biotechnology (Santa Cruz, CA). An ELISA kit for IL-8 (DuoSet ELISA Development System) was from R&D Systems (Minneapolis, MN). An in vitro toxicology assay kit for lactate dehydrogenase (LDH) was from Sigma (St. Louis, MO), as were other standard reagents unless otherwise specified.
Preparation of Ps. a. filtrate
Ps. a. strain ATCC 27853 was grown in trypticase soy broth (TSB) for 72 hours at 37°C with shaking at 250 RPM. Following centrifugation at 5,500 × G (4°C) for 30 minutes, the supernatant was 0.45 μm filtered, aliquoted and stored at -20°C. TSB treated similarly was used as a control. When used in experiments with cell lines cultured on plastic, Ps. a. filtrates or TSB were boiled for 10 minutes to eliminate protease activity. Consistent with prior reports [40], we found IL-8 stimulatory activity in Ps. a. filtrates to be heat-resistant.
Cell culture
Under an Institutional Review Board-approved protocol, hTBE cell cultures were prepared as previously described[41]. Briefly, epithelial cells were removed from the lower trachea and bronchi by protease XIV digestion and cells were plated in BEGM medium on collagen-coated dishes. Passage 2 cells were cultured on type VI collagen (Sigma) coated Millicell CM inserts (0.4 μM pore size, Millipore Corporation, Bedford, MA) in ALI medium. The cell seeding density for 10 mm and 30 mm diameter inserts was 0.15 × 106 and 1 × 106 cells per insert, respectively. Following confluence after 5–7 days, cultures were maintained with an air-liquid interface until well-differentiated and were used at 21 days. Endotoxin in ALI medium was less than 100 pg/ml (LAL assay, Bio-Whittaker, Walkersville, MD). A recently described immortalized cell line, referred to as AALEB, was derived from hTBE cells by infection with retroviruses expressing SV40 early region and telomerase reverse transcriptase [42]. AALEB cells were grown on plastic dishes in BEGM medium under standard culture conditions.
Ps. a. filtrate challenge
Experiments with well-differentiated hTBE cells on 12 and 30 mm Millicell inserts were performed in 12 or 6 well plates, respectively. Before challenge, the apical culture surface was rinsed once with Dulbecco's PBS. Ps. a. filtrate in ALI medium supplemented with 10% human serum (Sigma #H4522) was added to the apical culture surface, using 100 μl for 12 mm inserts and 500 μl for 30 mm inserts. One or two ml of ALI medium was added to the basolateral side of the 12 or 6 well plates, respectively, and the cultures were incubated at 37°C in 5% CO2 for 24 hours. Following removal of basolateral medium for IL-8 or LDH assay, the apical and basolateral surfaces were washed twice with PBS. After incubation at 37°C for 1–2 hours with fresh basolateral ALI medium, the cultures were re-challenged apically with Ps. a. filtrate plus serum as described above. Challenges with IL-1β (10 ng/ml) or TNFα (25 ng/ml) were performed in ALI medium without serum. IL-8 and LDH assays were performed with commercial kits as specified previously [11] and are based on results from triplicate wells using cells from at least three different individuals, unless stated otherwise.
Western blot analysis
At specified times following challenge, cells were harvested from 30 mm inserts into ice-cold lysis buffer (100 mM TrisHCl pH 8.0, 100 mM NaCl, 5.0 mM NaF, 2 mM EDTA, 1% NP-40, 1 mM Na3VO4, 100 μM TPCK, 100 μM quercetin, 1 mM PMSF, 1 μg/ml leupeptin, and 1 μg/ml pepstatin) using a cell scraper, transferred to tubes and set on ice for 20 minutes. Following centrifugation, protein concentrations were determined using the BCA Protein Assay Reagent (Pierce, Rockford, IL). Samples were resolved by SDS-PAGE (4–20% tris-glycine gels, Invitrogen, San Diego, CA) and blotted onto Immobilon-P membranes (Millipore Corp., Bedford, MA). Blots were blocked in TBS with 0.05% Tween 20 and 5% dry milk powder, incubated with primary then secondary antibodies (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA) followed by chemiluminescence detection of peroxidase (Pierce).
Nuclear extracts
Cells were scraped from 30 mm inserts into 0.8 ml ice-cold PBS containing protease inhibitors (1 μM PMSF, 1 μg/ml leupeptin, 1 μg/ml pepstatin, and 0.5 μM DTT) and were centrifuged. The cell pellet was re-suspended in buffer (10 mM Hepes, pH 7.9, 10 mM KCl, 1.5 mM MgCl2, 0.2 mM EDTA, 0.5 mM DTT, 0.25% NP-40, 1 μM PMSF, 1 μg/ml leupeptin and 1 μg/ml pepstatin) on ice for 10 minutes and cells were lysed with a Dounce homogenizer (Kontes Scientific Glassware, Vineland, NJ). The nuclei were extracted with high salt buffer (20 mM Hepes, pH 7.9, 450 mM NaCl, 1.5 mM MgCl2, 0.2 mM EDTA, 0.5 mM DTT, 25% glycerol, 1 μM PMSF, 1 μg/ml leupeptin and 1 μg/ml pepstatin) on ice for 20 minutes with occasional vortexing. Supernatants were prepared by centrifugation and protein concentration was determined as above.
Electrophoretic mobility shift assays
NF-κB-specific consensus oligonucleotide (5' AGTTGAGGGGACTTTCCCAGGC3') and AP1-specific consensus oligonucleotide (5' CGCTTGATGAGTCAGCCGGAA3') were from Promega (Madison, WI). DNA probes were 32P end labeled with T4 polynucleotide kinase (Promega). Nuclear extracts (2.5 μg) were incubated with 40,000–60,000 cpm of 32P end labeled oligonucleotide probe in binding buffer (final volume of 10 μl) containing 1 μg poly dI-dC (Sigma), 10 mM TrisHCl, pH 7.9, 50 mM KCl, 1 mM DTT, 0.25 mg/ml BSA, 4% glycerol for 20 minutes at room temperature. For supershift analysis, nuclear extracts were preincubated with 0.5 μl of antisera against NF-κB p50 or p65 sub-units for 10 minutes in binding buffer. Unlabeled NF-κB, AP-1 or SP1 (5' ATTCGATCGGGGCGGGGCGAGC3') oligonucleotides were used as competitors. Complexes were separated on 5% non-denaturing polyacrylamide-urea gels, which were dried and exposed to a PhosphorImager screen (Amersham Pharmacia Biotech, Piscataway, NJ).
IRAK1 in vitro kinase assay
Cells were harvested from 30 mm inserts into 600 μl ice-cold lysis buffer (50 mM HEPES, pH 7.6, 150 mM NaCl, 1 mM EDTA, 1% NP-40, 1 mM Na3VO4, 20 mMβ-glycerophosphate, 1 mM NaF, 1 mM benzamidine, 5 mM para-nitrophenylphosphate, 1 mM DTT, 1 mM PMSF, 1 μg/ml leupeptin, 1 μg/ml aprotinin, and 1 μg/ml pepstatin) using a cell scraper. After brief vortexing and incubation on ice for 20 minutes, tubes were centrifuged and supernatant protein concentrations were determined as above. For immunoprecipitation, 1000 μg of protein extract was precleared with 2 μg of normal rabbit IgG and 20 μl of protein G-agarose slurry. Protein G beads were pelleted and the supernatant was incubated with 2 μg of rabbit IgG against IRAK1. Protein G-agarose slurry (20 μl) was added and incubated for 1 hour and the beads were washed 3 times with lysis buffer and twice with kinase buffer without 32P ATP (see below). Beads were suspended in 20 μl of kinase buffer (20 mM Tris-HCl, pH 7.6. 20 mM MgCl2, 20 mM β-glycerophosphate, 1 mM benzamidine, 20 mM para-nitrophenylphosphate, 0.4 mM PMSF; 1 mM sodium metabisulfite, 2 μM cold ATP and 10 μCi γ-32P ATP). Reactions were allowed to proceed at 30°C for 30 min and terminated with SDS sample buffer. Samples were then run on 4–20% polyacrylamide gels, dried, and exposed to a PhosphorImager screen as above.
NF-κB reporter assay and expression of dominant negative IRAK1
Adenoviral vectors constitutvely expressing the LacZ gene from the CMV promoter (Ad.CMV-lacZ) and NF-κB-responsive firefly luciferase (Ad.NF-κB-fLuc) have been described previously [43]. We created an adenoviral vector expressing the DD domain of IRAK1 (NCBI accession # NM_001569, amino acids 1–80), which is reported to function as a dominant negative (dn) [44]. A plasmid encoding human IRAK1 was kindly provided by Dr. X. Li (The Cleveland Clinic, Cleveland, OH) and was used as a template for PCR using primers (forward: 5'-CTC GAG GTG CCA GGC TGT GA-3', reverse: 5'-GCT AGC CGG CAG CCA TGG-3') adding 5' and 3' XhoI and NheI sites, respectively. The amplified fragment was cloned into the pCR2.1 vector (Invitrogen). The resulting plasmid was digested with XhoI /NheI and the fragment was ligated into XhoI /NheI-digested pShuttle-IRES-hrGFP-1 adenoviral expression vector plasmid (Stratagene, LaJolla, CA). This strategy placed the DD domain of IRAK1 in frame with the 3X Flag tag of the vector, and the final construct was verified by sequencing. Adenoviral vectors were created, plaque-purified and amplified using conventional methods. AALEB or hTBE cells were transfected with Ad.CMV-lacZ and Ad.NF-κB-fLuc and pShuttle-IRES-hrGFP-1 empty vector or pShuttle-dnIRAK1-IRES-hrGFP-1, using a 1:10 ratio of reporter and expression vectors, respectively. Cells were exposed to viruses for 2 hours, 48 hours prior to experimental challenge, using transient permeabilization [45] for hTBE cells. Eight hours after challenge, cells were lysed and fLuc and β-galactosidase activity measured as described previously [46,47].
Results
The initial response of hTBE cells to Ps. a. products
To simulate in vivo host-bacterial product interactions, we challenged the apical surface of polarized, well-differentiated hTBE cells with late stationary phase Ps. a. filtrates. Exposure of the apical culture surface to 20% TSB (the negative control bacterial broth) in the presence of 10% human serum added as a source of LPS binding protein and soluble CD14, for 24 hours, caused a modest, < 1 fold, culture-dependent increase in IL-8 secretion ([11] and data not shown). Challenge with Ps. a. filtrate induced IL-8 secretion as a function of dose (Figure 1A), causing an average 6.3 ± 1.4 fold increase over the TSB control at the 20% Ps. a. filtrate dose (mean ± SEM, cells from 3 independent donors). We have previously shown that Ps. a. treatment similarly induces IL-6, but not IL-10 or RANTES [11]. Ps. a. challenge of well-differentiated hTBE cultures was non-toxic, as assessed by lack of LDH release into the medium (Figure 1B and 1C) and maintenance of a patent air-liquid interface [for more information, including a positive control, see the online supplement of reference [11], ].
Figure 1 Treatment of well-differentiated hTBE cells with Ps. a. culture filtrates induced IL-8 secretion and was non-toxic. A) The apical surface of well-differentiated hTBE cell culures was challenged with the indicated concentration of Ps. a. filtrate or 20% TSB as a control (0% Ps. a. group) in the presence of 10% human serum, and IL-8 was measured in the 24-hour conditioned basolateral medium. The results are the mean + SEM from 3 independent experiments with cell cultures from 3 different donors. B and C) Ps. a. filtrate was essentially non-toxic as illustrated by lack of LDH secretion into the medium and retention of cellular LDH. The results are from three replicate wells (mean + SD).
In several cell types, inflammatory mediator synthesis following bacterial product challenge is regulated by TLR activation of NF-κB and AP-1 transcription factors. We determined whether Ps. a. filtrates stimulated the TLR signal transduction pathway in well-differentiated hTBE cells. Challenge of hTBE cultures with 20% Ps. a. filtrate promptly induced IRAK1 phosphorylation (Figure 2A), IκBα protein degradation (Figure 2B), and MAP kinase phosphorylation, including JNK, p38, and ERK (Figure 2C). On the basis of equivalent protein input to the immunoprecipitaiton reaction, the LPS-treated THP-1 monocytic cells, used as positive and negative controls, contained much more phospho-IRAK1 than hTBE cells. We did not reduce the concentrations of growth factors in the culture media prior to challenge thus, ERK phosphorylation was relatively high at baseline. Nonetheless, we still observed enhanced ERK phosphorylation after Ps. a. challenge. As predicted, IκBα degradation and MAP kinase activation preceded elevated nuclear content of NF-κB and AP-1 transcription factors, respectively (Figure 3A and 3B). The predominant shifted NF-κB band was composed of p50/p65 heterodimers. We did not attempt supershift assays to identify the subset of AP-1 transcription factors. IRAK1 phosphorylation and stimulation of the NF-κB and MAP kinase pathways is consistent with Ps. a. filtrate activation of the MyD88-dependent component of the TLR receptor signal transduction pathway in hTBE cells.
Figure 2 Ps. a. filtrate induced IRAK1 autophosphorylation, IκBα degradation and MAP kinase phosphorylation. A)Proteins were harvested from day-21 hTBE cell cultures treated apically with 20% Ps. a. filtrate or from THP1 monocytic cells treated with LPS at the times indicated. Equal amounts of protein were immunoprecipitated with anti-IRAK1 or control IgG for each cell type. Precipitates were subjected to in vitro kinase assay and polyacrylamide gel electrophoresis and exposed to a Phosphoimager screen. B and C) Day-21 hTBE cell cultures were treated with Ps. a. filtrate, and harvested for western blot of IκBα (B) or phospho- and total MAP kinases (C) at the indicated time points. Equal amounts of protein were run per lane on each gel. The βactin and total MAP kinase western blots represent re-probing of the IκBα and phospho-MAP kinase blots, respectively. The results are representative of 3 separate experiments with cells derived from 3 different donors.
Figure 3 Ps. a. filtrate increased nuclear content of NF-κB and AP-1 transcription factors. Nuclear extracts were prepared from day-21 hTBE cell cultures challenged apically with Ps. a. filtrate or IL-1β at the times noted and equal amounts of nuclear protein per lane were subjected to EMSA using oligonucleotide probes for NF-κB (A) or AP-1 (B). The inhibitor and supershift assays were all performed with nuclei extracted 2 hours after TSB or Ps. a. challenge. The results are representative of 3 separate experiments with cells derived from 3 different donors.
The response of hTBE cells to repeated Ps. a. challenge
To assess whether prior exposure to Ps. a. filtrates induced tolerance, we re-challenged hTBE cell cultures 24 hours after the initial exposure, studying the 4 possible combinations (Figure 4A). As illustrated in figure 4B, initial treatment with TSB caused minimal IL-8 secretion and no change following a second TSB treatment. Initial TSB treatment followed by Ps. a. resulted in enhanced IL-8 production during the second 24-hour period. Initial exposure to Ps. a. induced a five-fold increase in IL-8 secretion and when these cells were re-challenged with TSB they displayed lesser IL-8 secretion, but still elevated over the initial TSB control. Importantly, cells initially challenged with Ps. a. and re-challenged with Ps. a. showed less stimulation of IL-8 secretion during the second 24-hour period, almost identical to that of TSB-re-challenged cells and, thus, were tolerant. Tolerance was consistently observed, and IL-8 production by cells from 3 different individuals during the second 24 hour period was normalized by dividing the Ps. a. -pretreated IL-8 value by the corresponding value in TSB-pretreated cultures (Fig 4C). This analysis confirmed no significant differences in IL-8 secretion during the second 24-hour period between Ps. a. and TSB treated cells that were pre-exposed to Ps. a. filtrate. Re-challenge of Ps. a.-exposed cells with Ps. a. reduced IL-8 secretion by an average of 44 ± 3% (p < 0.05) compared to TSB pretreated cells. Interestingly, Ps. a. induced significant tolerance to a subsequent stimulation with IL-1β (47 ± 9 % reduction; p < .05), but not TNFα (20 ± 14 % reduction; p > .05). Hetero-tolerance between Ps. a. and IL-1β, but not TNFα, suggests that tolerance occurs at a point in common to the TLR and IL-1β pathways, but upstream of the convergence with the TNFα pathway (see Discussion).
Figure 4 Ps. a. filtrate induced tolerance to a second Ps. a. challenge and cross-tolerance with IL-1β but not TNFα. The apical surface of well-differentiated, day-21 hTBE cell cultures was challenged with Ps. a. filtrate or TSB as schematically illustrated in (A) and IL-8 production was measured in the basolateral medium at 24 or 48 hours. The results of a representative experiment are given as means ± SD in panel (B) and are from duplicate assays of three replicate wells. Similar results were obtained in 3 separate experiments with cell cultures from 3 different donors. C) The apical surface of well-differentiated, day-21 hTBE cell cultures (triplicate wells) was challenged with Ps. a. filtrate or TSB. Following washing and basolateral media change, the cells were re-challenged as indicated. IL-8 production during the second 24 hour period was normalized by dividing the Ps. a.-pretreated IL-8 value by the corresponding value in TSB-pretreated cultures (eg. Ps. a. → Ps. a./TSB → Ps. a.). The results are the average ± SEM of three independent experiments using cells from different donors. P values are based on ANOVA and Tukeys test, n.s. = not significant.
To explore the mechanism of hTBE cell tolerance to Ps. a. products, we studied IκBα degradation, MAP kinase phosphorylation, and generation of nuclear NF-κB and AP-1 binding activity at an optimal time following the second challenge based on the prior time course studies. As shown in figure 5A and 5B, pre-exposure to Ps. a. for 24 hours largely prevented IκBα degradation and phosphorylation of JNK, p38 and ERK MAP kinases in response to a second Ps. a. challenge, while cells pre-exposed to TSB responded vigorously. Likewise, pre-exposure to Ps. a. prevented the activation of NF-κB or AP-1 transcription factors (Figure 5C).
Figure 5 IκBα degradation, MAP kinase activation and induction of nuclear NF-κB or AP-1 transcription factors were strongly attenuated following a second Ps. a. filtrate challenge of tolerant hTBE cells. The apical surface of well-differentiated, day-21 hTBE cell cultures was challenged with Ps. a. filtrate or TSB as indicated and cultures were incubated for 24 hours. Following washing and basolateral media change, the apical surfaces were re-challenged as indicated and proteins were harvested for IκBα and βactin (A) or phospho- and total MAP kinase (B) western blot analysis 20 minutes later. Equal amounts of protein were run per lane on each gel. The βactin and total MAP kinase western blots represent re-probing of the IκBα and phospho-MAP kinase blots, respectively. C) hTBE cells were re-challenged as indicated and nuclear proteins were harvested 2 hours following the second challenge. Equal amounts of nuclear protein per lane were subjected to EMSA using oligonucleotide probes for NF-κB or AP-1, only the shifted bands are shown. Similar results were obtained in 3 separate experiments with cell cultures from 3 different donors.
IRAK-1 as a critical determinant of hTBE cell sensitivity to Ps. a.
IL-1β, but not TNFα hetero-tolerance and the elimination of both MAP kinase and NF-κB responses in tolerant hTBE cells focused our attention on upstream elements of the signal transduction pathway. Prior studies have suggested that decreased IRAK1 protein is associated with LPS tolerance [26-31]. As illustrated in Figure 6A and 6B, IRAK1 protein content decreased during the initial stimulation and was reduced 62% ± 4% (mean ± SEM, n = cell cultures from 3 different individuals) at 24 hours following the initial Ps. a. challenge. Studying the same four exposure permutations described above, we found that IRAK1 remaining 24 hours following the initial Ps. a. challenge could not be autophosphorylated after a second Ps. a. challenge (Figure 6C). Thus, tolerance to Ps. a. products in hTBE cells was associated with reduced IRAK1 protein content and kinase activity. Furthermore, similar to the previously noted hetero-tolerance in IL-8 secretion between Ps. a. filtrate and IL-1β (Figure 3C), prior exposure to Ps. a. prevented IRAK1 phosphorylation in response to IL-1β, and vice versa (Figure 6D).
Figure 6 IRAK1 protein content was reduced in tolerant hTBE cells and remaining IRAK1 was not autophosphorylated after Ps. a. or IL-1β treatment. A) Proteins were harvested from day-21 hTBE cell cultures following apical challenge with Ps. a. filtrate at the times indicated. Equal amounts of protein were run per lane and subjected to western blot for IRAK1. B) Protein samples were obtained from cultures representing 3 different donors and IRAK protein content was examined as described in A, above. C and D) The apical surface of well-differentiated, day-21 hTBE cell cultures was challenged with Ps. a. filtrate, TSB or IL-1β as indicated and cultures were incubated for 24 hours. Following washing and basolateral media change, the apical surfaces were re-challenged as indicated and cellular protein was harvested 20 minutes following the second challenge. Equal amounts of protein per lane were immunoprecipitated with anti-IRAK1. Precipitates were subjected to in vitro kinase assay, run on polyacrylamide gels and exposed to a Phosphoimager screen. The results in panels C and D are representative of 3 separate experiments using cells derived from 3 different donors.
To facilitate molecular approaches towards understanding tolerance mechanisms in airway epithelial cells, we studied AALEB cells, a novel SV40 and telomerase immortalized hTBE derived cell line [42]. AALEB cells on plastic detached from the culture dishes when exposed to unboiled Ps. a. filtrates, presumably due to protease activity. Since prior studies demonstrated that Ps. a.-derived IL-8 stimulatory activity for A549 cells was heat resistant [40], we tested boiled Ps. a. filtrates on both hTBE and AALEB cells. Similar to unboiled filtrates in the presence of serum, hTBE cells challenged with 20% boiled Ps. a. filtrate, without serum, secreted IL-8 and developed tolerance (Figure 7A). Removal of the Ps. a. stimulus, followed by PBS washing and media change, and a 24 or 48 hour "rest time" promoted re-sensitization of hTBE cells. Although the ratio of cells to media, and thus absolute IL-8 levels, are much lower in conventional plastic cultures of AALEB cells than in air-liquid interface cultures of hTBE cells, tolerance and loss of tolerance regarding IL-8 secretion was almost identical (Figure 7B). Thus, the AALEB cell line is apparently a good model to study tolerance in airway epithelial cells.
Figure 7 Tolerance and loss of tolerance in IL-8 secretion and NF-κB driven transcription is similar in hTBE cells and the AALEB cell line, and dnIRAK1 inhibits responses to Ps. a. and IL-1β but not TNFα. A-D) The apical surface of well-differentiated, day-21 hTBE cell cultures (A, C) or AALEB cells in conventional culture on plastic (B, D) were challenged with boiled Ps. a. filtrate (P) or TSB (T) in the absence of human serum as indicated. In A and B, cultures were incubated for 24 hours, and following washing and media change, the apical surfaces were re-challenged when indicated. IL-8 was measured in the media 24 hours following the second challenge. C and D) Cells were transfected with adenoviral vectors expressing constitutive or NF-κB driven reporter genes as described in Methods and were challenged with TSB or Ps. a. and lysed for reporter gene assay as indicated. Note the steep slope of the Ps. a. response in naive or TSB-pretreated cells versus the attenuated slope in Ps.a.-pretreated cells. E, F) hTBE and AALEB cells were transfected with adenoviral vectors expressing constitutive and NF-κB driven reporter genes and dnIRAK1 or a control vector as described in the Methods and were lysed for reporter gene assay 8 hours after treatment as indicated. dnIRAK1 inhibited approximately 50% of the Ps. a. response in both cell types, and, in AALEB cells, >90% of the IL-1β response, but not the TNFα response. All points represent the mean of triplicate wells ± SD. The results in panel E ands F are representative of 2 and 3 separate experiments, respectively
To examine the effects of Ps. a. on NF-κB dependent gene transcription, we transfected hTBE and AALEB cells with adenoviruses expressing NF-κB-driven luciferase and constitutively expressed LacZ reporter genes. Transient permeabilization with caproic acid [45] was necessary for efficient gene transfer to hTBE cells. Luciferase activity normalized for β gal was measured at baseline and 8 and 24 hours following stimulation with the 4 possible combinations of Ps. a. or TSB. Ps. a. strongly simulated NF-κB-driven luciferase activity in naive or TSB pre-treated hTBE or AALEB cells, but the response was strongly attenuated in Ps. a. pretreated cells (compare the slopes of the 0–8 hour and 25–33 hour groups in Figure 7C and 7D). These results suggest a strong NF-κB driven transcriptional component of Ps. a. stimulation that was inhibited in tolerant airway epithelial cells.
Having established a reporter assay in a relevant cell type, it was now possible to examine whether IRAK1 down-regulation was correlative or causal in airway epithelial tolerance. AALEB cells were transfected with a 1:10 ratio of adenoviral particles expressing both reporter genes and dnIRAK1 [44] or the empty vector control, respectively. In both hTBE and AALEB cells, NF-κB driven luciferase activity was approximately 5-fold greater in response to Ps. a. in cells infected with the control, empty vector. Luciferase activity was reduced significantly (52% in hTBE cells, n = 2 independent experiments; 55 ± 8 % in AALEB cells, n = 3 independent experiments), but not completely, by expression of dnIRAK1. Representative experiments are illustrated in Figure 7E and 7F, respectively. In AALEB cells, the dnIRAK1 construct caused an average 86 ± 1% decrease in IL-1β-stimulated NF-κB-driven luciferase activity but did not affect TNFα stimulated activity. These results confirm an important role for IRAK1 in the response to Ps. a. products, presumably via its integral function in the MyD88-dependent portion of the TLR pathway, and strongly suggest that loss or inhibition of IRAK1 is an important component of tolerance.
Discussion
The host must eradicate pathogens while preventing tissue injury due to inflammation. Repeated exposure of airway epithelium to microbial products is a hallmark of chronic infectious lung diseases but the adaptation has not been previously studied in these key cells. Polarized, well-differentiated hTBE cell cultures used in the current studies recapitulate the morphology and mucus transport function found in vivo [39]. We challenged the apical surface of these cultures with soluble products of Ps. a., an important airway pathogen. Prior studies focused on the direct interaction of live Ps. a. bacteria with epithelial cells (see [48] for review). However, bacteria in chronically infected CF lungs are present mostly as intra-luminal masses distal from the airway epithelial surface [49]. Thus, we chose to model the interaction of the apical membrane of polarized hTBE cells with diffusible Ps. a. products, rather than live bacteria. We used late stationary phase TSB cultures of a common strain of Ps. a. instead of clinical isolates. Ps. a. in CF patients frequently become mucoid, produce modified forms of LPS [50] and also secrete quorum sensing molecules [51]. Future studies will be necessary to determine whether modifications of Ps. a. typically found in chronic human infections alter epithelial responses. Although we used non-inoculated TSB processed in parallel as a negative control, we cannot completely exclude that a portion of the stimulation was due to TSB breakdown products induced by bacterial metabolism of the broth. Well-differentiated hTBE cells likely have greater physiologic relevance than cell lines grown directly on plastic. However, well-characterized cell lines facilitate mechanistic studies, so we also used a newly developed airway epithelial cell line (AALEB).
Initial exposure of hTBE or AALEB cells to Ps. a. filtrate induced IL-8 release. This response was preceded by phosphorylation of IRAK1, JNK, p38 and ERK, degradation of IκBα, and generation of NF-κB and AP-1 and transcriptional activity in hTBE cells. IRAK1 phosphorylation leading to activation of NF-κB and AP-1 transcription factors is consistent with activation of the MyD88-dependent portion of the TLR signal transduction pathway. A main finding of the present study in both hTBE and AALEB cells was that cells previously exposed to Ps. a. products for 24 hours had a greatly diminished IL-8 response. Likewise, IRAK1 phosphorylation, IκBα degradation, MAP kinase activation and generation of NF-κB and AP-1 and transcriptional activity after a second Ps. a. exposure were all markedly reduced in Ps. a.-tolerant cells. NF-κB reporter assays strongly suggested that Ps. a.-induced cytokine secretion was transcriptionally regulated and that tolerance was associated with decreased activation of NF-κB transcription factors. The cells were hetro-tolerant to IL-1β, but not TNFα, suggesting that tolerance occurs at a point in common to the TLR and IL-1β pathways, but upstream of the convergence with the TNFα pathway, which is precisely the location of IRAK1 (Figure 8). Ps. a. tolerance was associated with decreased IRAK1 protein content and failure to phosphorylate IRAK1 which is reminiscent of LPS tolerance in other cell types [26-31]. A very recent study demonstrates that macrophage tolerance to the prototypical TLR2 agonist Pam3Cys is due to ablation of IRAK1 [52]. Additional evidence for a key role of IRAK1 in the airway epithelial response to Ps. a. was suggested by expression of dnIRAK1, which diminished NF-κB-driven transcription approximately 50%. In AALEB cells, dnIRAK1 did not alter the response to TNFα, but reduced IL-1β effects nearly 90%. Thus, it is likely that factors present in Ps. a. filtrates activate both IRAK1-dependent and -independent pathways leading to NF-κB activation. It is interesting to consider whether tolerant airway epithelial cells were generally "run down" and incapable of responding, but IL-8 release and NF-κB-driven luciferase activity after TNFα treatment of Ps. a.-tolerant cells indicates that the downstream signalling apparatus was fully capable of responding. Components upstream of the convergence of the TLR and TNFα pathways may indeed be "run down", in fact, it is our hypothesis that decreased IRAK1 protein results in Ps. a. tolerance in airway epithelial cells.
Figure 8 IRAK1 functions in the TLR and IL-1β pathway upstream of convergence with the TNFα pathway. A simplified diagram illustrating portions of the TLR, IL-1β and TNFα signalling pathways. Consistent with our observations, disruption of IRAK1 inhibits TLR and IL-1β, but not TNF, activation of NF-κB.
Generic IRAK was first recognized as a crucial component of the IL-1β signalling pathway and molecular cloning revealed homology to the Drosophila protein kinase Pelle [4]. More recent studies have revealed 4 IRAK isoforms [53], a critical role for tandem action of IRAK1 and 4 [54] and an inhibitory role for IRAK-M [34]. Ligation of TIR domain-containing receptors induces IRAK1 phosphorylation both auto-catalytically and by unidentified kinases. Phosphorylated IRAK1 becomes degraded by proteasomes in cells stimulated by IL-1β but not TNFα, which correspondingly results in desensitization of the IL-1β, but not TNFα response [55]. The precise mechanisms regulating IRAK1 localization and function during the propagation and termination of the TLR signal in hTBE cells are not fully understood, and require further study.
While the kinetics of the initial response of naive cells suggested a direct interaction of Ps. a. products to trigger the TLR pathway, airway epithelial cells at 24 hours were likely exposed to released autocrine/paracrine factors that may have also contributed to tolerance. However, this was probably not mediated by IL-1β, since this cytokine was consistently undetectable in hTBE cell media after Ps. a. stimulation (SHR and MWV unpublished observations). Moreover, such factors would have to selectively target the upstream TLR/IL-1β, pathway since TNFα responses were not affected.
Ps. a. is a complex and adaptable bacterium that secretes many factors known to damage host cells and/or induce inflammation such as pilin, flagellin, pyocyanin, hemolysins, autoinducer, LPS, proteases, and small unidentified heat-stable factors [9,40,50,56-59]. There is a broad array of mechanisms by which these agents may act, including TLR activation, generalized oxidative stress due to redox cycling of pyocyanin, and proteolytic modification of target cells by Ps. a. elastase or alkaline protease. IRAK activation in a human bronchial cell line has been shown following neutrophil elastase [60]. Rapid activation of IRAK1, in conjunction with desensitization of the IL-1β but not the TNFα response, suggests that Ps. a. products directly trigger an MyD88-dependent TLR pathway. Activation of airway epithelial cells by a variety of live bacteria and bacterial products is TLR2-dependent [10,12-14]. Polymyxin B treatment is commonly used to block LPS effects [61], but it did not reduce Ps. a.-stimulated hTBE cell IL-8 secretion, and Ps. a. derived LPS is a poor inducer of IL-8 in these cells (SHR and MWV unpublished observations), suggesting that LPS stimulation of TLR4 was not predominant. Further studies are needed to chemically identify the most quantitatively important TLR pathway-stimulating substance(s) produced by Pseudomonas aeruginosa. In this regard, recent studies indicate a potential role for TLR2 activation by lipopeptides encoded by Ps. a. genes [62].
While controversial, several lines of evidence suggest that inflammatory responses in the CF airway may be intrinsically enhanced or protracted due to the absence of functioning CFTR (see [11] and references therein). Thus, examination of differences in the development of tolerance in CF cells is worthy of further scrutiny. Although hTBE cells in our in vitro model became tolerant to Ps. a., chronically infected airways in CF humans are typically severely inflamed. Whether tolerance exists in vivo, if it is helpful or harmful, different in CF, or if it is overwhelmed by the heavy bacterial burden are important, but unanswered questions. Multiple cell types, which are constantly turning over, contribute to in vivo responses. Bacterial products may penetrate beyond the epithelial barrier to trigger inflammation or systemic responses. We speculate that epithelial tolerance to Ps. a. occurs in vivo, and that inflammation would be even more severe in its absence. However, further clinical and/or animal model studies are needed to address these key questions.
Conclusion
We have shown that exposure of primary, well-differentiated hTBE cells to Ps. a. products causes selective tolerance. Ps. a. products elicit a response via the TLR signal transduction pathway that becomes down-regulated, likely due to decreased IRAK1 protein content and inhibition of IRAK1 phosphorylation. A greater understanding of the precise mechanisms decreasing airway epithelial cell generation of inflammatory mediators will suggest new avenues for anti-inflammatory therapies to minimize lung destruction typical of CF and other chronic infectious airway diseases.
Authors' contributions
QW conceived the studies, performed culture challenges, most of the biochemical assays, and drafted the manuscript. ZL cloned the DD domain of IRAK-1 into the viral vector, characterized the response of AALEB cells to Ps. a. and performed all aspects of the reporter assays on both AALEB and hTBE cells. MWV performed specific biochemical studies, most of the statistical analysis and critically contributed to the experimental design and written manuscript. SHR participated in the design, coordination and analysis of the experiments, performed western blots, prepared the figures, and wrote and submitted the final manuscript. All authors read and approved the final manuscript.
Acknowledgements
The authors thank the staff of the UNC CF Center Tissue Procurement and Cell Culture Core for providing human airway epithelial cells, Kelly Plonk for technical assistance and Lisa Brown for help preparing the manuscript. We thank Dr. X. Li for IRAK plasmids and Dr. J. Engelhardt for the Ad.NF-κB-fLuc vector. The UNC Gene Therapy Center Vector Core assisted by producing adenovirus. Supported by grants from the Cystic Fibrosis Foundation and NIH to SHR.
==== Refs
Janeway CAJ Medzhitov R Innate immune recognition Annu Rev Immunol 2002 20 197 216 11861602 10.1146/annurev.immunol.20.083001.084359
Beutler B Hoebe K Du X Ulevitch RJ How we detect microbes and respond to them: the Toll-like receptors and their transducers J Leukoc Biol 2003 74 479 485 12960260 10.1189/jlb.0203082
Takeda K Kaisho T Akira S Toll-like receptors Annu Rev Immunol 2003 21 335 376 12524386 10.1146/annurev.immunol.21.120601.141126
Dunne A O'Neill LA The interleukin-1 receptor/Toll-like receptor superfamily: signal transduction during inflammation and host defense Sci STKE 2003 2003 re3 12606705
Baeckhed F Soederhaell M Ekman P Normark S Richter-Dahlfors A Induction of innate immune responses by Escherichia coli and purified lipopolysaccharide correlate with organ- and cell-specific expression of Toll-like receptors within the human urinary tract Cell Microbiol 2001 3 153 158 11260138 10.1046/j.1462-5822.2001.00101.x
Abreu MT Vora P Faure E Thomas LS Arnold ET Arditi M Decreased expression of Toll-like Receptor-4 and MD-2 correlates with intestinal epithelial cell protection against dysregulated proinflammatory gene expression in response to bacterial lipopolysaccharide J Immunol 2001 167 1609 1617 11466383
Cario E Rosenberg IM Brandwein SL Beck PL Reinecker HC Podolsky DK Lipopolysaccharide activates distinct signaling pathways in intestinal epithelial cell lines expressing Toll-like receptors J Immunol 2000 164 966 972 10623846
Arbour NC Lorenz E Schutte BC Zabner J Kline JN Jones M Frees K Watt JL Schwartz DA TLR4 mutations are associated with endotoxin hyporesponsiveness in humans Nat Genet 2000 25 187 191 10835634 10.1038/76048
Becker MN Diamond G Verghese MW Randell SH CD14-dependent lipopolysaccharide-induced beta-defensin-2 expression in human tracheobronchial epithelium J Biol Chem 2000 275 29731 29736 10882713 10.1074/jbc.M000184200
Hertz CJ Wu Q Porter EM Zhang YJ Weismuller KH Godowski PJ Ganz T Randell SH Modlin RL Activation of Toll-like receptor 2 on human tracheobronchial epithelial cells induces the antimicrobial peptide human beta defensin-2 J Immunol 2003 171 6820 6826 14662888
Becker MN Sauer MS Muhlebach MS Hirsh AJ Wu Q Verghese MW Randell SH Cytokine secretion by cystic fibrosis airway epithelial cells Am J Respir Crit Care Med 2004 169 645 653 14670800 10.1164/rccm.200207-765OC
Shuto T Xu H Wang B Han J Kai H Gu XX Murphy TF Lim DJ Li JD Activation of NF-kappaB by nontypeable Hemophilus influenzae is mediated by toll-like receptor 2-TAK1-dependent NIK-IKKalpha/beta-IkappaBalpha and MKK3/6-38 MAP kinase signaling pathways in epithelial cells Proc Natl Acad Sci U S A 2001 98 8774 8779 11438700 10.1073/pnas.151236098
Muir A Soong G Sokol S Reddy B Gomez M van Heeckeren A Prince A Toll like receptors in normal and cystic fibrosis airway epithelial cells Am J Respir Cell Mol Biol 2004 30 777 783 14656745 10.1165/rcmb.2003-0329OC
Adamo R Sokol S Soong G Gomez M Prince A P. aeruginosa flagella activate airway epithelial cells through asialoGM1 and TLR2 as well as TLR5 Am J Respir Cell Mol Biol 2004 30 627 634 14607814 10.1165/rcmb.2003-0260OC
Soong G Reddy B Sokol S Adamo R Prince A TLR2 is mobilized into an apical lipid raft receptor complex to signal infection in airway epithelial cells J Clin Invest 2004 113 1482 1489 15146246 10.1172/JCI200420773
Sha Q Truong-Tran AQ Plitt JR Beck LA Schleimer RP Activation of airway epithelial cells by toll-like receptor agonists Am J Respir Cell Mol Biol 2004 31 358 364 15191912 10.1165/rcmb.2003-0388OC
Suzuki M Hisamatsu T Podolsky DK Gamma interferon augments the intracellular pathway for lipopolysaccharide (LPS) recognition in human intestinal epithelial cells through coordinated up-regulation of LPS uptake and expression of the intracellular Toll-like receptor 4-MD-2 complex Infect Immun 2003 71 3503 3511 12761135 10.1128/IAI.71.6.3503-3511.2003
Monick MM Yarovinsky TO Powers LS Butler NS Carter AB Gudmundsson G Hunninghake GW Respiratory syncytial virus up-regulates TLR4 and sensitizes airway epithelial cells to endotoxin J Biol Chem 2003 278 53035 53044 14565959 10.1074/jbc.M308093200
Dobrovolskaia MA Vogel SN Toll receptors, CD14, and macrophage activation and deactivation by LPS Microbes Infect 2002 4 903 914 12106783 10.1016/S1286-4579(02)01613-1
Medvedev AE Henneke P Schromm A Lien E Ingalls R Fenton MJ Golenbock DT Vogel SN Induction of tolerance to lipopolysaccharide and mycobacterial components in Chinese hamster ovary/CD14 cells is not affected by overexpression of toll-like receptors 2 or 4 J Immunol 2001 167 2257 2267 11490013
Lehner MD Morath S Michelsen KS Schumann RR Hartung T Induction of cross-tolerance by lipopolysaccharide and highly purified lipoteichoic acid via different toll-like receptors independent of paracrine mediators J Immunol 2001 166 5161 5167 11290799
Medvedev AE Kopydlowski KM Vogel SN Inhibition of lipopolysaccharide-induced signal transduction in endotoxin-tolerozed mouse macrophages: dysregulation of cytokine, chemokine, and Toll-like receptor 2 and 4 gene expression J Immunol 2000 164 5564 5574 10820230
Nomura F Akashi S Sakao Y Sato S Kawai T Matsumoto M Nakanishi K Kimoto M Miyake K Takeda K Akira S Cutting edge: endotoxin tolerance in mouse peritoneal macrophages correlates with down-regulation of surface toll-like receptor 4 expression J Immunol 2000 164 3476 3479 10725699
Medvedev A Vogel SN Overexpression of CD14, TLR4, and MD-2 in HEK 293T cells does not prevent induction of in vitro endotoxin tolerance J Endotoxin Res 2004 9 60 64 12691621 10.1179/096805103125001360
Brint EK Xu D Liu H Dunne A McKenzie AN O'Neill LA Liew FY ST2 is an inhibitor of interleukin 1 receptor and Toll-like receptor 4 signaling and maintains endotoxin tolerance Nat Immunol 2004 5 373 379 15004556 10.1038/ni1050
Li L Coursat S Hu J McCall CE Characterization of IRAK in normal and endotoxin tolerant cells J Biol Chem 2000 275 23340 23345 10811644 10.1074/jbc.M001950200
Jacinto R Hartung T McCall C Li L Lipopolysaccharide- and lipoteichoic acid-induced tolerance and cross-tolerance: distinct alterations in IL-1 receptor-associated kinase J Immunol 2002 168 6136 6141 12055225
Yamashina S Rusyn I Ikejima K Sato N Thurman RG Tolerance and sensitization to endotoxin in Kupffer cells caused by acute ethanol involve interleukin-1 receptor-associated kinase Biochem Biophys Res Commun 2000 277 686 690 11062014 10.1006/bbrc.2000.3738
Sato S Takeuchi O Fujita T Tomizawa H Takeda K Akira S A variety of microbial components induce tolerance to lipopolysaccharide by differentially affecting MyD88-dependent and -independent pathways Int Immunol 2002 14 783 791 12096038 10.1093/intimm/dxf046
Yeo SJ Yoon JG Hong SC Yi AK CpG DNA induces self and cross-hyporesponsiveness of RAW264.7 cells in response to CpG DNA and lipopolysaccharide: alterations in IL-1 receptor-associated kinase expression J Immunol 2003 170 1052 1061 12517973
Otte JM Cario E Podolsky DK Mechanisms of cross hyporesponsiveness to Toll-like receptor bacterial ligands in intestinal epithelial cells Gastroenterology 2004 126 1054 1070 15057745 10.1053/j.gastro.2004.01.007
Medvedev AE Lentschat A Wahl LM Golenbock DT Vogel SN Dysregulation of LPS-induced Toll-like receptor 4-MyD88 complex formation and IL-1 receptor-associated kinase 1 activation in endotoxin-tolerant cells J Immunol 2002 169 5209 5216 12391239
Mizel SB Snipes JA Gram-negative flagellin-induced self-tolerance is associated with a block in interleukin-1 receptor-associated kinase release from toll-like receptor 5 J Biol Chem 2002 277 22414 22420 11953430 10.1074/jbc.M201762200
Kobayashi K Hernandez LD Galan JE Janeway CAJ Medzhitov R Flavell RA IRAK-M is a negative regulator of Toll-like receptor signaling Cell 2002 110 191 202 12150927 10.1016/S0092-8674(02)00827-9
Neish AS Gewirtz AT Zeng H Young AN Hobert ME Karmali V Rao AS Madara JL Prokaryotic regulation of epithelial responses by inhibition of IkappaB-alpha ubiquitination Science 2000 289 1560 1563 10968793 10.1126/science.289.5484.1560
Yamazaki S Muta T Takeshige K A novel IkappaB protein, IkappaB-zeta, induced by proinflammatory stimuli, negatively regulates nuclear factor-kappaB in the nuclei J Biol Chem 2001 276 27657 27662 11356851 10.1074/jbc.M103426200
Kastenbauer S Ziegler-Heitbrock HW NF-kappaB1 (p50) is upregulated in lipopolysaccharide tolerance and can block tumor necrosis factor gene expression Infect Immun 1999 67 1553 1559 10084986
Dobrovolskaia MA Medvedev AE Thomas KE Cuesta N Toshchakov V Ren T Cody MJ Michalek SM Rice NR Vogel SN Induction of in vitro reprogramming by Toll-like receptor (TLR)2 and TLR4 agonists in murine macrophages: effects of TLR "homotolerance" versus "heterotolerance" on NF-kappa B signaling pathway components J Immunol 2003 170 508 519 12496438
Matsui H Grubb BR Tarran R Randell SH Gatzy JT Davis CW Boucher RC Evidence for periciliary liquid layer depletion, not abnormal ion composition, in the pathogenesis of cystic fibrosis airways disease Cell 1998 95 1005 1015 9875854 10.1016/S0092-8674(00)81724-9
Massion PP Inoue H Richman-Eisenstat J Grunberger D Jorens PG Housset B Pittet JF Wiener-Kronish JP Nadel JA Novel Pseudomonas product stimulates interleukin-8 production in airway epithelial cells in vitro J Clin Invest 1994 93 26 32 8282796
Fulcher ML Gabriel S Burns KA Yankaskas JR Randell SH Well-differentiated human airway epithelial cell cultures Methods Mol Med 2005 107 183 206 15492373
Lundberg AS Randell SH Stewart SA Elenbaas B Hartwell KA Brooks MW Fleming MD Olsen JC Miller SW Weinberg RA Hahn WC Immortalization and transformation of primary human airway epithelial cells by gene transfer Oncogene 2002 21 4577 4586 12085236 10.1038/sj.onc.1205550
Sanlioglu S Williams CM Samavati L Butler NS Wang G McCray PBJ Ritchie TC Hunninghake GW Zandi E Engelhardt JF Lipopolysaccharide induces Rac1-dependent reactive oxygen species formation and coordinates tumor necrosis factor-alpha secretion through IKK regulation of NF-kappa B J Biol Chem 2001 276 30188 30198 11402028 10.1074/jbc.M102061200
Medzhitov R Preston-Hurlburt P Kopp E Stadlen A Chen C Ghosh S Janeway CAJ MyD88 is an adaptor protein in the hToll/IL-1 receptor family signaling pathways Mol Cell 1998 2 253 258 9734363 10.1016/S1097-2765(00)80136-7
Coyne CB Kelly MM Boucher RC Johnson LG Enhanced epithelial gene transfer by modulation of tight junctions with sodium caprate Am J Respir Cell Mol Biol 2000 23 602 609 11062138
Brasier AR Tate JE Habener JF Optimized use of the firefly luciferase assay as a reporter gene in mammalian cell lines BioTechniques 1989 7 1116 1122 2698191
Felgner JH Kumar R Sridhar CN Wheeler CJ Tsai YJ Border R Ramsey P Martin M Felgner PL Enhanced gene delivery and mechanism studies with a novel series of cationic lipid formulations J Biol Chem 1994 269 2550 2561 8300583
Goldberg JB Pier GB The role of the CFTR in susceptibility to Pseudomonas aeruginosa infections in cystic fibrosis Trends Microbiol 2000 8 514 520 11121762 10.1016/S0966-842X(00)01872-2
Baltimore RS Christie CD Smith GJ Immunohistopathologic localization of Pseudomonas aeruginosa in lungs from patients with cystic fibrosis. Implications for the pathogenesis of progressive lung deterioration Am Rev Respir Dis 1989 140 1650 1661 2513765
Ernst RK Yi EC Guo L Lim KB Burns JL Hackett M Miller SI Specific lipopolysaccharide found in cystic fibrosis airway Pseudomonas aeruginosa Science 1999 286 1561 1565 10567263 10.1126/science.286.5444.1561
Singh PK Schaefer AL Parsek MR Moninger TO Welsh MJ Greenberg EP Quorum-sensing signals indicate that cystic fibrosis lungs are infected with bacterial biofilms Nature 2000 407 762 764 11048725 10.1038/35037627
Siedlar M Frankenberger M Benkhart E Espevik T Quirling M Brand K Zembala M Ziegler-Heitbrock L Tolerance induced by the lipopeptide Pam3Cys is due to ablation of IL-1R-associated kinase-1 J Immunol 2004 173 2736 2745 15294992
Janssens S Beyaert R Functional diversity and regulation of different interleukin-1 receptor-associated kinase (IRAK) family members Mol Cell 2003 11 293 302 12620219 10.1016/S1097-2765(03)00053-4
Suzuki N Suzuki S Eriksson U Hara H Mirtosis C Chen NJ Wada T Bouchard D Hwang I Takeda K Fujita T Der S Penninger JM Akira S Saito T Yeh WC IL-1R-associated kinase 4 is required for lipopolysaccharide-induced activation of APC J Immunol 2003 171 6065 6071 14634120
Yamin TT Miller DK The interleukin-1 receptor-associated kinase is degraded by proteasomes following its phosphorylation J Biol Chem 1997 272 21540 21547 9261174 10.1074/jbc.272.34.21540
Denning GM Wollenweber LA Railsback MA Cox CD Stoll LL Britigan BE Pseudomonas pyocyanin increases interleukin-8 expression by human airway epithelial cells Infect Immun 1998 66 5777 5784 9826354
Oishi K Sar B Wada A Hidaka Y Matsumoto S Amano H Sonoda F Kobayashi S Hirayama T Nagatake T Matsushima K Nitrite reductase from Pseudomonas aeruginosa induces inflammatory cytokines in cultured respiratory cells Infect Immun 1997 65 2648 2655 9199432
Smith RS Fedyk ER Springer TA Mukaida N Iglewski BH Phipps RP IL-8 production in human lung fibroblasts and epithelial cells activated by the Pseudomonas autoinducer N-3-oxododecanoyl homoserine lactone is transcriptionally regulated by NF-kappaB and activator protein-2 J Immunol 2001 167 366 374 11418672
DiMango E Zar HJ Bryan R Prince A Diverse Pseudomonas aeruginosa gene products stimulate respiratory epithelial cells to produce interleukin-8 J Clin Invest 1995 96 2204 2210 7593606
Walsh DE Greene CM Carroll TP Taggart CC Gallagher PM O'Neill SJ McElvaney NG Interleukin-8 up-regulation by neutrophil elastase is mediated by MyD88/IRAK/TRAF-6 in human bronchial epithelium J Biol Chem 2001 276 35494 35499 11461907 10.1074/jbc.M103543200
Warner SJ Mitchell D Savage N McClain E Dose-dependent reduction of lipopolysaccharide pyrogenicity by polymyxin B Biochem Pharmacol 1985 34 3995 3998 2998395 10.1016/0006-2952(85)90378-8
Firoved AM Ornatowski W Deretic V Microarray analysis reveals induction of lipoprotein genes in mucoid Pseudomonas aeruginosa: implications for inflammation in cystic fibrosis Infect Immun 2004 72 5012 5018 15321993 10.1128/IAI.72.9.5012-5018.2004
| 15804356 | PMC1087506 | CC BY | 2021-01-04 16:36:26 | no | Respir Res. 2005 Apr 1; 6(1):26 | utf-8 | Respir Res | 2,005 | 10.1186/1465-9921-6-26 | oa_comm |
==== Front
Virol JVirology Journal1743-422XBioMed Central London 1743-422X-2-281582320010.1186/1743-422X-2-28ResearchNew Vibrio cholerae O1 Biotype ElTor bacteriophages Sen Anindito [email protected] Amar N [email protected] Division of Electron Microscopy, National Institute of Cholera and Enteric Diseases, P-33, C.I.T. Road, Scheme- XM, Beleghata, Kolkata- 700010. India2 (Present Address) Laboratory of Structural Biology, Room 1504, Building 50, NIAMS/NIH Bethesda, MD, 20852, USA2005 11 4 2005 2 28 28 17 2 2005 11 4 2005 Copyright © 2005 Sen and Ghosh; licensee BioMed Central Ltd.2005Sen and Ghosh; 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.
We report the presence of three new O1 ElTor vibriophages named AS1, AS2 and AS3, isolated from the sewage and pond waters of the outskirts of Kolkata. A few phages, named AS4, with hexagonal heads and abnormally long tails with typical curly projections were also found in the water samples.
VibriophageElectron microscopy
==== Body
Vibrio cholerae, the causative agent of cholera in humans, is classified into two serotypes: O1 and nonO1 [1]. The O1 strains are divided into two biotypes: Classical and ElTor. Before 1961 most epidemics had been caused by the classical biotype. But with the passage of time the classical biotype disappeared from the scenario and the ElTor emerged as the major biotype causing the Vibrio cholerae in humans. In 1993, Vibrio cholerae serogroup O139 made an explosive appearance and caused a severe epidemic in the Indian continent [2]. The disease cholera spreads rapidly to far off places from the epicenter of its emergence. From the epidemiological point of view it is important to track down the spread of the disease. Phage typing is a widely accepted method for tracking down cholera epidemic [3]. The international phage-typing scheme of Basu and Mukerjee [3] includes five phages (I, II, III, IV and V). But in course of time this typing scheme proved inadequate as a large number of Vibrio cholerae strains were found to be untypeable using this scheme. In order to overcome this problem a new typing scheme for ElTor strains was proposed in 1993 [4]. In the recent times vibriophages are found to occur in amazingly in large numbers in the environment around the globe [5-8] which, prompted us to search for new cholera phages from the environmental resources.
Sewage and pond water were collected from different places from the outskirts of Kolkata. During the study period the recorded temperature was about 32–38°C and pH ranged from 7.8 to 10. The sample waters were processed for phage isolation as described previously [9] using Vibrio cholerae O1 ElTor (MAK 757) as the propagating strain. The procedure was repeated on nutrient agar or the appearance of plaques. The phages were purified from a single discrete plaque by the soft agar (0.9 %) overlay method. Phage lysates were prepared in nutrient broth. A few drops of chloroform were added to the freshly prepared phage lysate to remove bacterial content in it. The phage lysates (nearly 108 – 109phages/ml) was subjected to ultracentrifugation at 35,000 r.p.m. for 90 minutes in a Sorval T 865 rotor and phage pellets were obtained. The phage pellets were resuspended in 1 ml of 50 mM; Tris-HCl pH 7.5, 20 mM, MgCl2 (TM buffer) to concentrate and the phage was stored at 4°C. The phages was purified on a sucrose step gradient of 10% to 40% as described previously [6] using a Sorval TW 668 swing-out rotor at revolution speed of 35,000 r.p.m. for 75 minutes. The purified phage pellets were resuspended in 1 ml of TM buffer and stored at 4°C. Five microliters of the purified suspensions were deposited directly on Pioloform coated 300-mesh Nickel grids, stabilized with a thin layer of carbon, allowed to adsorb for two minutes and the excess liquid was blotted out. The samples were stained with 2% aqueous uranyl acetate (pH 4.5). Grids were examined in FEI Tecnai 12 Biotwin transmission electron microscope. Measurements of the dimensions of the head (distances between the opposite apices), length and thickness of the tail were done using 'analySIS' software (SIS GmbH, Germany). Calibration was done using catalase crystal with alternate lattice plane spacing of 8.75 nm and 6.85 nm (Agar Scientific Ltd. England). Several enteropathogens were included to test the susceptibility against these phages. Cultures of these enteropathogens were grown to their mid log phases and were plated as lawn on (0.5% NaCl) nutrient agar plate. The lawns are spotted with about 5–7 μl of the lysates. Table 1 shows the result of the phage sensitivity to the different pathogens. It is seen that phages are only sensitive to Vibrio cholerae O1 and are resistant to rest of the species.
Table 1 Table showing the sensitivity of the newly isolated phages to the different species of enteropathogens
Species Phage AS1 Phage AS2 Phage AS3
Vibrio cholerae O1 Sensitive Sensitive Sensitive
Vibrio cholerae O139 Resistant Resistant Resistant
Vibrio cholerae non-O1 Resistant Resistant Resistant
non-O139
Vibrio parahaemolyticus Resistant Resistant Resistant
Escherichia coli Resistant Resistant Resistant
Salmonella enterica Resistant Resistant Resistant
serovar Typhimurium
We have not included the phage AS4 for the sensitivity study as their number is very small which, may give erroneous results.
Three different types of phages, named AS1, AS2 and AS3 were found. All the three phages have hexagonal heads with long tail (figure 1). Phages AS1 and AS3 have hexagonal heads and contractile tails and falls in the family of Myoviridae while phage AS2 has a hexagonal head with non-contractile tail and falls within the family Siphoviridae (according to International Committee for the Taxonomy of Viruses; 1982). The dimensions of the head (distances between the opposite apices), length and thickness of the tail of these phages are summarized in table 1. Morphological comparisons of these three phages were made with several other important typing vibriophages that possess hexagonal heads and long tails have been made and are given in table 1.
Figure 1 Vibrio cholerae O1 Biotype ElTor bacteriophages AS1-3. Panels A and B show the vibriophage AS1. They are contractile in nature and possess similar pattern as seen in the tail of another O1 ElTor typing vibriophage D10 (Chakrabarti et al., (1993). Panels A and B are shown at the same magnification. Panel C show the vibriophage AS2. The tails are non-contractile in nature. Panels D and E show the vibriophage AS3. Panels D and E are shown at the same magnification. The bars in Panels A, C and D: 50 nm.
From table 1 we find that phages AS1, AS2 and AS3 are morphologically different from the other typing vibriophages. While studying these phages we came across few phages (extremely small in number), named AS4, that have the head diameter of nearly equal to 65.24 ± 3.1 nm, straight-tail length nearly 460.20 ± 11.2 nm long and typical curly projection of length 230.20 ± 12.4 nm attached to the free end of the tail. In fact each of the curly projection has a constant contour length of 38.8 ± 5.72 nm. The thickness of the tails is about 10.52 ± 0.86 nm (figure 2). To best of our knowledge, till date, no vibriophage with such abnormally long tails are reported. However, Ackermann and DuBow [10] reported two non-cultivated rumen bacteriophages with such long tails but they do not have any curly projections as seen in AS4.
Figure 2 Vibrio cholerae O1 Biotype ElTor bacteriophage AS4. The tails are enormously long and non contractile in nature. Typical curly projections are seen at the end of the tails. The number of such phages is extremely rare in the water samples. Bar: 50 nm.
Isolation of these new cholera phages from the sewage and pond waters collected from the outskirts of Kolkata, a high cholera-endemic region, (where a cholera outbreak took place nearly two years back) carries additional significance. Detailed physiochemical studies like host specificity, immunological analysis, study of structural proteins, thermal and light inactivation, growth characteristics and extensive study of their genomes of these newly isolated vibriophages will prove helpful in modifying the present phage typing scheme of Vibrio cholerae O1 biotype ElTor untypeable strains in future as it was needed for the old Basu and Mukerjee [3] O1 biotype ElTor typing scheme almost a decade back.
Table 2 Morphology of different Vibriophages
Phage Host Diameter of head (nm) Length of tail (nm) Thickness of tail (nm) Nature of tail Reference
AS1 VC O1 ElTor MAK 757 43.60 ± 2.34 85.21 ± 3.80 13.54 ± 0.91 Contractile tail Present study
AS2 VC O1 ElTor MAK 757 44.93 ± 1.35 123.88 ± 5.21 8.83 ± 0.43 Non-contractile tail) Present study
AS3 VC O1 ElTor MAK 757 90.1 ± 2.21 193.5 ± 14.5 22.8 ± 1.25 Contractile tail Present study
M4 (O1 ElTor typing phage) VC O1 ElTor MAK 757 97.7 ± 0.03 109.6 ± 0.2 18.2 ± 0.4 Contractile tail Chattopadhyay et al., 1993
D10 (O1 ElTor typing phage) VC O1 ElTor MAK 757 62.9 ± 0.06 101.4 ± 0.3 15.8 ± 0.4 Contractile tail Chattopadhyay et al., 1993
MAD-5 O139 typing phage VC O139 NPR-4 58.0 ± 2.7 141.2 ± 4.8 8.33 ± 0.4 Contractile tail Chakrabarti et al., 2000
VE-2 O139 typing phage VC O139 NPR-4 112.5 ± 1.8 204.0 ± 2.8 23.0 ± 0.2 Contractile tail Chakrabarti et al., 2000
Group II classical typing phage VC Classical 154 62.1 ± 3.1 & 65.5 ± 3.7 nm for the widest & narrowest sections 81.0 ± 3.2 16.6 ± 2.0 Contractile tail Chatterjee and Maiti, 1984
Group IV classical typing phage VC Classical 154 73.8 ± 3.3 & 83.6 ± 4.0 for the widest & narrowest sections 152.8 ± 8.2 10.7 ± 1.4 Non-contractile tail (Chatterjee and Maiti, 1984)
DR1 VC O26 77.5 ± 0.3 100.0 ± 0.6 19.0 ± 0.4 Contractile tail (Sarkar et al., 2004) [11]
DR2 VC O39 83.3 ± 0.3 111.0 ± 0.8 17.0 ± 0.5 Contractile tail (Sarkar et al., 2004)
ΦP15 VC O1 ElTor Inaba 52.9 ± 9.0 & 40.5 ± 9.5 for the widest & narrowest sections 105.4 ± 3.2 22.5 nm Contractile tail (Talledo et al., 2003)
(VC stands for Vibrio cholerae)
Acknowledgements
Authors are grateful to Dr. S. K. Bhattacharya, Director of the institute, for encouragement and kind cooperation.
==== Refs
Chatterjee SN Maiti M Vibriophages and Vibriocins: physical, chemical and biological properties Adv Virus Res 1984 29 263 312 6397050
Ramamurthy T Garg R Sharma R Bhattacharya SK Nair GB Simada T Takeda T Karasawa T Kurazano H Pal A Takeda Y Emergence of novel strain of Vibrio cholerae with epidemic potential in southern and eastern India Lancet 1993 314 703 704 8095620 10.1016/0140-6736(93)90480-5
Basu S Mukerjee S Bacteriophage typing of Vibrio ElTor Experimenta 1968 24 299 300
Chattopadhyay DJ Sarkar BL Ansari MQ Chakrabarti BK Roy MK Ghosh AN Pal SC New phage typing scheme for Vibrio cholerae O1 biotype El Tor strains J Clin Microbiol 1993 31 1579 1585 8315000
Chakrabarti AK Ghosh AN Nari GB Niyogi SK Bhattacharya SK Sarkar BL Development and evaluation of a phage-typing scheme for Vibrio choleare O139 J Clin Microbiol 2000 38 44 49 10618061
Chakrabarti BK Chattopadhyay DJ Ghosh AN Vibriophage D10 contains non-permutated DNA with cohesive ends J Gen Virol 1993 74 2749 2752 8277281
Ghosh AN Ansari MQ Dutta GC Isolation and morphological characterization of El Tor cholera phages J Gen Virol 1989 70 2241 2243 2769238
Talledo M Rivera ING Lipp EK Neale A Karolis D Huq A Colwell R Characterization of a Vibrio cholerae phage isolated from coastal water of Peru Environ Microbiol 2003 5 350 354 12713461 10.1046/j.1462-2920.2003.00411.x
Adams MH Bacteriophages 1959 Interscience publishers, Inc New York
Ackerman HW DuBow MS Vibriophages Viruses of Prokaryotes 1987 I CRC press, Inc Florida, USA
Sarkar BL Ghosh AN Sen A Rodrigues DP Newly isolated Vibro cholerae non-O1 non-O139 phages Emerg Infect Dis 2003 10 754 756 15212001
| 15823200 | PMC1087507 | CC BY | 2021-01-04 16:38:58 | no | Virol J. 2005 Apr 11; 2:28 | utf-8 | Virol J | 2,005 | 10.1186/1743-422X-2-28 | oa_comm |
==== Front
Virol JVirology Journal1743-422XBioMed Central London 1743-422X-2-291582320910.1186/1743-422X-2-29MethodologyQuantitative assessment of the effect of uracil-DNA glycosylase on amplicon DNA degradation and RNA amplification in reverse transcription-PCR Kleiboeker Steven B [email protected] Veterinary Medical Diagnostic Laboratory and Department of Veterinary Pathobiology, College of Veterinary Medicine, University of Missouri, Columbia, Missouri 65211, USA2005 11 4 2005 2 29 29 2 3 2005 11 4 2005 Copyright © 2005 Kleiboeker; licensee BioMed Central Ltd.2005Kleiboeker; 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.
Although PCR and RT-PCR provided a valuable approach for detection of pathogens, the high level of sensitivity of these assays also makes them prone to false positive results. In addition to cross-contamination with true positive samples, false positive results are also possible due to "carry-over" contamination of samples with amplicon DNA generated by previous reactions. To reduce this source of false positives, amplicon generated by reactions in which dUTP was substituted for dTTP can be degraded by uracil DNA glycosylase (UNG). UNG does not degrade RNA but will cleave contaminating uracil-containing DNA while leaving thymine-containing DNA intact. The availability of heat-labile UNG makes use of this approach feasible for RT-PCR. In this study, real-time RT-PCR was used to quantify UNG degradation of amplicon DNA and the effect of UNG on RNA detection. Using the manufacturers' recommended conditions, complete degradation of DNA was not observed for samples containing 250 copies of amplicon DNA. Doubling the UNG concentration resulted in degradation of the two lowest concentrations of DNA tested, but also resulted in an increase of 1.94 cycles in the CT for RNA detection. To improve DNA degradation while minimizing the effect on RNA detection, a series of time, temperature and enzyme concentrations were evaluated. Optimal conditions were found to be 0.25 U UNG per 25 μl reaction with a 20 min, 30°C incubation prior to RT-PCR. Under these conditions, high concentrations of amplicon DNA could be degraded while the CT for RNA detection was increased by 1.2 cycles.
==== Body
Background
Molecular techniques have provided a valuable approach for detection of pathogens in both human and veterinary medicine [1-5]. As with any diagnostic technique, quality control of individual steps is critical to ensure the accuracy of results. When performing diagnostic PCR or reverse transcription (RT)-PCR, elimination of false positive results is crucial to ensuring diagnostic accuracy. False positives can occur due to contamination at any point in sample preparation and amplification procedures [6]. For example, cross-contamination between positive and negative samples may occur during sample collection, nucleic acid extraction, PCR or RT-PCR reaction assembly or during agarose gel electrophoresis analysis. In addition to contamination by positive samples, false positive results are also possible due to contamination of samples at any point in the protocol with DNA generated by previous positive amplification reactions. This source of contamination is of particular concern since a positive amplification reaction can generate in excess of 1011 molecules of product (amplicon) DNA per reaction. Given that ten or fewer DNA template molecules can generate a positive result by PCR or RT-PCR, even minute levels of amplicon contamination can result in false positive results. Furthermore, the inherent stability of DNA under a variety of environmental conditions could potentially lead to false positive results weeks or months after contamination of reagents or equipment with amplicon DNA.
Uracil-DNA glycosylase (UNG) is a DNA repair enzyme that will cleave uracil-containing DNA while leaving the natural, thymine-containing DNA unaffected [7,8]. During PCR, deoxyuridine triphosphate (dUTP) can be substituted for deoxythymidine triphosphate (dTTP) in the synthesis of product DNA. Thus to reduce the frequency of false positive results due to amplicon contamination, one common recommendation [9-12] has been to substitute dUTP for dTTP as a source of nucleotides for the PCR reaction. Amplicon DNA that has incorporated dUTP can then be degraded with uracil-DNA glycosylase prior to subsequent amplification reactions, thus preventing these molecules from producing false positive results by acting as template. This approach to elimination of carry-over contamination has led several manufacturers of commercial PCR and RT-PCR reagents to substitute dUTP in the reaction mixture in place of dTTP and in the case of some manufacturers to include UNG as a standard reagent in kits. The success of this approach for elimination of contaminating amplicon DNA depends on the availability of a heat-labile UNG enzyme, the heat-inactivation of which prevents cleavage of product DNA amplified from the target template. The half-life of heat-labile UNG has been estimated to be 2 minutes at 40°C [13] thus making the use of this enzyme feasible for both PCR and RT-PCR applications since the reverse transcription step is commonly performed at 45 – 50°C for 30 min or longer.
Real-time RT-PCR utilizes fluorescence to detect the presence of amplification products as the reaction occurs. The cycle at which a positive reaction is first detectible, termed the cycle threshold (CT) is proportionate to the concentration of template in a sample. Real-time PCR and RT-PCR amplification products are typically much shorter (e.g. 75 – 150 bases in length) compared to those generated by standard PCR and RT-PCR assays. In addition to rapid quantification of template RNA, real-time PCR and RT-PCR offers significant advantages over standard PCR and RT-PCR for detection of DNA or RNA in terms of reduced sample handling, the time required for analysis and analytical sensitivity. However, most importantly real-time assays have reduced (though certainly not eliminated) the opportunities for false positive results due to cross-contamination of samples since real-time assays are conducted in a "closed-tube" system, in which the tubes are not opened after amplification is complete. Nonetheless, given the rigorous standards in place for both human and veterinary diagnostic laboratories and the significant consequences of false positive results, even laboratories using real-time methods may employ strategies such as UNG addition prior to RT-PCR to reduce the potential for false positive results.
While the use of UNG to eliminate amplicon contamination has been previously reported for RT-PCR assays [14,15], the effect of UNG on quantitative assay sensitivity for RNA detection has not been investigated to date. Nor has a quantitative assessment of the concentrations of contaminating DNA that can be degraded prior to RT-PCR been performed. Real-time (quantitative) RT-PCR detection of Porcine arterivirus (family Arteriviridae, order Nidovirales) RNA was used for these assessments. This virus is an important pathogen of swine and is thus frequently the target of diagnostic investigation with RT-PCR representing the principal assay for pathogen detection in many laboratories. Some high-value swine herds are free of this virus thus making the report of false positive results particularly troublesome since depopulation is a common method used to eliminate this virus from a herd. In this study, it was demonstrated that heat-labile UNG had a concentration, temperature and time-dependent effect on quantitative RT-PCR sensitivity and DNA degradation. Conditions were optimized so that minimal effects on target RNA amplification sensitivity were observed while maximizing the ability to degrade carry-over amplicon DNA contamination in a sample.
Results
Effect of UNG concentration on DNA degradation and RT-PCR amplification of RNA
To assess the effect of UNG on DNA degradation and RNA detection, reactions were performed under conditions recommended but the supplier of UNG (Table 1). The template for amplification was 10-fold serial dilutions of viral RNA or amplicon DNA. The amplicon DNA was from a previous RT-PCR reaction in which dUTP was used in place of dTTP. An enzyme concentration and temperature-dependent increase was observed for the CT of both DNA and RNA detection. However, at the enzyme concentration recommended by the supplier (0.5 U UNG per 25 μl reaction), complete degradation of amplicon DNA was not observed at 15°C – 25°C, even in reactions containing less than 250 copies (Table 1). At double the recommended enzyme concentration (1.0 U UNG per 25 μl reaction), complete degradation of amplicon DNA was observed only at 25°C in the two dilutions containing the lowest concentrations of amplicon DNA. Using UNG concentrations recommended by the supplier, increases in RNA CT values were noted ranging from 0.28 cycles, for 0.5 U UNG/reaction and a 15°C incubation, to 1.94 cycles for 1 U UNG/reaction and a 25°C incubation.
Table 1 Effect of UNG concentration and incubation temperature on DNA degradation and RNA detectiona
0.5 U UNG 1.0 U UNG
15°C 20°C 25°C 15°C 20°C 25°C
Analyte (dil.) Controlc CT incr. CT incr. CT incr. CT incr. CT incr. CT incr.
RNA (undil.)b 21.21 21.77 0.56 21.38 0.17 21.81 0.60 22.04 0.83 21.97 0.76 23.34 2.13
RNA (1:10) 24.91 25.11 0.20 25.29 0.38 25.44 0.53 25.10 0.19 26.23 1.32 26.69 1.78
RNA (1:100) 28.32 28.05 -0.27 28.41 0.09 29.23 0.91 28.47 0.15 29.73 1.41 30.13 1.81
RNA (1:1000) 31.16 31.78 0.62 31.59 0.43 32.15 0.99 32.70 1.54 32.70 1.54 33.19 2.03
DNA (1:107)b 21.20 24.22 3.02 26.04 4.84 27.54 6.34 24.67 3.47 28.09 6.89 30.45 9.25
DNA (1:108) 23.91 28.01 4.10 29.20 5.29 31.46 7.55 28.06 4.15 32.30 8.39 34.98 11.07
DNA (1:109) 27.09 31.71 4.62 32.89 5.80 33.50 6.41 32.00 4.91 36.02 8.93 No CTe -
DNA (1:1010) 30.48 34.43 3.95 36.00 5.52 36.24 5.76 36.78 6.30 39.03 8.55 No CT -
Mean CT increase for RNA 0.28 0.27 0.76d 0.68 1.26d 1.94d
Mean CT increase for DNA 3.92 5.36 6.52 4.71 8.19 10.16
a Incubations were performed before RT-PCR with the indicated concentrations of UNG per 25 μl reaction at the indicated temperatures for 10 min.
b The undiluted RNA sample and the 1:107 dilution of the DNA sample contained 250,000 copies of RNA or DNA, respectively
c Control reactions did not contain UNG and were not incubated prior to RT-PCR
d P < 0.05 by paired t-test compared to control values
e Indicates that amplification was not detected through 40 cycles in each of three replicate reactions
To further optimize the effect of UNG on DNA degradation and minimize the effect of UNG on RNA amplification, RT-PCR reactions were performed containing a range of UNG concentrations with a longer incubation and higher temperature than recommended by the enzyme supplier (Table 2). A concentration-dependent effect was observed for DNA degradation, with the lowest UNG concentration tested (0.1 U per 25 μl reaction) increasing the CT for DNA detection but failing to completely eliminate DNA contamination even at the lowest concentration tested. The highest concentrations tested, 0.5 and 1.0 units of UNG per 25 ul reaction, completely eliminated all detectible uracil-containing DNA, including the highest concentration of amplicon DNA tested which contained approximately 250,000 copies in a 25 μl reaction.
Table 2 Effect of UNG concentration on DNA degradation and RNA detectiona
0.1 U UNG 0.25 U UNG 0.5 U UNG 1.0 U UNG
Analyte (dil.) Controlc CT incr. CT incr. CT incr. CT incr.
RNA (undil.)b 21.26 22.30 1.04 22.79 1.53 23.09 1.83 23.86 2.60
RNA (1:10) 26.25 26.86 0.61 27.57 1.32 27.95 1.70 27.31 1.06
RNA (1:100) 31.99 32.52 0.53 33.15 1.16 33.47 1.48 34.04 2.05
RNA (1:1000) 35.08 35.25 0.17 36.73 1.65 36.86 1.78 No Ct -
DNA (1:107)b 21.53 26.56 5.03 35.04 13.51 No Ct - No Ct -
DNA (1:108) 24.53 30.24 5.71 38.34 13.81 No Ct - No Ct -
DNA (1:109) 28.01 33.38 5.37 No Cte - No Ct - No Ct -
DNA (1:1010) 30.82 36.44 5.62 No Ct - No Ct - No Ct -
Mean CT increase for RNA 0.59d 1.42d 1.70d 1.90d
Mean CT increase for DNA 5.43 13.66 N/A N/A
a Incubation was performed before RT-PCR at 30°C for 30 min at the indicated enzyme concentrations
b The undiluted RNA sample and the 1:107 dilution of the DNA sample contained 250,000 copies of RNA or DNA, respectively
c Control reactions did not contain UNG and were not incubated prior to RT-PCR
d P < 0.05 by paired t-test compared to control values
e Indicates that amplification was not detected through 40 cycles in each of three replicate reactions
An UNG concentration-dependent effect was also observed for RNA detection by RT-PCR (Table 2). The lowest UNG concentration tested, 0.1 U per 25 μl reaction, increased the mean CT of RNA detection by 0.59 cycles, while the highest concentration of UNG tested, 1.0 U per 25 μl reaction, increased the mean CT of RNA detection by 1.90 cycles. Intermediate CT increases were observed with 0.25 and 0.5 U UNG per reaction. The increases in CT values for RNA detection following incubation with UNG correspond to 1.5 – 3.7-fold decreases in detectible RNA.
Effect of incubation temperature and time on DNA degradation and RT-PCR amplification of RNA
The effect of temperature on DNA degradation and RNA amplification in the presence of UNG was assessed at three incubation temperatures prior to RT-PCR (Table 3). Incubation at 25°C prior to RT-PCR increased the CT for DNA detection by 8.67 cycles but did not completely eliminate a positive reaction even at the lowest concentration of amplicon DNA. Incubation at 30°C and 35°C eliminated positive reactions at the three lowest concentrations of DNA and increased the CT for the highest DNA concentration by 16.03 and 17.09 cycles, respectively. A temperature dependent increase in CT for RNA amplification was also noted, with incubation temperatures of 25°C and 30°C resulting in smaller increases in CT values for RNA detection than incubation at 35°C.
Table 3 Effect of incubation temperature on DNA degradation and RNA detection in the presence of UNGa
25°C 30°C 35°C
Analyte (dil.) Controlc CT incr. CT incr. CT incr.
RNA (undil.)b 21.22 22.40 1.18 23.17 1.95 23.79 2.57
RNA (1:10) 26.15 26.93 0.78 27.54 1.39 28.25 2.10
RNA (1:100) 31.37 32.51 1.14 32.05 0.68 33.79 2.42
RNA (1:1000) 33.62 35.46 1.84 35.24 1.62 37.43 3.81
DNA (1:107)b 21.48 32.11 10.63 37.51 16.03 38.57 17.09
DNA (1:108) 24.89 34.33 9.44 No Cte - No Ct -
DNA (1:109) 28.22 36.08 7.86 No Ct - No Ct -
DNA (1:1010) 31.75 38.48 6.73 No Ct - No Ct -
Mean CT increase for RNA 1.24d 1.41d 2.73d
Mean CT increase for DNA 8.67 16.03 17.09
a The UNG concentration was 0.25 units per 25 μl reaction and incubation time before RT-PCR was 30 min at the indicated temperatures.
b The undiluted RNA sample and the 1:107 dilution of the DNA sample contained 250,000 copies of RNA or DNA, respectively
c Control reactions did not contain UNG and were not incubated prior to RT-PCR
d P < 0.05 by paired t-test compared to control values
e Indicates that amplification was not detected through 40 cycles in each of three replicate reactions
Incubation times with UNG of 10, 20, and 30 minutes were evaluated for DNA degradation and the effect on RNA amplification by RT-PCR (Table 4). An incubation of 10 min eliminated DNA detection at the lowest concentration and increased the CT by a mean of 6.58 cycles for the other DNA dilutions. Incubation times of 20 and 30 min eliminated progressively more DNA from the reactions. A time-dependent increase in CT values was also observed for RNA detection by RT-PCR, with a 30 min incubation with UNG resulting in the greatest mean increase of 1.42 cycles in the CT.
Table 4 Effect of incubation time on DNA degradation and RNA detection in the presence of UNGa
10 min 20 min 30 min
Analyte (dil.) Controlc CT incr. CT incr. CT incr.
RNA (undil.)b 21.07 21.79 0.72 22.69 1.62 22.74 1.67
RNA (1:10) 25.80 25.96 0.16 26.83 1.03 27.51 1.71
RNA (1:100) 31.28 31.74 0.46 32.15 0.87 32.78 1.50
RNA (1:1000) 34.81 36.16 1.35 36.01 1.20 35.6 0.79
DNA (1:107)b 21.46 28.19 6.73 32.71 11.25 35.93 14.47
DNA (1:108) 24.73 31.22 6.49 36.52 11.79 No Ct -
DNA (1:109) 28.08 34.60 6.52 No Ct - No Ct -
DNA (1:1010) 31.14 No Cte - No Ct - No Ct -
Mean CT increase for RNA 0.67 1.18d 1.42d
Mean CT increase for DNA 6.58 11.52 14.47
a The UNG concentration was 0.25 units per 25 μl reaction and incubation before RT-PCR was performed at 30°C for the indicated times.
b The undiluted RNA sample and the 1:107 dilution of the DNA sample contained 250,000 copies of RNA or DNA, respectively
c Control reactions did not contain UNG and were not incubated prior to RT-PCR
d P < 0.05 by paired t-test compared to control values
e Indicates that amplification was not detected through 40 cycles in each of three replicate reactions
Simultaneous detection of RNA amplification and DNA degradation
To ensure that significant levels of contaminating DNA could be degraded at the same time that RNA was amplified and quantified by real-time RT-PCR, viral RNA was contaminated with amplicon DNA prior to UNG incubation and RT-PCR amplification (Fig. 1). To accomplish this, a constant amount of amplicon DNA generated from a heterologous competitor (which has the viral oligonucleotide primer binding sites but a different recognition sequence for the dual-labeled TaqMan oligonucleotide probe) was added to ten-fold serial dilutions of viral RNA. In reactions which did not contain UNG, approximately 1,000 copies of contaminating DNA were detected in each dilution of viral RNA (Fig. 1A). However, addition of 0.25 U UNG per reaction with incubation at 30°C for 20 min prior to RT-PCR completely eliminated the signal for contaminating DNA in each reaction (Fig. 1B). The amplification curves for viral RNA detection were unaffected other than an increase in CT by a mean of 1.2 cycles, a value consistent with results shown above.
Figure 1 Simultaneous detection of viral RNA and contaminating DNA. Serial ten-fold dilutions of viral RNA with approximately 1,000 copies of heterologous competitor (amplicon) DNA per reaction were amplified following incubation for 20 min at 30°C either without (A) or with (B) 0.25 U UNG per reaction. Reactions were performed in triplicate and contained approximately 250,000 copies viral RNA (●), 25,000 copies viral RNA (▼), 2,500 copies viral RNA (■) or 250 copies viral RNA (◆). Each reaction contained 1,000 copies of amplicon DNA. The fluorescence signal generated by amplicon DNA is indicated by the open form of the same symbol for each respective reaction. The horizontal line at approximately 0.008 fluorescence units (dRn) indicates the threshold for a positive reaction. dRn, baseline-corrected normalized fluorescence.
Effect of UNG and incubation prior to RT-PCR on standard curves for RNA quantification
To assess the effect of UNG addition to reaction mixtures on standard curves used for RNA quantification, reactions were performed using 10-fold dilutions of heterologous competitor RNA (Fig. 2). Standard curves were analyzed for reactions without UNG or containing 0.25 units UNG and were incubated 30 min at 30°C prior to RT-PCR. Regression analyses for these curves were compared to a standard curve using the same RNA dilutions which were not incubated at 30°C for 20 min prior to RT-PCR. Regression analysis for all three reaction conditions demonstrated parallel lines with R2 values >0.99. The slopes of the regression lines were essentially equal at approximately -3.330, which indicate amplification efficiencies near 100%. Only the y-intercepts differed between the three reaction conditions, with the values from the reactions containing UNG and incubated at 30°C for 20 min slightly greater than the standards that did not contain UNG and were not incubated prior to RT-PCR. The y-intercept value for the reactions not containing UNG and incubated at 30°C for 20 min prior to RT-PCR was intermediate of the other two values.
Figure 2 Effect of UNG on Standard curves for competitor RNA. Linear regression analysis for each standard curve was performed within the analysis software (Stratagene Mx4000 version 4.00). Symbols represent means for samples analyzed in triplicate. Addition of 0.25 units UNG per reaction followed by incubation at 30°C for 20 min prior to RT-PCR (●), no addition of UNG but incubation at 30°C for 20 min prior to RT-PCR (○), no addition of UNG and no incubation prior to RT-PCR (▼). CT, cycle threshold. dRn, baseline-corrected normalized fluorescence.
Detection of low viral RNA concentrations in reactions containing UNG
To determine if UNG addition and incubation prior to RT-PCR would result in false negative results under the conditions described, 96 replicate samples containing approximately 20 copies of viral RNA per replicate were amplified by RT-PCR. Of the 96 replicate samples tested, positive amplification (defined as CT < 40) was not detected in three samples that contained 0.25 units UNG and were incubated at 30°C for 30 min prior to RT-PCR (data not shown). Of 96 control reactions (i.e. no UNG added and no incubation prior to RT-PCR), amplification was detected in all but one of the replicates. The mean CT increase of samples containing UNG and incubated at 30°C for 30 min prior to RT-PCR was 1.56 cycles, a value in agreement with results shown above.
Discussion
The techniques of PCR and RT-PCR offer several advantages when compared to traditional viral diagnostic techniques, especially in terms of analytical sensitivity and time for assay completion. Unfortunately the high level of sensitivity, which approaches the single molecule level, also makes this technique prone to false positive results. Amplicon generated by previous positives reactions in which dUTP was substituted for dTTP can be degraded by UNG and thus theoretically eliminated as a source of template that would cause false positives in PCR or RT-PCR. UNG degrades contaminating uracil-containing DNA while leaving the natural, thymine-containing DNA intact. The precise, reproducible quantification of real-time RT-PCR provides a rapid method to assess and optimize the use of UNG to eliminate or reduce the impact of amplicon DNA in RT-PCR as well as determine the effects of UNG on RNA detection. The manufacturers' recommendation for the use of heat-labile UNG include addition of 1 unit enzyme per 50 μl reaction followed by an incubation time of 10 min at 15 – 25°C. In this study, this enzyme concentration and the incubation conditions were not found to eliminate amplicon DNA, presumably due at least in part to the short length of amplicon DNA (114 bases). To consistently degrade approximately 500 copies of amplicon DNA, a two-fold increase in UNG concentration was necessary. However, this concentration of UNG resulted in a nearly two cycle increase to reach the threshold for RNA detection. These results are corroborated by a recent publication using real-time PCR for detection of single-copy genes in which it was concluded that manufacturers' recommended conditions were not adequate to consistently degrade even minimal (e.g. < 30 copies) amounts of amplicon DNA [16].
To determine optimal conditions for amplicon degradation with minimal effect on RNA detection by RT-PCR, a series of time, temperature and enzyme concentrations were evaluated. From these results, it was shown that UNG degradation of amplicon DNA in RT-PCR was concentration, time and temperature depended, as previously described for real-time PCR [16]. To allow degradation of considerable concentrations of carry-over amplicon DNA while having minimal effects on RNA detection, the optimal concentration of UNG was found to be 0.25 U per 25 μl reaction with a pre-amplification incubation at 30°C for 20 min. This incubation was performed after the complete reaction had been assembled in the reaction tubes, just prior to the RT step. These conditions completely degraded approximately 2,500 copies of carry-over amplicon DNA while reducing the detectible concentration of viral RNA template by 2.2-fold (i.e. CT values increased by a mean of 1.2 cycles). If higher levels of carry-over contamination are encountered, increasing the incubation time by an additional 10 min demonstrated degradation of a 10-fold higher concentration of amplicon DNA while only slightly increasing the CT for RNA detection. It is interesting to note that, based on analysis of the standard curves, increases observed in the CT for RNA were due both to the presence of UNG and the incubation at 30°C prior to RT-PCR.
Given the relatively long time required for the reverse transcriptase step, a heat-labile UNG that is rapidly and effectively inactivated at temperatures below that of the RT step must be used for this approach to be applied to control false positive reactions in RT-PCR. The commercially available heat-labile UNG used in this study is rapidly inactivated at 40°C with a half-life of 2 min [13]. At the end of the UNG incubation, the samples were held at 55°C to rapidly inactivate UNG just prior to the RT step which was 50°C for 30 min. For the RT-PCR reagents used (as well as many other commercially availably RT-PCR reagents), the manufacture states that the 30 min RT step can be performed at temperatures of up to 55°C (or higher for some reagent systems), and reactions analyzed in preliminary experiments (data not shown) demonstrated no effect on RT-PCR analytical sensitivity following a 2 min incubation at 55°C prior to the RT step.
Results presented herein demonstrated that incubation with UNG appears to increase the CT equally for in vitro transcribed RNA and viral RNA, thus quantification through the use of a standard curve can remain accurate provided that all reactions are performed under the same conditions. However, a small percentage of weak positive samples may not be detected due to an increase in CT caused by UNG and incubation prior to RT-PCR. To the reduce the possibility of false negatives with dilute RNA samples, results from this study suggest that replicates of two or more reactions should be sufficient, given that the false negative rate was only increased by 2% for samples containing UNG and incubated prior to RT-PCR. Also, two – five additional cycles of amplification may provide positive amplification of target RNA in samples with very low concentrations.
Conclusion
Quantitative assessment of the effect of UNG on DNA degradation and RNA amplification over a range of enzyme concentrations, temperatures and times demonstrated that optimization of reaction conditions allows selection of conditions that maximize carry-over amplicon degradation while minimizing the effect on RNA detection. While this study was performed with real-time RT-PCR to provide accurate quantification of the effects of UNG, these findings are potentially useful to both standard and real-time RT-PCR amplification methods.
Methods
Quantitative (TaqMan) RT-PCR
Amplification reactions were performed using the Qiagen QuantiTect Probe RT-PCR kit (Qiagen, Inc., Valencia, CA) with thermocycling and detection performed in a Stratagene Mx4000 real-time PCR machine (Stratagene, Inc., La Jolla, CA). Samples were analyzed in triplicate. The amplification protocol, oligonucleotide primers and dual-labeled probe used for 5' exonuclease (TaqMan) amplification of the North American PRRSV Ingelvac MLV were as previously described [17] and amplified a 114-bp fragment. The amplified in the presence of dUTP, the sense strand and anti-sense strand will contain 36 and 26 uracil residues, respectively. The dual-labeled probe used for detection of heterologous competitor RNA (and amplicon DNA derived from the competitor RNA) was: 5'-HEX-TGTGCTGCAAGGCGATTAAGTTGGGT-BHQ2-3'. All oligonucleotide primers and dual-labeled probes were synthesized by Integrated DNA Technologies, Inc. (Coralville, IA). Negative control reactions, in which RNA extracted from normal (unaffected) swine tissues or serum was added as template to the RT-PCR reaction mixture, did not produce a signal for the quantitative RT-PCR assay.
Preparation of heterologous competitor RNA
Specific oligonucleotide primer binding sites for the PRRSV real-time RT-PCR assay were incorporated as 5' extensions in PCR primers and in vitro transcribed heterologous competitor RNA was prepared and spectrophotometrically quantified using methods previously described [18].
Viral RNA extraction
Extraction of Porcine arterivirus RNA from cell culture stocks of the vaccine strain Ingelvac MLV (derived by serial passage of U.S. prototype strain VR-2332) was performed using the Qiagen viral RNA kit (Qiagen, Inc., Valencia, CA) according to the manufacturer's instructions. Viral stocks were prepared in MARC-145 cells maintained in Dulbecco's Modified Eagle medium supplemented with 10% heat-inactivated fetal bovine serum and 2 mM L-glutamine, 0.25 μg/ml fungizone, and 0.5 mg/ml gentamycin (all supplied by Mediatech, Inc., Herndon, VA). The viral-infected cell cultures were maintained at 37°C in a humidified 5% CO2 incubator for approximately 2 days until viral cytopathic effect was readily identified throughout the culture. Extraction of heterologous competitor RNA after in vitro transcription was performed using the Qiagen RNeasy kit (Qiagen, Inc., Valencia, CA).
UNG treatment of reactions prior to RT-PCR
The indicated concentrations of heat-labile uracil-DNA glycosylase (Roche Applied Science, Indianapolis, IN), purified from the psychrophilic marine organism BMTU 3346 [13] were added to the amplification master mix prior to dispensing into individual amplification tubes. Samples were then held in the thermocycler at the indicated temperatures for the indicated times prior to RT-PCR. At the end of the incubation, the thermocycler was programmed to ramp the samples at the maximum rate (2.2°C/sec) to 55°C and hold at this temperature for 5 min prior to the 50°C, 30 min RT step. Control reactions, which did not contain UNG and were not subjected to incubation prior to RT-PCR, were placed in the thermocycler at the beginning of the 55°C phase, prior to the 50°C, 30 min RT step.
Competing interests
The author(s) declare that they have no competing interests.
Acknowledgements
The author wishes to thank Sunny J. Troxell for dedicated and expert technical contributions.
==== Refs
Wagar EA Direct hybridization and amplification applications for the diagnosis of infectious diseases J Clin Lab Anal 1996 10 312 325 8951597 10.1002/(SICI)1098-2825(1996)10:6<312::AID-JCLA2>3.0.CO;2-8
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
Sellon RK Update on molecular techniques for diagnostic testing of infectious disease Vet Clin North Am Small Anim Pract 2003 33 677 693 12910738
Ivnitski D O'Neil DJ Gattuso A Schlicht R Calidonna M Fisher R Nucleic acid approaches for detection and identification of biological warfare and infectious disease agents Biotechniques 2003 35 862 869 14579752
Yang S Rothman RE PCR-based diagnostics for infectious diseases: uses, limitations, and future applications in acute-care settings Lancet Infect Dis 2004 4 337 348 15172342 10.1016/S1473-3099(04)01044-8
Borst A Box AT Fluit AC False-positive results and contamination in nucleic acid amplification assays: suggestions for a prevent and destroy strategy Eur J Clin Microbiol Infect Dis 2004 23 289 299 15015033 10.1007/s10096-004-1100-1
Kwok S Hinguchi R Avoiding false positives with PCR Nature 1989 339 237 238 2716852 10.1038/339237a0
Longo MC Berninger MS Hartley JL Use of uracil DNA glycosylase to control carry-over contamination in polymerase chain reactions Gene 1990 93 125 128 2227421 10.1016/0378-1119(90)90145-H
Pang J Modlin J Yolken R Use of modified nucleotides and uracil-DNA glycosylase (UNG) for the control of contamination in the PCR-based amplification of RNA Mol Cell Probes 1992 6 251 256 1406734 10.1016/0890-8508(92)90024-R
Rys PN Persing DH Preventing false positives: quantitative evaluation of three protocols for inactivation of polymerase chain reaction amplification products J Clin Microbiol 1993 31 2356 2360 8408555
Poddar SK Sawyer MH Connor JD Optimized PCR amplification of influenza A virus RNA using Tth DNA polymerase, incorporating uracil N glycosylase (UNG) in a single tube reaction J Clin Lab Anal 1997 11 323 327 9406050 10.1002/(SICI)1098-2825(1997)11:6<323::AID-JCLA2>3.0.CO;2-6
Udaykumar Epstein JS Hewlett IK A novel method employing UNG to avoid carry-over contamination in RNA-PCR Nucleic Acids Res 1993 21 3917 3918 8367319
Sobek H Schmidt M Frey B Kaluza K Heat-labile uracil-DNA glycosylase: purification and characterization FEBS Lett 1996 388 1 4 8654578 10.1016/0014-5793(96)00444-9
Hofmann-Lehmann R Swenerton RK Liska V Leutenegger CM Lutz H McClure HM Ruprecht RM Sensitive and robust one-tube real-time reverse transcriptase-polymerase chain reaction to quantify SIV RNA load: comparison of one- versus two-enzyme systems AIDS Res Hum Retroviruses 2000 16 1247 5127 10957722 10.1089/08892220050117014
Taggart EW Carroll KC Byington CL Crist GA Hillyard DR Use of heat labile UNG in an RT-PCR assay for enterovirus detection J Virol Methods 2002 105 57 65 12176142 10.1016/S0166-0934(02)00080-0
Pierce KE Wangh LJ Effectiveness and limitations of uracil-DNA glycosylases in sensitive real-time PCR assays Biotechniques 2004 36 44 46 48 14740482
Kleiboeker SB Schommer SK Lee S-M Watkins S Chittick W Polson D Simultaneous detection of North American and European Porcine reproductive and respiratory syndrome virus using real-time quantitative RT-PCR J Vet Diagn Invest 2005 17 165 170 15825498
Kleiboeker SB Applications of competitor RNA in diagnostic reverse transcription-PCR J Clin Microbiol 2003 41 2055 2061 12734248 10.1128/JCM.41.5.2055-2061.2003
| 15823209 | PMC1087508 | CC BY | 2021-01-04 16:38:59 | no | Virol J. 2005 Apr 11; 2:29 | utf-8 | Virol J | 2,005 | 10.1186/1743-422X-2-29 | oa_comm |
==== Front
Virol JVirology Journal1743-422XBioMed Central London 1743-422X-2-301582320510.1186/1743-422X-2-30ResearchHost-derived pathogenicity islands in poxviruses Da Silva Melissa [email protected] Chris [email protected] Department of Biochemistry and Microbiology, University of Victoria, Victoria, Canada2005 11 4 2005 2 30 30 11 2 2005 11 4 2005 Copyright © 2005 Da Silva and Upton; licensee BioMed Central Ltd.2005Da Silva and Upton; 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
Poxviruses are important both as pathogens and as vaccine vectors. Poxvirus genomes (150–350 kb) consist of a single linear dsDNA molecule; the two polynucleotide strands are joined by short hairpin loops. The genomes encode highly conserved proteins required for DNA replication and mRNA transcription as well as a variable set of virulence factors; transcription takes place within the cytoplasm of the host cell. We are interested in evolution of poxvirus genomes and especially how these viruses acquire host-derived genes that are believed to function as virulence factors.
Results
Using a variety of bioinformatics tools, we have identified regions in poxvirus genomes that have unusual nucleotide composition (higher or lower than average A+T content) compared to the genome as a whole; such regions may be several kilobases in length and contain a number of genes. Regions with unusual nucleotide composition may represent genes that have been recently acquired from the host genome. The study of these genomic regions with unusual nucleotide content will help elucidate evolutionary processes in poxviruses.
Conclusion
We have found that dotplots of complete poxvirus genomes can be used to locate regions on the genome that differ significantly in A+T content to the genome as a whole. The genes in these regions may have been acquired relatively recently from the host genome or from another AT-rich poxvirus.
==== Body
Background
Poxviruses comprise a family of large dsDNA viruses that replicate in the cytoplasm of eukaryotic cells; they have a wide host range that includes insects, reptiles, birds and mammals [1]. The poxviruses are relatively self-sufficient and encode proteins responsible for the virion structure, DNA replication, mRNA transcription as well as a number of virulence factors [1]. Most poxviruses are moderately AT-rich (55–75%), but there are several exceptions. The known insect poxviruses are extremely AT-rich (82% A+T) whereas molluscum contagiosum virus (MOCV-1) and the known parapoxviruses are AT-poor (36–37% A+T) [1].
When making comparisons of large viral genomes such as poxviruses there are several approaches depending on the level of resolution that is required. Although dotplots can be used for any length of DNA sequence and also for protein sequences, they are especially useful when trying to get a global view of the relationship between large DNA sequences. Since they generate a graphical view of sequence similarity, some relationships between the sequences may be apparent that would otherwise be missed in text formatted alignments [2]. In the simplest form of a dotplot, one sequence is placed along the x-axis and the other sequence along the y-axis to create a matrix (Figure 1). Wherever one nucleotide of one sequence is identical to a nucleotide of the other sequence, the appropriate cell of the matrix is filled. Most dotplot programs, however, use a sliding window of user-defined size, and a measurement of similarity between the two windows instead of comparing each nucleotide along the genomes. When plotting the data, a greyscale can be used to record differing degrees of similarity. If the two sequences are identical, a fully black diagonal line is observed on the dotplot comprised of many dots, each drawn for a window that matches an identical window on the second sequence. The background of the dotplot is therefore comprised of greyscale dots representing random matches of varying similarity between the windows used to scan the two sequences. Figure 1 shows a simplified example of a typical dotplot. For each region of the sequence plotted on the x-axis that is identical to the sequence on the y-axis, a diagonal line is plotted and, for the sake of this example, shaded an appropriate colour to show where each region is identical. Dotter [2] and our java version of this program (JDotter) [3] are especially useful implementations because the window size and degree of similarity used for the dotplot display can both be manipulated on the fly, without recalculation of the whole plot which is CPU intensive. Changing these parameters can help visualize regions of relatively low similarity or to enhance/decrease the background matches in the plot.
Figure 1 An example of a typical dotplot. Regions on the x-axis that are identical or similar to regions on the y-axis are marked A, B and C and correspond to regions X, Y, and Z on the y-axis respectively. Regions that correspond to each other are also colour-coded in blue for regions A and X, pink for regions B and Y and turquoise for regions C and Z.
When viewing many of the dotplots comparing different poxvirus genomes plotted against each other and themselves, we observed unexpected striped patterns in the background of the dotplots (Figure 2). These unusual non-random banding patterns suggested that the composition of discrete regions of the genomes differed significantly from the overall composition of the poxvirus genomes.
Figure 2 Dotplot depicting a comparison of molluscum contagiosum virus genome to itself. Region 1 begins at position 71,550 bp and ends at position 74,790 bp and contains 5 genes. Region 2 begins at position 151,470 bp and ends at 157,950 bp and contains 5 genes. Blue and red bars across each axis represent each gene as it is located on the MOCV-1 genome with red bars representing genes located on the top strand and blue bars representing genes located on the bottom strand.
Results and discussion
A self-dotplot of MOCV-1, where the viral genome is plotted on both the x- and y-axes, is shown in Figure 2. Since the comparison uses the identical genome on each axis, there is an unbroken diagonal line running from the upper left corner to the lower right corner of the dotplot. A number of horizontal and vertical stripes can be seen scattered throughout the plot; such patterns are observed in most poxvirus genome self-plots although the intensity varies considerably between stripes in a single plot and are usually more intense in those genomes with the most extreme nucleotide composition, A+T or G+C rich genomes. Two of the most striking regions for MOCV-1 are marked on Figure 2. Regions 1 and 2 are located at 71,550 – 74,790 bp and 151,470 – 157,950 bp on the MOCV-1 genome, respectively; each region encompasses 5 genes. It was evident from the dotplot itself that sets of small DNA repeats, which would appear as short parallel lines, were not responsible for generating these unusual stripes in the background of the plot. Therefore, the hypothesis that variations in nucleotide composition were responsible for the patterns was tested. Figure 3 shows that these two regions marked in Figure 2 do indeed have a G+C composition significantly lower than the genome average. Regions 1 and 2 have G+C averages of 52.58% and 50.38% respectively, whereas the G+C composition of MOCV-1 is 63.36% (also shown as a horizontal line in Figures 3A and 3B).
Figure 3 G+C composition plots created using viral genome organizer (VGO) [14] using a window size of 135 bp for (A) region 1 of Figure 2 and (B) region 2 of Figure 2. Scale on left hand side of plot shows the maximum (91.85%) and minimum (36.30%) G+C content of the MOCV-1 genome. Line drawn through the plot indicates the average G+C content of the MOCV-1 genome (63.36%). Numbered blue bars represent the genes in each region and the scale across the top of each panel shows the location of each gene on the genome.
The genes present in regions 1 and 2 are listed in Table 1 together with the A+T content, G+C content and putative function. Region 1 contains 5 genes (MOCV-1-051L, MOCV-1-052R, MOCV-1-053L, MOCV-1-054L, MOCV-1-055R), two of which have unknown function (MOCV-1-052R and MOCV-1-055R). Interestingly, three of the five genes in region 1 show significant similarity to each other. The gene product of MOCV-1-054L is a secreted glycoprotein that binds interleukin-18 [4] and shows some sequence similarity to human and mouse IL-18 binding proteins with three conserved cysteine residues shared between the human and MOCV-1 IL-18 binding proteins [5]; although no function has been predicted for the hypothetical products of related genes MOCV-1-051L and MOCV-1-053L, they are predicted to be secreted from infected cells. Since the other poxviruses that encode an ortholog of the MOCV-1 interleukin-18 binding protein do not contain orthologs of MOCV-1-051L and MOCV-1-053L, it appears that these genes may have arisen from duplications of MOCV-1-054L after the divergence of MOCV from the other poxvirus families.
Table 1 Description of genes in regions 1 and 2. Gene names, protein length, AT%, GC% and putative function of all genes described located in regions 1 and 2 (Figure 2).
Region Gene Name Protein length (amino acids) G+C% A+T% Putative Function
1 MOCV-1-051L 120 55.92 44.08 Secreted glycoprotein
MOCV-1-052R 100 54.54 45.54 Unknown
MOCV-1-053L 133 55.47 44.53 Secreted glycoprotein
MOCV-1-054L 235 58.05 41.95 Secreted IL-18 binding protein
MOCV-1-055R 216 55.76 44.24 Unknown
2 MOCV-1-130L 451 57.01 42.99 A-Type inclusion (ATI) protein
MOCV-1-131L 513 56.48 43.52 Intracellular mature virion surface protein (ATI-factor)
MOCV-1-132L 229 58.7 41.31 Unknown
MOCV-1-133L 546 57.4 42.6 Intracellular mature virion surface protein (ATI-factor)
MOCV-1-134L 141 58.22 41.78 Intracellular mature virion membrane protein (associated with virus entry)
Region 2 is also comprised of 5 genes (MOCV-1-130L, MOCV-1-131L, MOCV-1-132L, MOCV-1-133L, and MOCV-1-134L). Three of the five genes (MOCV-1-130L, MOCV-1-131L and MOCV-1-133L) have low, but significant, similarity to each other and also to both the A-type inclusion (ATI) proteins and orthologs of the vaccinia virus P4c gene (A26L; strain Copenhagen). In some strains of cowpox virus, the ATI protein functions to surround intracellular mature virus (IMV) particles in the cytoplasm of the host cell with the P4c protein playing a role in directing the IMV particles to the A-type inclusions [6,7]. Some orthopoxviruses such as vaccinia and variola encode a presumed non-functional, truncated ATI protein that is incapable of forming occluded ATIs suggesting that ATI formation is not essential for virus survival in the host cell [8]. Little is known about the role of the ATI proteins in the MOCV-1 replication cycle, however, given that these proteins are truncated compared to their cowpox orthologs, it is likely that they have a somewhat different, if any, function in MOCV-1. There are two other genes in region 2; MOCV-1-132L is not found in any other poxvirus and its function is unknown, and MOCV-1-134L is conserved in all poxviruses and shows significant sequence similarity (50% amino acid identity) to the A28L gene of vaccinia virus strain Copenhagen and is an intracellular mature virion membrane protein which is associated with virus entry into the host cell, [9,10].
There are several possible explanations for the observed differences in G+C content of the genes in regions 1 and 2 compared to the rest of the genome. The MOCV-1-054L gene in region 1, for example, was most likely recently acquired from an AT-rich host or from an AT-rich virus with subsequent duplications and divergence resulting in the MOCV-1-051L and MOCV-1-053L genes. Since these genes make up a relatively AT-rich region on the MOCV-1 genome, they may have served as a target site for the acquisition, presumably through non-homologous recombination, of genes MOCV-1-052R and MOCV-1-055R neither of which have been assigned a function. Given that the ATI genes found in region 2 are relatively AT-rich compared to the MOCV-1 genome, these genes may have also been acquired relatively recently from another AT-rich poxvirus. Similar to the genes in region 1, an initial acquisition event leading to an AT-rich region may have led to further acquisition events and the creation of region 2. Although the MOCV-1-134L gene appears to have been acquired from an AT-rich poxvirus, the situation is not simple because the gene has been found to be essential for vaccinia virus entry into the host cell [9,10]. One explanation is that an essential MOCV-1 ortholog was replaced by a similarly functioning gene from an AT-rich poxvirus, thus explaining why it differs in G+C content despite being an essential gene.
It is interesting to note that there is a short sequence in the AT-rich region 1 that has a significantly higher G+C composition than the rest of the region (Figure 3A). This area can be seen on the dotplot as a thin, dark stripe located in the middle of the band for region 1 (Figure 2). This short spike of G+C rich DNA is located at the 3' end of the MOCV-1-054L gene in a region that does not align with either of the two related MOCV-1 genes (MOCV-1-051L and MOCV-1-053L) indicating that this region may have also been acquired relatively recently; this region results in an in-frame insertion in the predicted MOCV-1-054L polypeptide.
To confirm that the unusual stripes seen in the background of the MOCV-1 dotplot could be due to a lower than average G+C content of these particular regions of the genome, a dotplot was created with the MOCV-1 genome plotted on the x-axis against a 250 kb DNA sequence consisting of five 50 kb randomized DNA segments with specific G+C content on the y-axis (Figure 4). Regions 1 and 2 are marked on the figure and the individual 50 kb segments are clearly visible on the dotplot. As the G+C content increases along the 250 kb sequence on the y-axis, the stripes seen for regions 1 and 2 change colour from dark to light; this reflects a reduction in random matches between the two sequences. At 60% G+C content, the bands for both regions 1 and 2 appear similar to those observed in the MOCV-1 self-dotplot (Figure 2). When the randomized sequence has a G+C content of 50%, the stripes for regions 1 and 2 appear to disappear, as they merge with the background, indicating that the G+C content of these regions is approximately 50%. As the G+C content further decreases to less than 40%, the bands seen for regions 1 and 2 are again visible but are the negative image of those seen in Figure 2; the bands are darker than the background because these regions have a G+C content that is more similar to the randomized sequence than the rest of the MOCV-1 genome. Similarly, each of the 50 kb segments of the random sequence used for Figure 4 produce a different level of background matches in the dotplot and appears as a broad horizontal stripe.
Figure 4 Dotplot comparing molluscum contagiosum virus to a randomly generated 250 kb sequence made up of 50 kb segments with differing A+T content which is indicated on the dotplot. Regions corresponding to the regions seen in Figure 2 are indicated with arrows.
Locating regions of differing G+C content using dotplots can be a difficult task given that the intensity of the banding patterns observed are often diminished in poxvirus genomes that are not extremely AT- or GC-rich. However, we noticed that these banding patterns were more distinct when comparing the MOCV-1 genome to a random sequence with increasing A+T content (Figure 4). For example, areas on the MOCV-1 genome with higher than average G+C content that were previously undetected when observing a self-plot of the MOCV-1 genome can be visualized on the dotplot of the MOCV-1 genome versus a random sequence with decreasing A+T content (compare Figures 2 and 4). Thus this type of comparison can be used to enhance recognition of regions of unusual nucleotide composition.
In order to further illustrate the differences between the genes in regions 1 and 2 compared to the whole genome, the codon usage of 49 MOCV-1 genes that are conserved in all poxviruses was compared to the codon usage of the 10 genes in regions 1 and 2. The mean relative synonymous codon usage (RSCU) values of the two datasets were compared using a Student's T-test. The codons that were found to have statistically different codon usage between the two datasets are listed in Table 2. We found that the codon usage of 43 of 62 codons (69%) was statistically different between all genes in regions 1 and 2 and the 49 conserved MOCV-1 genes. Codon usage was deemed statistically different when the p-value was less than 0.05. The amino acids that have at least 1 codon with statistically equal RSCU values are leucine, isoleucine, valine, serine, proline, threonine, alanine, arginine, glycine and two stop codons (UAA and UAG).
Table 2 Codon usage differences between regions 1 and 2 and 49 MOCV-1 genes conserved in all poxviruses. The codon usage between regions 1 and 2 was first determined to be statistically equal and was then used to compare to the codon usage of 49 MOCV-1 genes that are conserved in all poxviruses. Student's t-test was used to compare codon usage with null hypothesis assuming codon usage was equal.
Amino Acid Fraction of codons with different usage Codons that have statistically different usage*
Phe 2/2 UUU, UUC
Tyr 2/2 UAU, UAC
His 2/2 CAU, CAC
Gln 2/2 CAA, CAG
Asn 2/2 AAU, AAC
Lys 2/2 AAA, AAG
Asp 2/2 GAU, GAC
Glu 2/2 GAA, GAG
Cys 2/2 UGU, UGC
Leu 4/6 UUA, UUG, CUU, CUG
Ile 1/3 AUC
Val 2/4 GUU, GUG
Ser 2/6 UCU, AGC
Pro 3/4 CCU, CCC, CCA
Thr 2/4 ACU, ACA
Ala 2/4 GCU, GCA
Arg 5/6 CGU, CGC, CGA, CGG, AGA
Gly 3/4 GGU, GGC, GGA
Stop 1/3 UGA
*Codon usage was considered different if p-value < 0.05
In order to determine whether the genes in regions 1 and 2 were recently acquired from the host genome, the codon usage of the genes in regions 1 and 2 was compared to the codon usage of 50 human genes (MOCV-1's natural host), using the same method used for the comparison of regions 1 and 2 with the 49 conserved MOCV-1 genes. The codon usage of the genes in regions 1 and 2 was found to be 68% (48/62 codons) similar to the codon usage of the 50 human genes tested (Table 3). As a control, the codon usage of the same 50 human genes was compared to the codon usage of the 49 conserved MOCV-1 genes and was found to be statistically different in 56 of 62 codons (90%) (Table 4).
Table 3 Codon usage differences between regions 1 and 2 and 50 human genes.
Amino Acid Fraction of codons with different usage Codons that have statistically different usage*
Phe 0/2 -
Tyr 0/2 -
His 1/2 CAC
Gln 0/2 -
Asn 0/2 -
Lys 0/2 -
Asp 2/2 GAU, GAC
Glu 2/2 GAA, GAG
Cys 0/2 -
Leu 2/6 UUG, CUG
Ile 0/3 -
Val 0/4 -
Ser 1/6 UCG
Pro 2/4 CCU, CCG
Thr 2/4 ACC, ACG
Ala 2/4 GCU, GCG
Arg 4/6 CGU, CGC, AGA, AGG
Gly 0/4 -
Stop 2/3 UAG, UGA
*Codon usage was considered different if p-value < 0.05
Table 4 Codon usage differences between 49 MOCV-1 genes that are conserved in all poxviruses and 50 human genes.
Amino Acid Fraction of codons with different usage Codons that have statistically different usage*
Phe 2/2 UUU, UUC
Tyr 2/2 UAU, UAC
His 2/2 CAU, CAC
Gln 2/2 CAA, CAG
Asn 2/2 AAU, AAC
Lys 2/2 AAA, AAG
Asp 2/2 GAU, GAC
Glu 2/2 GAA, GAG
Cys 2/2 UGU, GUC
Leu 6/6 UUA, UUG, CUU, CUC, CUA, CUG
Ile 3/3 AUU, AUC, AUA
Val 3/4 GUU, CUA, GUG
Ser 5/6 USU, UCA, UCG, AGU, AGC
Pro 4/4 CCU, CCC, CCA, CCG
Thr 3/4 ACU, ACA, ACG
Ala 3/4 GCU, CGA, GCG
Arg 6/6 CGU, CGC, CGA, CGG, AGA, AGG
Gly 3/4 GGU, GGC, GGA
Stop 2/3 UAG, UGA
*Codon usage was considered different if p-value < 0.05
Since the codon usage of the genes in regions 1 and 2 was found to be statistically different to the codon usage of 49 conserved MOCV-1 genes for 69% of the codons tested yet was found to be only 34% different from the codon usage of the 50 human genes tested, it is reasonable to assume that these genes may be recent acquisitions from the host genome or a poxvirus genome with higher A+T composition. Interestingly, since the initial submission of this manuscript, a paper has been published which identifies the MOCV-1 interleukin-18 binding protein (as well as several other poxvirus proteins) as a possible horizontally transferred gene [11], which further supports our hypothesis that the genes in regions 1 and 2 were likely acquired from external sources.
Pathogenicity islands (PAI) in bacterial genomes contain several distinct structural features, which include the presence of virulence genes within the PAI, a nucleotide composition of the PAI that is different from the remainder of the genome, and occupancy of large regions of the bacterial chromosome [12]. The genes in regions 1 and 2 also occupy a relatively large region of the MOCV-1 genome, and some of the genes in these regions are known virulence factors with the function of the remaining genes yet to be determined. Thus, we suggest that these regions may be analogous to bacterial pathogenicity islands. This notion is further supported by the observations in this manuscript, which showed the differences in nucleotide composition and codon usage of the genes in regions 1 and 2.
Conclusion
The data presented in this paper suggest that the unusual striped banding pattern seen on the dotplots of poxvirus genomes is due to differences in nucleotide content of the genes in these regions compared to the remainder of the genome. The difference between the genes in these regions and 49 MOCV-1 genes that are conserved in all poxviruses was further shown by comparing the codon usage of 62 codons in each of these datasets. The codon usage was found to differ significantly in 69% of the codons tested. The codon usage of the genes in regions 1 and 2 was found to be statistically similar to 50 human genes in 68% of the codons tested.
We conclude that the genes in regions 1 and 2 have a relatively low G+C content and that they may have been acquired from either an AT-rich host or virus making these regions possible novel pathogenicity islands in the MOCV-1 genome. A future survey of all other poxviruses is needed in order to determine the extent of these possible gene acquisitions.
Methods
Creation of dotplots
Dotplots for the molluscum contagiosum virus genome were created and visualized using JDotter with a default window size of 26 nucleotides using a minimum cut-off score of 40 and a maximum cut-off score of 100 on the GreyMap tool in order to better visualize the unique background patterns [3].
The dotplot shown in Figure 4 was created using the Dotter program with the complete molluscum contagiosum virus genome plotted on the x-axis and a 250 kb random sequence with increasing G+C content plotted on the y-axis [2]. The 250 kb sequence was created using DNACreator a program that creates a random sequence with a given G+C content [13]. Segments of 50 kb and varying G+C content were created and concatenated into one 250 kb sequence that contained an increasing G+C content.
G+C composition plots
The G+C composition of regions 1 and 2 was plotted using the "nucleotide base content" feature of the program Viral Genome Organizer (VGO) [14]. This feature calculates and plots the G+C content of the entire genome using a user-specified window size.
Codon usage
Comparison between genes in regions 1 and 2 and 49 conserved MOCV-1 genes
The program CodonW was used to calculate the Relative Synonymous Codon Usage (RSCU) values of 49 molluscum contagiosum virus genes that are conserved in all poxviruses, and the 5 genes comprising each of region 1 and 2 (Figure 2 and Table 1) [15]. The mean RSCU values for each codon excluding the codons for methionine and tryptophan (62 in total) from regions 1 and 2 were initially compared to each other using a two-tailed Student's T-test with the null hypothesis that the means are equal being rejected only if the p-value was less than 0.05. From the results of these T-tests, the RSCU values for each region were pooled in order to create a larger data set to perform subsequent Student's T-tests. The mean RSCU values for four codons in regions 1 and 2 were found to be statistically different and were therefore treated as individual data sets when subsequent T-tests were performed. These codons were CCA (proline), CAU and CAC (histidine), and CAG (glutamine).
With the RSCU values from regions 1 and 2 being pooled for all but 4 codons, the mean values for this pooled dataset were compared to the mean RSCU values obtained for the 49 conserved MOCV-1 genes again using a two-tailed Student's T-test. The null hypothesis that mean RSCU values from the two datasets are equal was rejected only when the p-value was less than 0.05.
Comparison between MOCV-1 and 50 human genes
Using the above outlined method, the mean RSCU values for the genes in regions 1 and 2 were compared to the mean RSCU values of 50 randomly selected human genes. The 50 human genes were randomly selected from the dataset used by the codon usage database at the Kazusa DNA Research Institute [16] that consists of 76,893 human coding sequences that were taken from the NCBI Genbank FlatFile release 145.0 (January 25, 2005). The comparison of the codon usage of 50 human genes with the codon usage of 49 MOCV-1 genes was performed using the same methods outlined above. In each of the analyses performed, the null hypothesis that mean RSCU values from the two datasets are equal was rejected when the p-value was less than 0.05
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
CU initiated the project and edited the manuscript; MDS performed all the work and wrote the manuscript.
Acknowledgements
The authors are grateful to Angelika Ehlers, Julius Litorco and Norman Jordan for software development and support and to David Esteban for critical review of the manuscript.
==== Refs
Moss B Fields BN, Knipe DM, Howley PM Poxviridae: The viruses and their replication Fundamental Virology 1996 Third Philadelphia , Lippincott-Raven Publishers 1163 1197
Sonnhammer EL Durbin R A dot-matrix program with dynamic threshold control suited for genomic DNA and protein sequence analysis Gene 1995 167 GC1 10 8566757 10.1016/0378-1119(95)00714-8
Brodie R Roper RL Upton C JDotter: a Java interface to multiple dotplots generated by dotter Bioinformatics 2004 20 279 281 14734323 10.1093/bioinformatics/btg406
Xiang Y Moss B IL-18 binding and inhibition of interferon gamma induction by human poxvirus-encoded proteins Proc Natl Acad Sci U S A 1999 96 11537 11542 10500212 10.1073/pnas.96.20.11537
Xiang Y Moss B Identification of human and mouse homologs of the MC51L-53L-54L family of secreted glycoproteins encoded by the Molluscum contagiosum poxvirus Virology 1999 257 297 302 10329540 10.1006/viro.1999.9676
Funahashi S Sato T Shida H Cloning and characterization of the gene encoding the major protein of the A-type inclusion body of cowpox virus J Gen Virol 1988 69 35 47 2826668
McKelvey TA Andrews SC Miller SE Ray CA Pickup DJ Identification of the orthopoxvirus p4c gene, which encodes a structural protein that directs intracellular mature virus particles into A-type inclusions J Virol 2002 76 11216 11225 12388681 10.1128/JVI.76.22.11216-11225.2002
Ulaeto D Grosenbach D Hruby DE The vaccinia virus 4c and A-type inclusion proteins are specific markers for the intracellular mature virus particle J Virol 1996 70 3372 3377 8648667
Senkevich TG Ward BM Moss B Vaccinia virus A28L gene encodes an essential protein component of the virion membrane with intramolecular disulfide bonds formed by the viral cytoplasmic redox pathway J Virol 2004 78 2348 2356 14963131 10.1128/JVI.78.5.2348-2356.2004
Senkevich TG Ward BM Moss B Vaccinia virus entry into cells is dependent on a virion surface protein encoded by the A28L gene J Virol 2004 78 2357 2366 14963132 10.1128/JVI.78.5.2357-2366.2004
Hughes AL Friedman R Poxvirus genome evolution by gene gain and loss Mol Phylogenet Evol 2005 35 186 195 15737590 10.1016/j.ympev.2004.12.008
Schmidt H Hensel M Pathogenicity islands in bacterial pathogenesis Clin Microbiol Rev 2004 17 14 56 14726454 10.1128/CMR.17.1.14-56.2004
HU Z DNACreator version 1.0 Center for Computational Research at State University of New York at Buffalo 2002
Upton C Hogg D Perrin D Boone M Harris NL Viral genome organizer: a system for analyzing complete viral genomes Virus Res 2000 70 55 64 11074125 10.1016/S0168-1702(00)00210-0
Peden J Analysis of Codon Usage PhD Dissertation, University of Nottingham 1999
Nakamura Y Gojobori T Ikemura T Codon usage tabulated from international DNA sequence databases: status for the year 2000 Nucleic Acids Res 2000 28 292 10592250 10.1093/nar/28.1.292
| 15823205 | PMC1087509 | CC BY | 2021-01-04 16:38:59 | no | Virol J. 2005 Apr 11; 2:30 | utf-8 | Virol J | 2,005 | 10.1186/1743-422X-2-30 | oa_comm |
==== Front
Virol JVirology Journal1743-422XBioMed Central London 1743-422X-2-351583311310.1186/1743-422X-2-35ReviewMolecular advances in the cell biology of SARS-CoV and current disease prevention strategies Stark Caren J [email protected] CD [email protected] Division of Viral Products, Center for Biologics Evaluation and Research, US Food and Drug Administration, Bethesda, MD 20892 USA2005 15 4 2005 2 35 35 13 4 2005 15 4 2005 Copyright © 2005 Stark and Atreya; licensee BioMed Central Ltd.2005Stark and Atreya; 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.
In the aftermath of the SARS epidemic, there has been significant progress in understanding the molecular and cell biology of SARS-CoV. Some of the milestones are the availability of viral genome sequence, identification of the viral receptor, development of an infectious cDNA clone, and the identification of viral antigens that elicit neutralizing antibodies. However, there is still a large gap in our understanding of how SARS-CoV interacts with the host cell and the rapidly changing viral genome adds another variable to this equation. Now the SARS-CoV story has entered a new phase, a search for preventive strategies and a cure for the disease. This review highlights the progress made in identifying molecular aspects of SARS-CoV biology that is relevant in developing disease prevention strategies. Authors conclude that development of successful SARS-CoV vaccines and antivirals depends on the progress we make in these areas in the immediate future.
AntiviralsCell biologyMolecular virologySARS-CoVVaccines
==== Body
Introduction
Following reports of the last case of the severe acute respiratory syndrome (SARS) epidemic in July 2003, there has been remarkable progress in several areas of research on the molecular identification of the pathogen and its pathogenesis, replication, genetics, and host immunogenicity, as well as elegant epidemiological studies. The sequence of epidemiological events that unfolded early in the outbreak gave researchers a glimpse into the first new pathogen of the era of globalization. As the year 2002 drew to a close, multiple reports of an "infectious atypical pneumonia" caught public health officials across the globe by surprise and suggested that a new human pathogen had emerged in the Guangdong Province in China [1]. By the end of February 2003, this outbreak of SARS had infected almost 800 patients and caused 31 deaths in the Province [2]. One month later, the disease had spread throughout Asia and into Europe and North America. This epidemic eventually affected more than 8000 people and resulted in approximately 800 deaths worldwide, with mortality rates reaching over 40% in certain populations [3,4].
Electron microscope analysis quickly identified the putative SARS agent as having features associated with coronaviruses. The SARS agent was later unambiguously identified as a new coronavirus member and named SARS-coronavirus (SARS-CoV) [5-7]. Coronaviruses are enveloped, plus-stranded RNA viruses with the largest RNA genomes known (on the order of 30 kb). Coronaviruses have long been important in the world of veterinary viral diseases. However, previously known human coronaviruses such as HCoV-229E and HCoV-OC43 cause only minor health problems such as the common cold and gastrointestinal diseases. In contrast, the SARS-CoV pathogen causes fever, pulmonary edema, and diffuse alveolar damage in severely affected individuals (collectively termed severe acute respiratory syndrome) [8]. SARS-CoV is also a unique coronavirus in that, to date, it is the only member known to cause severe morbidity and mortality in humans [8]. Demonstration that SARS-CoV can cause serious public health problems has focused attention on the need to understand the viral replicative strategy and devise prophylactic measures.
The clinical symptoms of SARS are those of a lower respiratory tract infection and are accompanied by damage to the lungs [6,9,10]. Gastrointestinal involvement is also common, with more than 20% of patients presenting with watery diarrhea [11]. Fecal samples from SARS patients taken up to 25 days after onset of disease contain viral RNA, which suggests viral shedding through the bowels [5]. Liver dysfunction has also been reported based on observed necrosis in hepatocytes [9,12]. Post-mortem tissue examination of SARS patients has found the virus presence in lung, bowel, lymph node, liver, heart, kidney, and skeletal muscle samples [13]. The primary mode of SARS-CoV transmission is airborne via droplets [14,15]. However, there are also reports of the presence of replicating virus in blood cells (peripheral blood mononuclear cells) and in the small and large intestine [11,16]. Alternative modes of transmission, such as blood-borne or fecal-oral are therefore possible.
The virus has been isolated from wild animals (Himalayan palm civets and raccoon dogs) found in the animal markets of Guangdong, China [17]. The actual natural reservoir for SARS-CoV is still unknown. Once transmitted to humans, SARS-CoV appears to evolve to facilitate to human-human transmission. Sequence analysis of different SARS-CoV isolates from early in the epidemic show deletion events occurring in open reading frame 8 (Orf 8) [18]. Identical deletions in Orf 8 have also been seen in animal coronaviruses supporting the idea that SARS-CoV was introduced to humans via an animal intermediate. In addition to deletion events occurring early and late in the epidemic, a slowing of missense mutations is seen over time, with the most extensive changes occurring in the S protein during the early stages of the outbreak [18]. This suggests the virus has undergone some level of adaptation but has ultimately stabilized at a time in the epidemic where SARS-CoV has become more virulent. Deciphering the evolutionary passage of this virus will undoubtedly provide valuable information on preventing future outbreaks.
In the wake of the SARS epidemic, a number of excellent review articles on the clinical and molecular aspects of SARS epidemiology have been published. These reviews have focused primarily on rapid advances made in the identification and characterization of SARS-CoV genomes as well as describing the etiology of the virus and clinical features of the disease [19-21]. Now the SARS-CoV story has entered a new phase, a search for preventative strategies and a cure. In this review, we highlight the progress made in revealing the molecular aspects of SARS-CoV biology and how such information may lead to strategies for disease prevention.
Brief overview of the SARS-CoV genome
Coronaviruses are subdivided into three groups based on genetic and serological markers [22]. Groups I, and II infect mammals while group III is specific for avian species. Group I members are the porcine transmissible gastroenteritis virus (TGEV) and epidemic diarrhea virus (PEDV), feline and canine coronavirus (FCoV and CCoV), and human coronavirus 229E (HCoV-229E). Group II includes porcine hemagglutinating encephalomyelitis virus (HEV), murine hepatitis virus (MHV), bovine, equine, and rat coronavirus (BCoV, ECoV, and RtCoV), and human coronavirus OC43 (HCoV-OC43). Group III includes the turkey coronavirus (TCoV), pheasant coronavirus and avian infectious bronchitis virus (IBV). Although most closely related to Group II coronaviruses, SARS-CoV, with some of its unique genetic features, represents a distinct phylogenetic group [22-24].
To date, approximately 61 SARS-CoV genomic sequences have been analyzed representing different phases of the epidemic (early, middle, and late) and two isolates obtained from palm civets [18]. The SARS-CoV genomic RNA is approximately 30 kb and is organized into 13 to 15 open reading frames (ORFs) [25-27]. The SARS CoV structural gene arrangement follows the same pattern as most coronavirus genomes: 5'- Replicase (ORF 1a)-Protease (ORF 1b)-Spike (S)-Envelope (E)-membrane (M)-Nucleocapsid (N)-3' [27]. However, in contrast to other coronaviruses, two ORFs of unknown function are located between the S and E ORFs and 3–5 ORFs are located between M and N. In addition, despite the evolutionary overlap between SARS-CoV and Group II coronavirus genome sequences, the SARS genome lacks a gene for hemagglutinin-esterase (HE) protein, which is common to a majority of Group II coronaviruses [25]. For an excellent pictorial representation of SARS-CoV genome with functions (or lack of) assigned to each ORF, please refer to the recent review by Tan et al [21]. A significant milestone in SARS-CoV molecular biology was the construction of a SARS-CoV full-length cDNA-containing plasmid from which infectious viral RNA can be produced [28]. This development facilitates the study of SARS-CoV gene functions and should promote the elucidation of function for ORFs whose function is still unknown [29]. Although it has been the perception that these ORFs are not essential for viral replication, they may play a role in the manifestation or severity of disease.
Progress in SARS-CoV genome-based evolutionary biology
RNA viruses utilize a variety of mechanisms to exchange their genetic repertoire. The viral RNA dependent RNA polymerases (RdRP) have a built in error rate that allows diversification of the genomic sequence as replication proceeds. Estimates put the error rate of an RdRp at 10-3 to 10-5 per nucleotide [30]. Coronaviruses also undergo high rates of RNA recombination, providing an additional mechanism by which the viruses can rapidly amplify genomic diversity. The SARS-CoV polymerase gene has a recombination breakpoint, suggesting multiple genetic origins for this molecule. [31]. These evolutionary mechanisms may have facilitated the adaptation of the animal-borne SARS-CoV ancestor to the human host, suggesting that such events in the future could lead to a virus with increased pathogenicity for humans or one capable of infecting multiple species. Recent evidence indicates that the human-adapted SARS virus has crossed into another species. Sequence and epidemiological analyses revealed that a SARS-CoV isolated from a pig was derived from a human strain. Complete nucleotide sequencing of the pig virus isolate (designated TJF) and an S gene-based phylogenetic tree analysis revealed a closer relationship with human SARS-CoV isolates than with animal coronaviruses [32].
Progress in cell biology of SARS-CoV: Signaling pathways
Successful viral replication depends upon the ability of the virus to subvert cellular processes to their advantage and counteract cellular defense mechanisms. Such virus-cell interactions represent potential targets for the development of virus-specific antiviral drugs, therapeutics, and prophylactic vaccines. Different viruses, based on their target cell types and entry pathways, differ in their cellular exploitation mechanisms. The mechanism of SARS virus pathogenesis in vivo may reflect both the effect of viral replication in target cells and host immune responses. The molecular basis for SARS-CoV replication, the signaling pathways affected, and the inflammatory responses provoked by viral infection are not yet clearly understood. Progress in these areas should lead to more effective preventive strategies to counter SARS-CoV infections.
It has been shown that the SARS-CoV N protein selectively activates the Activator Protein-1 (AP-1) signal transduction pathway, which regulates a wide variety of cellular processes including cell proliferation, differentiation, and apoptosis [33]. Such viral induced modifications of the AP-1 pathway may play a significant role in the viral replicative strategy. Recently, another group demonstrated that the S protein alone induces AP-1 activation and that the region from 324–688 amino acids within the S protein is essential for AP-1 activation-dependent IL-8 induction [34]. Another SARS-CoV protein, the U122 ORF of unknown function (also known as X4), was shown to be produced in virus infected Vero E6 cells and expression of this protein alone was shown to induce apoptosis in cell culture [35,36]. This raises the question of how apoptosis of SARS-CoV infected cells is balanced in order for the virus to survive and propagate (Figure 1). This has been addressed to some extent in recent studies which indicate that SARS-CoV infection of Vero E6 cells induces both pro-apoptotic [activation of p38 mitogen-activated protein kinase (MAPK)] and anti-apoptotic [activation of the protein kinase B (PKB, also known as Akt)] signaling pathways, although Akt induction appears to be insufficient to prevent the virus-induced apoptosis [37,38]. Exactly how SARS-CoV manipulates these cellular signaling pathways to facilitate viral replication remains to be determined.
Figure 1 The balance of cell survival and cell death in response to SARS-CoV infection. SARS-CoV is shown approaching a cell with ACE2 receptors (blue "Y"s) on the surface. The virus enters the cell, uncoats, and the viral RNA is replicated and translated. The SARS-CoV U122 protein induces apoptosis in cells. SARS-CoV S and N proteins each can activate the cellular AP-1 protein, which regulates apoptosis, as well as other cellular processes. AP-1 also activates IL-8, a cellular cytokine. SARS-CoV infection induces both MAPK (pro-apoptotic) and Akt (anti-apoptotic) pathways. How this balance between cell survival and apoptosis is maintained is yet unknown. Cellular proteins are labeled in blue, viral proteins in black.
As mentioned above, IL-8 induction was shown to be dependent upon AP-1 activation by SARS-CoV S protein and in this process NF-κB was not involved [34]. This may partially explain the clinical observation of dramatic cytokine storm (high serum levels of IL-6 and IL-8) and inflammation responses observed in SARS patients in the acute stage associated with lung lesions; it has been also suggested that the elevations of IL-6 and IL-8 due to SARS-CoV infection of the respiratory tract can induce the hyper-innate inflammatory response [39]. It is established that cellular MAPKs regulate AP-1 activation-dependent IL-8 induction in viral infections [40-42]. In SARS-CoV infection, the IL-8 induction signaling pathway is perhaps related to angiotensin-converting enzyme 2 (ACE2), as anti-ACE2 antibodies inhibit IL-8 induction/release [34]. ACE2 is the cellular receptor for the SARS-CoV and the receptor-binding sites on the virion are located in the 12–672 amino acid region of the S protein [43].
Current advances towards SARS-CoV prevention strategies
During the SARS outbreak that occurred in 2002–2003, the spread of the disease was primarily controlled by strict quarantine protocols and patient-isolation measures as well as by broad-spectrum antibiotics and antiviral regimens with or without administration of corticosteroids [44,45]. Since then, the wealth of information that has emerged on SARS-CoV molecular and cellular biology, as updated in the preceding sections of this review, now offers potential avenues for developing more efficient anti-viral as well as vaccine strategies.
a. Antiviral agents
Coronavirus genome structure and major gene-product functions have been known for years, but since they cause mild disease, selection of the virus-specific antiviral drugs was not a priority in the past. The SARS-CoV epidemic changed this selective view. Tan et al, 2004, tabulated a screen of available antiviral agents against SARS virus in detail in their recent review [46]. The obvious molecular targets for SARS-CoV antiviral agents are the viral polymerase/replicase, protease, receptor, the viral mRNA cap-1 methyl transferase and NTPase/helicase [47-54]. In addition, a 32-nucleotide long, highly conserved RNA structure in the 3' untranslated region of coronaviruses and astroviruses was identified [55]. This structure resembles the 530 loop of 16s rRNA involved in translation initiation suggesting a possible role for this element in sequestering host translation machinery. The tertiary interactions of this structure create a tunnel lined with negative charge where Mg2+ can bind. This unique structure presents an attractive target for tunnel binding antiviral drugs [55]. Finally, since the functional details of most coronavirus replicase gene products are not known, random screening of potential antiviral compound libraries will be a key area of drug discovery for SARS virus in the near future [47].
b. Vaccine development
Vaccines are the best and least expensive prophylactic measures against pathogens that cause epidemics in humans. The fact that high titers of virus neutralizing antibody to SARS-CoV are found in sera of patients recovering from infection and that those infected with the virus show improvement after passive antibody administration suggests a SARS-CoV vaccine is possible and points toward antibody based treatments for the disease [47,56-58]. However, in developing SARS CoV vaccines, there are lessons to be learned from the world of veterinary CoV vaccines. In a review by Saif, it was pointed out that coronaviruses in general target mucosal surfaces and therefore eliciting local (mucosal) immunity is a major consideration in the development of SARS-CoV vaccines; this largely depends on the type of vaccine, delivery systems, and immuno-modulatory adjuvants used [59]. Further, immunity against animal CoV is usually short term, necessitating periodic boosting, which in the end may not be sufficient to prevent re-infection.
Despite these potential pitfalls in the development of a human vaccine, efforts to develop a vaccine to prevent another SARS outbreak are underway. Several laboratories around the globe are working at an unprecedented pace to develop a SARS vaccine utilizing essentially two different types of SARS-CoV-derived immunogens, 1) inactivated whole virus, and 2) SARS-CoV encoded N and S proteins using recombinant DNA methods. The possibility of producing an engineered live, attenuated SARS-CoV has also been considered.
1. Inactivated whole virus
Takasuka et al (2004) have reported that subcutaneous administration of UV-inactivated purified SARS-CoV virion elicits a high level of humoral immunity, resulting in long-term antibody secretion and memory B cells [60]. The antibodies elicited in mice recognized both the spike (S) and nucleocapsid (N) proteins of the virus. The inactivated virus also induced regional lymph node T-cell proliferation and significant levels of cytokine production upon restimulation with inactivated virus in vitro [60]. These studies suggest that whole-killed virion may have the potential as a candidate antigen for SARS vaccine to elicit both humoral and cellular immunity. When SARS-CoV inactivated by beta-propiolactone was used as antigen in mice and rabbits, the animals elicited antibodies against the receptor-binding domain (RBD) present in the S1 region of SARS-CoV. These antibodies effectively inhibited the S-protein mediated SARS-pseudovirus entry up to 50%, suggesting the potential of the inactivated SARS-CoV as antigen for vaccine development [61]. Depletion of RBD-specific antibodies from patient or rabbit immune sera by immunoadsorption, significantly reduced the virus neutralizing ability of the sera, suggesting that the RBD epitope in the S protein is a critical determinant in developing vaccine strategies [62].
2.1. Cloned N protein
The N protein of SARS-CoV appears to be more conserved than S and M proteins and it has been suggested that this protein may play a role in cell-mediated immunity in SARS-CoV infections and also is an important viral antigen for the early diagnosis. Vaccination of C57BL/6 mice with a SARS-CoV N protein expressed by an E1/partially E3-deleted, replication-defective human adenovirus 5 vector was shown to produce potent SARS-CoV-specific humoral and T cell-mediated immune responses, suggesting the potential of this construct to be used as SARS-CoV vaccine [63]. Along the same line, intra-muscular immunization of BALB/c mice with a plasmid DNA construct encoding the full-length N protein was shown to elicit serum anti-N antibodies and spenocyte proliferative responses against the N protein [64]. The immunized mice also produced strong delayed-type hypersensitivity (DTH) and CD8 (+) CTL responses to the N protein, suggesting that the N protein is not only an important B cell immunogen, but also can elicit broad-based cellular immune responses [64]. In another novel strategy, the N protein was expressed in the cytoplasm of Lactococcus lactis bacterium and the N-expressing bacteria were administered to mice by intranasal or oral route [65]. In this case, significant levels of N-specific IgG in the mice sera were detected, suggesting that the engineered bacteria may serve as a mucosal vaccine against SARS-CoV [65].
2.2. Cloned viral S spike protein or, S-containing pseudovirions
Although immunization with inactivated viral vaccine provides significant protection in animals against challenge with certain corresponding pathogenic CoVs, in the case of SARS-CoV there remains the threat of introducing live virus into the environment from partially inactivated vaccine, as there are no validated and effective inactivation measures developed yet. To circumvent this obstacle, Chen et al have introduced the S protein into the deletion III region of the live, attenuated modified vaccinia virus Ankara (MVA) vector [66]. This recombinant virus elicits potent neutralizing antibodies in mice, rabbits, and monkeys and the major epitope is mapped to the virus receptor-binding region [66]. In another approach, it has been demonstrated that co-expression of SARS-CoV S, M and N expression plasmids in human 293T cells result in the formation of SARS-CoV pseudoparticles (virus-like particles or VLPs) [67]. These findings help us understand the viral morphogenesis as well as offer a safer alternative to using live, replicating SARS virus in the development of vaccines.
3. Attenuated live virus
The third possibility is a genetically engineered version of live SARS-CoV for traits such as attenuated phenotype, increased immunogenicity, and safe handling (out of BL3+ facility). A full-length SARS-CoV cDNA-containing plasmid has been developed from which synthetic infectious viral RNA can be produced [28]. This system allows for the functional analysis of each gene in the context of infection and can be used for making attenuated strains for vaccine development.
Conclusions: Limitations to current SARS vaccine strategies
SARS-CoV clearly has pandemic potential. Although progress in SARS-CoV molecular and cell biology research has been remarkable, there remain clear limitations regarding vaccine development due to a lack of complete understanding in the areas of animal models of the disease as well as host immune responses to the evolving molecular diversity of this newly emerged human virus. Caution is warranted when utilizing experimental data originating from one SARS-CoV strain infection in one animal species or cell line in the development of a human vaccine. The rapid development of an effective SARS-CoV vaccine depends upon continuing basic research.
A study on the evolving S protein molecular diversity in SARS-CoV isolates and its unexpected profound immuno-functional effects illustrates this point [68]. The S protein exhibited minor genetic diversity among 8 strains transmitted during human outbreaks in early 2003. Synthetic versions of these S variants with human preferred codons were tested for 1) their ability to bind the receptor (hACE-2), and 2) their sensitivity to antibody neutralization with viral pseudotypes. In these sets of experiments, substantial functional differences were found in S derived from a Guangdong province case -isolate and two palm civets isolates. Antibodies that neutralized most human isolates-derived S proteins unexpectedly enhanced entry mediated by the civet virus-derived S proteins [68]. This novel observation emphasizes the need to understand the molecular potential of the SARS-CoV genome in developing vaccines to prevent human disease. As mentioned previously, studies also point to the fact that variability in the S protein from early to late disease outbreak stages has been detected [18]. There is a large gap in our understanding of how SARS-CoV interacts with the host cell and the rapidly changing genome of SARS-CoV indicates the potential variability of such interactions [25]. Development of successful vaccines against SARS virus therefore depends on the progress we make in these areas in the immediate future.
Competing Interests
The author(s) declare that they have no competing interests.
Authors' Contributions
Authors contributed equally to the intellectual content of this review article.
Disclaimer
The views presented in this article do not necessarily reflect those of the Food and Drug Administration or United States government.
Acknowledgements
We thank Stephen Feinstone and Ron Lundquist of CBER, FDA for their critiques and the National Vaccines Program Office (NVPO) for a grant to CDA. CJS is supported by a postdoctoral fellowship administered by the Oak Ridge Institute for Science and Education (ORISE).
==== Refs
Zhong NS Zeng GQ Our strategies for fighting severe acute respiratory syndrome (SARS) Am J Respir Crit Care Med 2003 168 7 9 12773318 10.1164/rccm.200305-707OE
Organization WH Cumulative Number of Reported Cases (SARS)
Nie QH Luo XD Zhang JZ Su Q Current status of severe acute respiratory syndrome in China World J Gastroenterol 2003 9 1635 1645 12918094
Donnelly CA Ghani AC Leung GM Hedley AJ Fraser C Riley S Abu-Raddad LJ Ho LM Thach TQ Chau P Chan KP Lam TH Tse LY Tsang T Liu SH Kong JH Lau EM Ferguson NM Anderson RM Epidemiological determinants of spread of causal agent of severe acute respiratory syndrome in Hong Kong Lancet 2003 361 1761 1766 12781533 10.1016/S0140-6736(03)13410-1
Drosten C Gunther S Preiser W van der Werf S Brodt HR Becker S Rabenau H Panning M Kolesnikova L Fouchier RA Berger A Burguiere AM Cinatl J Eickmann M Escriou N Grywna K Kramme S Manuguerra JC Muller S Rickerts V Sturmer M Vieth S Klenk HD Osterhaus AD Schmitz H Doerr HW Identification of a novel coronavirus in patients with severe acute respiratory syndrome N Engl J Med 2003 348 1967 1976 12690091 10.1056/NEJMoa030747
Peiris JS Lai ST Poon LL Guan Y Yam LY Lim W Nicholls J Yee WK Yan WW Cheung MT Cheng VC Chan KH Tsang DN Yung RW Ng TK Yuen KY Coronavirus as a possible cause of severe acute respiratory syndrome Lancet 2003 361 1319 1325 12711465 10.1016/S0140-6736(03)13077-2
Ksiazek TG Erdman D Goldsmith CS Zaki SR Peret T Emery S Tong S Urbani C Comer JA Lim W Rollin PE Dowell SF Ling AE Humphrey CD Shieh WJ Guarner J Paddock CD Rota P Fields B DeRisi J Yang JY Cox N Hughes JM LeDuc JW Bellini WJ Anderson LJ A novel coronavirus associated with severe acute respiratory syndrome N Engl J Med 2003 348 1953 1966 12690092 10.1056/NEJMoa030781
Poutanen SM Low DE Henry B Finkelstein S Rose D Green K Tellier R Draker R Adachi D Ayers M Chan AK Skowronski DM Salit I Simor AE Slutsky AS Doyle PW Krajden M Petric M Brunham RC McGeer AJ Identification of severe acute respiratory syndrome in Canada N Engl J Med 2003 348 1995 2005 12671061 10.1056/NEJMoa030634
Ding Y Wang H Shen H Li Z Geng J Han H Cai J Li X Kang W Weng D Lu Y Wu D He L Yao K The clinical pathology of severe acute respiratory syndrome (SARS): a report from China J Pathol 2003 200 282 289 12845623 10.1002/path.1440
Tsang KW Ho PL Ooi GC Yee WK Wang T Chan-Yeung M Lam WK Seto WH Yam LY Cheung TM Wong PC Lam B Ip MS Chan J Yuen KY Lai KN A cluster of cases of severe acute respiratory syndrome in Hong Kong N Engl J Med 2003 348 1977 1985 12671062 10.1056/NEJMoa030666
Leung WK To KF Chan PK Chan HL Wu AK Lee N Yuen KY Sung JJ Enteric involvement of severe acute respiratory syndrome-associated coronavirus infection Gastroenterology 2003 125 1011 1017 14517783 10.1016/S0016-5085(03)01215-0
Chan-Yeung M Yu WC Outbreak of severe acute respiratory syndrome in Hong Kong Special Administrative Region: case report Bmj 2003 326 850 852 12702616 10.1136/bmj.326.7394.850
Farcas GA Poutanen SM Mazzulli T Willey BM Butany J Asa SL Faure P Akhavan P Low DE Kain KC Fatal severe acute respiratory syndrome is associated with multiorgan involvement by coronavirus J Infect Dis 2005 191 193 197 15609228 10.1086/426870
Seto WH Tsang D Yung RW Ching TY Ng TK Ho M Ho LM Peiris JS Effectiveness of precautions against droplets and contact in prevention of nosocomial transmission of severe acute respiratory syndrome (SARS) Lancet 2003 361 1519 1520 12737864 10.1016/S0140-6736(03)13168-6
Wong TW Lee CK Tam W Lau JT Yu TS Lui SF Chan PK Li Y Bresee JS Sung JJ Parashar UD Cluster of SARS among medical students exposed to single patient, Hong Kong Emerg Infect Dis 2004 10 269 276 15030696
Li L Wo J Shao J Zhu H Wu N Li M Yao H Hu M Dennin RH SARS-coronavirus replicates in mononuclear cells of peripheral blood (PBMCs) from SARS patients J Clin Virol 2003 28 239 244 14522061 10.1016/S1386-6532(03)00195-1
Guan Y Zheng BJ He YQ Liu XL Zhuang ZX Cheung CL Luo SW Li PH Zhang LJ Guan YJ Butt KM Wong KL Chan KW Lim W Shortridge KF Yuen KY Peiris JS Poon LL Isolation and characterization of viruses related to the SARS coronavirus from animals in southern China Science 2003 302 276 278 12958366 10.1126/science.1087139
Chinese SMEC Molecular evolution of the SARS coronavirus during the course of the SARS epidemic in China Science 2004 303 1666 1669 14752165 10.1126/science.1092002
Peiris JS Yuen KY Osterhaus AD Stohr K The severe acute respiratory syndrome N Engl J Med 2003 349 2431 2441 14681510 10.1056/NEJMra032498
Ziebuhr J Molecular biology of severe acute respiratory syndrome coronavirus Curr Opin Microbiol 2004 7 412 419 15358261 10.1016/j.mib.2004.06.007
Tan YJ Lim SG Hong W Characterization of viral proteins encoded by the SARS-coronavirus genome Antiviral Res 2005 65 69 78 15708633 10.1016/j.antiviral.2004.10.001
Gonzalez JM Gomez-Puertas P Cavanagh D Gorbalenya AE Enjuanes L A comparative sequence analysis to revise the current taxonomy of the family Coronaviridae Arch Virol 2003 148 2207 2235 14579179 10.1007/s00705-003-0162-1
Gorbalenya AE Snijder EJ Spaan WJ Severe acute respiratory syndrome coronavirus phylogeny: toward consensus J Virol 2004 78 7863 7866 15254158 10.1128/JVI.78.15.7863-7866.2004
Snijder EJ Bredenbeek PJ Dobbe JC Thiel V Ziebuhr J Poon LL Guan Y Rozanov M Spaan WJ Gorbalenya AE Unique and conserved features of genome and proteome of SARS-coronavirus, an early split-off from the coronavirus group 2 lineage J Mol Biol 2003 331 991 1004 12927536 10.1016/S0022-2836(03)00865-9
Groneberg DA Hilgenfeld R Zabel P Molecular mechanisms of severe acute respiratory syndrome (SARS) Respir Res 2005 6 8 15661082 10.1186/1465-9921-6-8
Rota PA Oberste MS Monroe SS Nix WA Campagnoli R Icenogle JP Penaranda S Bankamp B Maher K Chen MH Tong S Tamin A Lowe L Frace M DeRisi JL Chen Q Wang D Erdman DD Peret TC Burns C Ksiazek TG Rollin PE Sanchez A Liffick S Holloway B Limor J McCaustland K Olsen-Rasmussen M Fouchier R Gunther S Osterhaus AD Drosten C Pallansch MA Anderson LJ Bellini WJ Characterization of a novel coronavirus associated with severe acute respiratory syndrome Science 2003 300 1394 1399 12730500 10.1126/science.1085952
Thiel V Ivanov KA Putics A Hertzig T Schelle B Bayer S Weissbrich B Snijder EJ Rabenau H Doerr HW Gorbalenya AE Ziebuhr J Mechanisms and enzymes involved in SARS coronavirus genome expression J Gen Virol 2003 84 2305 2315 12917450 10.1099/vir.0.19424-0
Yount B Curtis KM Fritz EA Hensley LE Jahrling PB Prentice E Denison MR Geisbert TW Baric RS Reverse genetics with a full-length infectious cDNA of severe acute respiratory syndrome coronavirus Proc Natl Acad Sci U S A 2003 100 12995 13000 14569023 10.1073/pnas.1735582100
Baric RS Sims AC Development of mouse hepatitis virus and SARS-CoV infectious cDNA constructs Curr Top Microbiol Immunol 2005 287 229 252 15609514
Domingo E Holland JJ RNA virus mutations and fitness for survival Annu Rev Microbiol 1997 51 151 178 9343347 10.1146/annurev.micro.51.1.151
Rest JS Mindell DP SARS associated coronavirus has a recombinant polymerase and coronaviruses have a history of host-shifting Infect Genet Evol 2003 3 219 225 14522185 10.1016/j.meegid.2003.08.001
Chen W SARS-associated Coronavirus Transmitted from Human to Pig Emerg Infect Dis 2005 11 446 448 15757562
He R Leeson A Andonov A Li Y Bastien N Cao J Osiowy C Dobie F Cutts T Ballantine M Li X Activation of AP-1 signal transduction pathway by SARS coronavirus nucleocapsid protein Biochem Biophys Res Commun 2003 311 870 876 14623261 10.1016/j.bbrc.2003.10.075
Chang YJ Liu CY Chiang BL Chao YC Chen CC Induction of IL-8 release in lung cells via activator protein-1 by recombinant baculovirus displaying severe acute respiratory syndrome-coronavirus spike proteins: identification of two functional regions J Immunol 2004 173 7602 7614 15585888
Fielding BC Tan YJ Shuo S Tan TH Ooi EE Lim SG Hong W Goh PY Characterization of a unique group-specific protein (U122) of the severe acute respiratory syndrome coronavirus J Virol 2004 78 7311 7318 15220404 10.1128/JVI.78.14.7311-7318.2004
Tan YJ Fielding BC Goh PY Shen S Tan TH Lim SG Hong W Overexpression of 7a, a protein specifically encoded by the severe acute respiratory syndrome coronavirus, induces apoptosis via a caspase-dependent pathway J Virol 2004 78 14043 14047 15564512 10.1128/JVI.78.24.14043-14047.2004
Mizutani T Fukushi S Saijo M Kurane I Morikawa S Phosphorylation of p38 MAPK and its downstream targets in SARS coronavirus-infected cells Biochem Biophys Res Commun 2004 319 1228 1234 15194498 10.1016/j.bbrc.2004.05.107
Mizutani T Fukushi S Saijo M Kurane I Morikawa S Importance of Akt signaling pathway for apoptosis in SARS-CoV-infected Vero E6 cells Virology 2004 327 169 174 15351204 10.1016/j.virol.2004.07.005
Hsueh PR Chen PJ Hsiao CH Yeh SH Cheng WC Wang JL Chiang BL Chang SC Chang FY Wong WW Kao CL Yang PC Patient data, early SARS epidemic, Taiwan Emerg Infect Dis 2004 10 489 493 15109419
Chen W Monick MM Carter AB Hunninghake GW Activation of ERK2 by respiratory syncytial virus in A549 cells is linked to the production of interleukin 8 Exp Lung Res 2000 26 13 26 10660833 10.1080/019021400269934
Alcorn MJ Booth JL Coggeshall KM Metcalf JP Adenovirus type 7 induces interleukin-8 production via activation of extracellular regulated kinase 1/2 J Virol 2001 75 6450 6459 11413312 10.1128/JVI.75.14.6450-6459.2001
Griego SD Weston CB Adams JL Tal-Singer R Dillon SB Role of p38 mitogen-activated protein kinase in rhinovirus-induced cytokine production by bronchial epithelial cells J Immunol 2000 165 5211 5220 11046054
Li W Moore MJ Vasilieva N Sui J Wong SK Berne MA Somasundaran M Sullivan JL Luzuriaga K Greenough TC Choe H Farzan M Angiotensin-converting enzyme 2 is a functional receptor for the SARS coronavirus Nature 2003 426 450 454 14647384 10.1038/nature02145
So LK Lau AC Yam LY Cheung TM Poon E Yung RW Yuen KY Development of a standard treatment protocol for severe acute respiratory syndrome Lancet 2003 361 1615 1617 12747883 10.1016/S0140-6736(03)13265-5
Tsang K Zhong NS SARS: pharmacotherapy Respirology 2003 8 Suppl S25 30 15018130 10.1046/j.1440-1843.2003.00525.x
Tan EL Ooi EE Lin CY Tan HC Ling AE Lim B Stanton LW Inhibition of SARS coronavirus infection in vitro with clinically approved antiviral drugs Emerg Infect Dis 2004 10 581 586 15200845
Oxford JS Balasingam S Chan C Catchpole A Lambkin R New antiviral drugs, vaccines and classic public health interventions against SARS coronavirus Antivir Chem Chemother 2005 16 13 21 15739618
Tanner JA Watt RM Chai YB Lu LY Lin MC Peiris JS Poon LL Kung HF Huang JD The severe acute respiratory syndrome (SARS) coronavirus NTPase/helicase belongs to a distinct class of 5' to 3' viral helicases J Biol Chem 2003 278 39578 39582 12917423 10.1074/jbc.C300328200
von Grotthuss M Wyrwicz LS Rychlewski L mRNA cap-1 methyltransferase in the SARS genome Cell 2003 113 701 702 12809601 10.1016/S0092-8674(03)00424-0
Yang H Yang M Ding Y Liu Y Lou Z Zhou Z Sun L Mo L Ye S Pang H Gao GF Anand K Bartlam M Hilgenfeld R Rao Z The crystal structures of severe acute respiratory syndrome virus main protease and its complex with an inhibitor Proc Natl Acad Sci U S A 2003 100 13190 13195 14585926 10.1073/pnas.1835675100
Xu X Liu Y Weiss S Arnold E Sarafianos SG Ding J Molecular model of SARS coronavirus polymerase: implications for biochemical functions and drug design Nucleic Acids Res 2003 31 7117 7130 14654687 10.1093/nar/gkg916
Kontoyiannis DP Pasqualini R Arap W Aminopeptidase N inhibitors and SARS Lancet 2003 361 1558 12737884 10.1016/S0140-6736(03)13186-8
Chou KC Wei DQ Zhong WZ Binding mechanism of coronavirus main proteinase with ligands and its implication to drug design against SARS Biochem Biophys Res Commun 2003 308 148 151 12890493 10.1016/S0006-291X(03)01342-1
Anand K Ziebuhr J Wadhwani P Mesters JR Hilgenfeld R Coronavirus main proteinase (3CLpro) structure: basis for design of anti-SARS drugs Science 2003 300 1763 1767 12746549 10.1126/science.1085658
Robertson MP Igel H Baertsch R Haussler D Ares MJ Scott WG The structure of a rigorously conserved RNA element within the SARS virus genome PLoS Biol 2005 3 e5 15630477 10.1371/journal.pbio.0030005
Sui J Li W Murakami A Tamin A Matthews LJ Wong SK Moore MJ Tallarico AS Olurinde M Choe H Anderson LJ Bellini WJ Farzan M Marasco WA Potent neutralization of severe acute respiratory syndrome (SARS) coronavirus by a human mAb to S1 protein that blocks receptor association Proc Natl Acad Sci U S A 2004 101 2536 2541 14983044 10.1073/pnas.0307140101
Li G Chen X Xu A Profile of specific antibodies to the SARS-associated coronavirus N Engl J Med 2003 349 508 509 12890855 10.1056/NEJM200307313490520
Pearson H Clarke T Abbott A Knight J Cyranoski D SARS: what have we learned? Nature 2003 424 121 126 12853923 10.1038/424990a
Saif LJ Animal coronavirus vaccines: lessons for SARS Dev Biol (Basel) 2004 119 129 140 15742624
Takasuka N Fujii H Takahashi Y Kasai M Morikawa S Itamura S Ishii K Sakaguchi M Ohnishi K Ohshima M Hashimoto S Odagiri T Tashiro M Yoshikura H Takemori T Tsunetsugu-Yokota Y A subcutaneously injected UV-inactivated SARS coronavirus vaccine elicits systemic humoral immunity in mice Int Immunol 2004 16 1423 1430 15314040 10.1093/intimm/dxh143
He Y Zhou Y Siddiqui P Jiang S Inactivated SARS-CoV vaccine elicits high titers of spike protein-specific antibodies that block receptor binding and virus entry Biochem Biophys Res Commun 2004 325 445 452 15530413 10.1016/j.bbrc.2004.10.052
He Y Zhu Q Liu S Zhou Y Yang B Li J Jiang S Identification of a critical neutralization determinant of severe acute respiratory syndrome (SARS)-associated coronavirus: importance for designing SARS vaccines Virology 2005 334 74 82 15749124 10.1016/j.virol.2005.01.034
Zakhartchouk AN Viswanathan S Mahony JB Gauldie J Babiuk LA Severe acute respiratory syndrome coronavirus nucleocapsid protein expressed by an adenovirus vector is phosphorylated and immunogenic in mice J Gen Virol 2005 86 211 215 15604448 10.1099/vir.0.80530-0
Zhao P Cao J Zhao LJ Qin ZL Ke JS Pan W Ren H Yu JG Qi ZT Immune responses against SARS-coronavirus nucleocapsid protein induced by DNA vaccine Virology 2005 331 128 135 15582659 10.1016/j.virol.2004.10.016
Pei H Liu J Cheng Y Sun C Wang C Lu Y Ding J Zhou J Xiang H Expression of SARS-coronavirus nucleocapsid protein in Escherichia coli and Lactococcus lactis for serodiagnosis and mucosal vaccination Appl Microbiol Biotechnol 2005
Chen Z Zhang L Qin C Ba L Yi CE Zhang F Wei Q He T Yu W Yu J Gao H Tu X Gettie A Farzan M Yuen KY Ho DD Recombinant modified vaccinia virus ankara expressing the spike glycoprotein of severe acute respiratory syndrome coronavirus induces protective neutralizing antibodies primarily targeting the receptor binding region J Virol 2005 79 2678 2688 15708987 10.1128/JVI.79.5.2678-2688.2005
Huang Y Yang ZY Kong WP Nabel GJ Generation of synthetic severe acute respiratory syndrome coronavirus pseudoparticles: implications for assembly and vaccine production J Virol 2004 78 12557 12565 15507643 10.1128/JVI.78.22.12557-12565.2004
Yang ZY Werner HC Kong WP Leung K Traggiai E Lanzavecchia A Nabel GJ Evasion of antibody neutralization in emerging severe acute respiratory syndrome coronaviruses Proc Natl Acad Sci U S A 2005 102 797 801 15642942 10.1073/pnas.0409065102
| 15833113 | PMC1087510 | CC BY | 2021-01-04 16:38:58 | no | Virol J. 2005 Apr 15; 2:35 | utf-8 | Virol J | 2,005 | 10.1186/1743-422X-2-35 | oa_comm |
==== Front
World J Surg OncolWorld Journal of Surgical Oncology1477-7819BioMed Central London 1477-7819-3-181580198010.1186/1477-7819-3-18ResearchIncrease in number of circulating disseminated epithelial cells after surgery for non-small cell lung cancer monitored by MAINTRAC® is a predictor for relapse: A preliminary report Rolle Axel [email protected]ünzel Rainer [email protected] Ulrich [email protected] Babette [email protected]öffken Klaus [email protected] Katharina [email protected] Fachkrankenhaus Coswig/Dresden, der Friedrich-Schiller-Universität Jena, Germany2 TZB Transfusionsmedizinisches Zentrum Bayreuth, der Friedrich-Schiller-Universität Jena, Germany3 Klinik für Innere Medizin II der Friedrich-Schiller-Universität Jena, Germany2005 31 3 2005 3 18 18 29 11 2004 31 3 2005 Copyright © 2005 Rolle et al; licensee BioMed Central Ltd.2005Rolle 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
Lung cancer still remains one of the most commonly occurring solid tumors and even in stage Ia, surgery fails in 30% of patients who develop distant metastases. It is hypothesized that these must have developed from occult circulating tumor cells present at the time of surgery, or before. The aim of the present study was to detect such cells in the peripheral blood and to monitor these cells following surgery.
Methods
30 patients treated for lung cancer with surgery were monitored for circulating epithelial cells (CEC) by taking peripheral blood samples before, 2 weeks and 5 months after surgery and/or radiotherapy (RT) chemotherapy (CT) or combined RT/CT using magnetic bead enrichment and laser scanning cytometry (MAINTRAC®) for quantification of these cells.
Results
In 86% of the patients CEC were detected before surgery and in 100% at 2 weeks and 5 months after surgery. In the control group, which consisted of 100 normal donors without cancer, 97 % were negative for CEC. A significantly higher number of CEC was found preoperatively in patients with squamous cell carcinoma than in those with adenocarcinoma. In correlation to the extent of parenchymal manipulation 2 weeks after surgery, an increase in numbers of CEC was observed with limited resections (18/21) whereas pneumonectomy led to a decrease (5/8) of CEC, 2 weeks after surgery. The third analysis done 5 months after surgery identified 3 groups of patients. In the group of 5 patients who received neo- or adjuvant chemo/radiotherapy there was evidence that monitoring of CEC can evaluate the effects of therapy. Another group of 7 patients who underwent surgery only showed a decrease of CEC and no signs of relapse. A third group of 11 patients who had surgery only, showed an increase of CEC (4 with an initial decrease after surgery and 7 with continuous increase). In the group with a continuous increase during the following 24 months, 2 early relapses in patients with stage Ia adenocarcinoma were observed. The increase of CEC preceded clinical detection by six months.
Conclusion
We consider, therefore, that patients with adenocarcinoma and a continuous increase of CEC after complete resection for lung cancer are at an increased risk of early relapse.
==== Body
Background
Lung cancer still remains the most frequent solid tumour in men and represents the leading cause of death in women since the 1980s in the USA [1]. Due to a steady increase in cigarette smoking, even in Europe and Germany, lung cancer has become three times more common in women during the last 20 years [2].
Despite good progress in diagnosis, operative and multimodality treatment of patients with lung cancer the prognosis has remained persistently bad during the last 50 years, with 5 years survival rates of 9% for all patients [3]. The therapeutic dilemma becomes evident if results from stage Ia lung cancers are examined. If patients undergo a complete resection, the therapy of choice, then the 5 year survival is 70%. But treatment has failed for the 30% of patients who develop distant metastases in the two years following surgery [4]. The metastases in these patients must have developed from occult circulating tumor cells present at the time of their original surgery, or before, leading to an underestimation of the true tumor stage. Pantel et al [5] detected epithelial cells in 20–25 % of biopsies taken intra-operatively from bone marrow after resection of lung cancers. Nevertheless, even today the role of cytokeratin positive cells in bone marrow remains unclear. They may have come from the tumor at an early stage, or seeded into the circulation during surgery. lt is also not clear whether these cells have proliferative activity or are so-called dormant cells [6].
Using a method for rapid determination of circulating epithelial cells (CEC) in peripheral blood [7], which allows monitoring of these cells under therapy, we demonstrate that a single examination is not sufficient to determine outcome. However, it is necessary to monitor the behavior of these cells over time to be able to predict tumor recurrence.
Methods and methods
Anti-coagulated peripheral blood samples were drawn after informed consent from a control group of 100 normal subjects aged between 17 and 73 years and from 30 successive patients before, and two weeks after, surgery for lung cancer. A third sample was obtained 3 to 5 months postoperatively when the patients were seen for follow-up. In all patients routine controls were continued up to 27 months or death with a median follow-up of 22 months.
Seven patients were excluded. In one patient lung cancer was not confirmed after histological examination and in another patient lung cancer was combined with advanced sarcoidosis. In 5 patients one or two samples were missing therefore these cases were omitted as well, despite there being no further discrepancy. Of the remaining 23 patients, for which all data and follow-up were complete, 20 were male (mean age 64 years, range 55 – 74 years) and 3 were female (mean age 53 years, range 40 – 74 years) with a balanced proportion of early to advanced stages of lung cancer (12:11). Staging was performed by chest x-ray, endobrochial bronchoscopy, computerized tomography (CT), bone scan, operative lymph node dissection and size of tumor-free margin. Further information about surgery, histological type of tumor stage and the number of CEC is depicted in table 1.
Table 1 Data from 23 patients operated for lung cancer
CEC Levels
No age sex tumor histology operation Pathologic stage 1st analysis pre op 2nd analysis 2 weeks pop 3nd analysis 5 month pop Change therapy follow up
1 55 M squamous pneumonectomy III a 769.091 664.210 853.333 ⇒ neoadj. CT Ø R 24 months
2 62 M squamous pneumonectomy II b 0 474.375 5.722.667 Ø R 27 months
3 68 M squamous bilobectomy I b 5.396.667 2.222 65.000 Ø R 25 months
4 68 M squamous pneumonectomy III a 1.514.583 80.000 35.185 Ø R 25 months
5 67 M squamous pneumonectomy I b 292.592 292.592 1.312.000 Ø R 26 months
6 57 M squamous lobectomy I b 64.942 476.667 610.000 Ø R 26 months
7 68 M adeno pneumonectomy III a 1.000.000 110.000 96.429 ⇒ adj. RT T † 26 months
8 66 M squamous pneumonectomy I b 0 5.000 203.333 Ø R 25 months
9 56 M adeno lobectomy I a 224.118 615.556 1.656.000 R † 24 months
10 74 M squamous explorative III b 127.059 147.000 70.000 adj.CT T alive 25 months
11 60 M adeno lobectomy I b 3.158 203.333 13.889 Ø R 22 months
12 61 M adeno lobectomy I b 28.500 6.667 46.667 Ø R 25 months
13 41 F adeno lobectomy I a 20.625 51.667 98.333 R † 11 months
14 67 M squamous pneumonectomy II b 16.667 9.259 93.333 Ø R 14 months
15 69 M adeno segmentectomy I b 6.667 25.000 100.000 Ø R 24 months
16 68 M adeno explorativ III b 71.667 140.000 263.571 adj. CT/RT T † 9 months
17 65 M adeno lobectomy I b 0 335.000 182.857 Ø R 13 months
18 71 M adeno lobectomy I b 22.941 17.778 1.605.185 Ø R 23 months
19 68 M adeno lobectomy III a 6.667 57.647 11.111 adj. RT Ø R 23 months
20 66 M adeno lobectomy I a 126.667 154.286 38.571 Ø R 22 months
21 57 M squamous pneumonectomy I b 473.333 1.031.250 503.333 Ø R 23 months
22 74 F squamous lobectomy II b 11.053 160.000 150.000 ⇒ Ø R 19 months
23 56 F adeno segmentectomy I b 5.000 2.670.370 576.190 Ø R 23 months
CT = Chemotherapy, RT = Radiotherapy, R = Recurrence, ∅ R = no Recurrence, T = Tumor, T † = died with tumor, adj. = adjuvant, neoadj. = neoadjuvant, Pop = Postoperatively, Adeno = Adenocarcinoma, Squamous = Squamous carcinoma
Cell preparation
Red blood cells from 10 ml samples of blood were lysed, the white blood cells spun down and epithelial cells were enriched using micro-beads from Miltenyi (Bergisch Gladbach, Germany) [7] or the method by Labsoft Diagnostics (Halle, Germany) [8]. Both methods yielded comparable results. The same method was always used in follow-up samples from the same patient. Cells were stained with FITC conjugated mouse anti-human epithelial antibody (HEA, Milteny, Bergisch Gladbach, Germany), reacted either with HEA-micro-beads Miltenyi (Bergisch Gladbach, Germany) or Cell Select-micro-beads (Labsoft Diagnostics Halle, Germany) and then either separated over the columns provided by the Miltenyi company according to their description or the vials containing the cells were applied to magnets provided by the Labsoft company. Negative cells not bound in the columns were washed out, the columns removed from the magnet and the retained cells eluted. In the Labsoft method cells not attached to the vial wall were sucked off, the vials removed from the magnet and the cells retained in the vials were then flushed off the wall and studied.
Analysis
For measurements the cells were applied to adhesion slides (Menzel, Gläser, Germany) and measurements started when the cells had settled using a Laser Scanning Cytometer (LSC® Compucyte Corporation, Cambridge, MA, USA). The cells could easily and unequivocally be contoured using forward scatter as a threshold parameter. Background fluorescence was determined dynamically to calculate both peak and integral fluorescence on a per-cell basis making the fluorescent calculation equivalent for all cells. The green fluorescence was displayed as scattergrams and histograms.
The LSC® enables the user to relocate cells contained within the positive population for visual examination through the microscope (Figure 1). In addition a CCD camera attached to the microscope allowed taking photo- and fluoromicrographs at the same time. More detailed morphology could be determined in conventional panoptic staining (Figure 1).
Figure 1 Example of an analysis of a slide with epithelial antigen positive and negative cells: a) fluorescence intensity displayed as a dot plot with all cells in region 1 and the positive cells gated in region 2 (green region). b) The scanning stage proceeds from one positive event to the next stopping at each position of a positive event and allowing the observer to control whether it represents a specific membrane stained cell as shown in figure 1c. c) Three typical cells shown with green fluorescence. In the lower panel the identical cell is shown in panoptic staining indicated by the arrow.
Results
Circulating epithelial cells (CEC) were detectable in blood samples from 25/29 patients (86%) before surgery, whereas 97% of normal controls were negative for these cells. The number of CEC was calculated in relation to the analyzed blood volume and not to white blood cells, which vary considerably during the course of disease. For 26 patients histological evaluation was available and revealed 15 cases of adenocarcinoma and 11 cases of squamous cell carcinoma. The comparison of preoperative numbers of CEC revealed higher values in patients with squamous cell than in those with adenocarcinoma and these values differed significantly (p = 0.046) (Figure 2). Neither tumor size nor positivity of lymph nodes correlated with the number of circulating tumor cells before surgery. With regard to the extent of resection 1 pleurectomy, 8 pneumonectomies, 1 bilobectomy, 12 lobectomies, 4 segmental resections and 3 explorative thoracotomies were performed. Lobectomy was more frequent than pneumonectomy in patients with adenocarcinoma (69% vs. 36%) and pneumonectomy more frequent than lobectomy in those with squamous cell carcinoma (54% vs. 6%) (Figure 3).
Figure 2 Number of preoperative circulating tumor cells in comparison to histological classification.
Figure 3 Frequency of surgical interventions in comparison to histological classification.
Fourteen days postoperatively, CEC were detected in all patients including the 3 where no cells had been detectable before surgery. In cases with limited resection (3 exploratory thoracotomies, 4 segmental resections, 12 lobectomies, 1 bilobectomy) parenchymal manipulation was more extensive. This resulted more frequently in an increase (85.7%) than in a decrease (14.3%) in CEC, whereas in cases of complete resection of the lung (pneumonectomy) numbers of CEC decreased (62.5%) more frequently than increased (37.5%) (Figure 4). We, also, investigated CEC in one patient who had surgery for pneumothorax and observed a comparable increase in CEC as in the tumor patients but with a rapid re-decrease already after another two weeks (Figure 5a).
Figure 4 Changes in cell numbers (increase or decrease) following surgical intervention.
Figure 5 Trends of CEC numbers in individual patients after complete resection without further treatment: a) patients with an increase in CEC 14 days after surgery and a subsequent decrease during the next five months (dotted lines: squamous cell cancer; solid lines: adenocarcinoma; fat line and squares: patient with benign lung disease pneumothorax) b) patients with an increase in CEC 14 days after surgery and a subsequent further increase during the next five months (dotted lines: squamous cell cancer; solid lines: adenocarcinoma; fat lines: relapsed patients) c) patients with a decrease in CEC 14 days after surgery (dotted lines: squamous cell cancer; solid lines: adenocarcinoma)
A third analysis of CEC was performed between 3 and 5 months after surgery. In 23 (76%) of 30 consecutive patients considered to have a resectable non small cell lung cancer (NSCLC) complete data were available with a median follow-up of 22 months (range 9 – 27 months) (Table 1).
Five out of 23 patients underwent chemotherapy (CT) or chemo/radiotherapy (CT/RT). These patients were considered separately from the 18 patients who had complete curative resection of their tumors without adjuvant treatment. Two patients underwent exploratory thoracotomy: one with squamous cell carcinoma (patient 10) was treated with adjuvant CT with a subsequent decrease in CEC and is still alive 25 months after surgery. The other patient with adenocarcinoma (patient 16) showed an increase in CEC in spite of combined RT/CT and died 9 months later with tumor progression.
Three of the 21 patients with tumor resection received additional treatment: 1 lobectomy patient with adenocarcinoma had N2 lymph node status (patient 19). This patient with one level N2 status received adjuvant RT. He subsequently showed a decrease in CEC and is alive, 23 months after surgery without signs of relapse. One patient with squamous cell carcinoma (patient 1) received neoadjuvant therapy and subsequently was treated with pneumonectomy. He showed a slight increase in CEC but is still alive without signs of relapse. Another patient with adenocarcinoma (patient 7) underwent RT but with no significant decrease in CEC and died 26 months later with tumor progression. Among these patients receiving additional radio- and/or chemotherapy relapses occurred only in those with adenocarcinoma and increasing CEC numbers in spite of treatment.
Of the remaining 18 patients who had complete curative resection of their tumors, 7 showed a decrease, and 11 showed an increase in CEC during the 3–5 months follow-up after surgery.
Of the former 7 patients 6 had had cell numbers increasing after surgery and subsequently decreasing during the following months (patients 11, 17, 20, 21, 22, 23), 4 with adeno- and 2 with squamous cell carcinoma and all but one (patient 21) had lobectomy (Figure 5a). All are in complete continuous remission. 1 patient (patient 4) with squamous cell carcinoma had pneumectomy and continuously decreasing cell numbers (Figure 5c) and he, too, is in complete continuous remission.
Among the 11 patients with increasing CEC there were 4 with an initial decrease after surgery followed by a subsequent increase (Figure 5c). None has shown early relapse so far. Seven had a continuous increase from pre-surgery to 5 months after treatment (Figure 5b), 4 with squamous cell carcinoma and 3 with adenocarcinoma. Two relapses were observed amongst these patients during the subsequent 2 years of follow-up. Both, a male (56 years of age) and a female (41 years of age), had adenocarcinoma stage Ia. They had continuously increasing CEC and the increase occurred six months before repeated clinical examinations could verify disease relapse. Both received chemotherapy after diagnosis of relapse but died despite 9 and 24 months after surgery with systemic spread of their tumor.
In summary, amongst the patients with adjuvant treatment after resection, as well as in curatively resected patients, patients with adenocarcinoma and a continuous increase in CEC during the course of their disease have a significantly higher risk of early relapse than those with decreasing cell numbers and than those with squamous cell carcinoma, even if the latter show increasing numbers of CEC.
In the relapsed patients not single values of circulating tumor cells at different time points but only the increase in cell numbers was a predictor of relapse.
Discussion
Analyzing circulating epithelial cells in peripheral blood from patients with non small cell lung cancer, we were able to detect live CEC in 86% of patients before surgery. This represents a significantly higher number and frequency than reported from other groups analyzing micrometastases taken from bone marrow [9,10] but cell numbers are in the same range as those reported by Cristofanilli and coworkers [11]. This discrepancy may be explained by several modifications of pre-analytical handling: 1) In contrast to the former authors [9,10] we used red blood cell lysis [7] for white blood cell preparation. During ficoll-separation used for preparation of cells from bone marrow aspirates we observed a loss of up to 90% epithelial antigen positive cells and there is no reason why circulating epithelial cells should have the same density properties as mononuclear cells (Pachmann submitted for publication]. 2) Micrometastatic cells can be hidden in the fat lumps of the bone marrow aspirates and thus be lost for analysis when preparing mononuclear cell suspensions. 3) The portion of blood admixed to marrow aspirates can vary between 10 and 90% resulting in a high variability of the material used for analysis. Although the method applied by us is also suitable for bone marrow samples, we preferred analysis of peripheral blood samples which are easy to obtain and can be drawn repeatedly.
The number of CEC detected by us in lung cancer patients is somewhat lower than the number reported recently by Cristofanilli et al [11] in metastatic breast cancer patients. In positive patients we found between 2000 and 6 × 106 epithelial cells calculating the number of cells in the circulation assuming that a patient has about 5 liters of blood. The values of 5 to 22000 cells per 7.5 ml blood reported by Cristofanilli et al [11] would translate into 3000 to 13 × 106 cells in the circulation. The frequency of preoperatively CEC positive patients was comparable to that reported by Racila et al [12] and Krag et al [13] who also used a similar method for analyzing blood from breast cancer patients.
We preferred detection of fluorescent antibody bound to the surface of live epithelial antigen positive cells performed by an automated Laser Scanning Cytometer LSC® to PCR based methods. PCR methods are currently regarded as the most sensitive procedures, but there is a reported high frequency of false positives [14-16]. In addition, our procedure allowed subsequent separation and even culturing of the isolated cells. Culturing will definitely allow determination whether the cells are dormant or have the potency of differentiating into tumor cells.
It should be mentioned that we observed significantly higher numbers of CEC preoperatively in patients with squamous cell carcinoma than in those with adenocarcinoma. This could be due to the former being more centrally located (next to the great vessels) which might lead to a higher dissemination than occurs in more peripherally located tumors.
Two weeks after surgery all patients were positive for CEC, but there was no correlation to stage, tumor size or lymph node involvement. Unexpectedly we found a strong correlation between postoperative increase and limited resections (segmental resections and lobectomies) and a decrease of CEC after pneumonectomies. We can only explain this observation by the extended parenchymal manipulation involving the separation of parenchymal bridges between lobes and late ligation of pulmonary vessels during lobectomies or segmentectomies. In contrast, pneumonectomies normally start with ligation of the vessels and subsequent resection of the lung. Perhaps our data confirm once more the importance of careful "No touch" preparation during oncological resections [17].
However, during parenchymal resections also, normal epithelial cells from the lung can be mobilized and washed out entering the vessels. This may be true in the cases where we observed a reduction of CEC during the following six months of observation. Our attempts to differentiate benign isolated epithelial cells from malignant cells by their morphology failed, and molecular biological techniques will need to be applied [18,19] to solve this problem. But generally the premise is accepted, that disseminated normal cells will die in a short time and only the portion of cells that are malignant will have a chance of survival for unknown periods of time. [18]. Monitoring the CEC after five months would, therefore, be expected to detect only surviving malignant epithelial cells. Once again we found CEC in all patients.
Analysis of the increase or reduction of CEC in the 5 patients receiving neoadjuvant and/or adjuvant CT and RT due to their nodal status indicated that this parameter can possibly become a tool for monitoring the effectiveness of therapeutic interventions, an effect that we are actually investigating now in patients receiving CT for lung cancer.
Of the 18 patients treated with complete curative resection all 7 patients of the group with decreasing numbers of CEC were without signs of relapse up to 27 months after surgery. In the group of 11 patients with increasing numbers of CEC, 6 had squamous cell carcinomas and none has relapsed during the observation time of 2 years in accordance with published data [20]. In contrast, in the group of adenocarcinomas 2/5 early relapses were observed, both in patients with stage Ia with the increase in CEC preceded overt clinical relapse by 6 months.
Conclusion
Thus, postoperative monitoring of CEC may become a reliable prognostic predictor of relapse. The observation that only 20% of the patients in the group at risk develop early relapse confirmed by other studies [21] and that patients with squamous cell carcinomas can have high and over time even increasing numbers of CEC necessitate additional analyses on the CEC to obtain more information about their metastatic potential [22]. Such analyses on enriched and cultured isolated CEC are now under way in our laboratory.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
AR: carried out the studies, and drafted the manuscript
RG: conceived of the study, and participated in its design and coordination
UP: participated in the design of the study and performed the statistical analysis
BW: carried out the immunoassays
KH: preparation of the manuscript
KP: participated in the design of the study and performed the statistical analysis
All authors read and approved the manuscript
Research grant
Not declared
Acknowledgements
None
==== Refs
Wingo PA Cardinez CJ Landis SH Greenlee RT Ries LA Anderson RN Thun MJ Long term trends in cancer mortality in the United States 1996–1998 Cancer 2003 97 3133 3275 12784323 10.1002/cncr.11380
Bray F Tykczynski JE Parkin DM Going up or coming down? The changing phases of the lung cancer epidemic from 1966–1999 in the 15 European Union countries Eur J Cancer 2004 40 96 125 14687795 10.1016/j.ejca.2003.08.005
Xie L Ugnat AM Morriss J Scmenciw R Mao Y Histology related variation in the treatment and survival of patients with lung carcinoma in Canada Lung Cancer 2003 42 127 139 14568680 10.1016/S0169-5002(03)00283-6
Takita H Pitoniak RF Induction therapy for locoregional lung cancer using paclitaxel combination. A preliminary report J Exp Clin Cancer Res 2000 19 291 293 11144521
Pantel K von Knebel Doeberitz M Izbicki JR Riethmüller G Disseminierte Tumourzellen: Diagnostik, prognostische Relevanz, Phänotypisierung und therapeutische Strategien Chirurg 1997 68 1241 1250 9483347 10.1007/s001040050351
Bessa X Elizalde JI Boix L Pinol V Lacy AM Salo J Pique JM Castells A Lack of prognostic influence of circulating tumor cells in peripheral blood of patients with colorectal cancer Gastroenterology 2001 120 1084 1092 11266372
Pachmann K Heiß P Demel U Tilz G Detection and quantification of small numbers of circulating tumor cells in peripheral blood using Laser Scanning Cytometry (LSC®) Clin Chem Lab Med 2001 39 811 817 11601678 10.1515/CCLM.2001.134
Pachmann K Lychatz S Oblonczek G Willen B Tumourzell-Analyse mit kombinierter Magnetanreicherung, Laser-Scanning Zytometrie und visueller Kontrolle Laborwelt 2001 Nr. N 23 26
Schlimok G Funke I Pantel K Strobel F Lindemann F Witte J Riethmüller G Micrometastatic tumour cells in bone marrow' of patients with gastric cancer: methodological aspects of detection and prognostic significance Eur J Cancer 1991 27 1461 1465 1720636
Cote RJ Beattie EJ Chaiwun B Sevi SR Harry J Chen SC Sherrod AF Groshen S Taylor CR Detection of occult bone marrow micrometastases in patients with operable lung carcinoma Ann Surg 1995 222 423 425
Cristofanilli M Budd GT Ellis MJ Stopeck A Matera J Miller MC Reuben JM Doyle GV Allard WJ Terstappen LW Hayes DF Circulating tumor cells, disease progression, and survival in metastatic breast cancer New Engl J Med 2004 351 781 791 15317891 10.1056/NEJMoa040766
Racila E Euhus D Weiss AJ Rao C Mc Connell J Terstappen LW Uhr JW Detection and characterization of carcinoma cells in the blood Proc Natl Acad Sci USA 1998 95 4589 4594 9539782 10.1073/pnas.95.8.4589
Krag DN Ashikaga T Moss TS Kusminsky RE Feldman S Carp NZ Moffat FL JrBeitsch PD Frazier TG Gaskin TA Shook JW Harlow SP Weaver DL Breast cancer cells in the blood; a pilot study The Breast J 1999 5 354 358 11348313 10.1046/j.1524-4741.1999.99016.x
Jung R Petersen K Krugar W Wolf M Wagener C Zander A Neumaier M Detection of micrometastasis by cytokeratin 20 RT-PCR is limited due to stable background transcription in granulocytes Br J Cancer 1999 81 870 873 10555760 10.1038/sj.bjc.6690778
Suchy B Austrup F Driesel G Eder C Kusiak I Uciechowski P Grill HJ Giesing M Detection of mammaglobin expressing cells in blood of breast cancer patients Cancer Lett 2000 158 171 178 10960767 10.1016/S0304-3835(00)00520-6
Wong IH Chan AT Johnson PJ Quantitative analysis of circulating tumor cells in peripheral blood of osteosarcoma patients using osteoblast specific messenger RNA markers: a pilot study Clin Canc Res 2000 6 2183 2188
Martin J Ginsberg RJ Venkatrman ES Bains MS Downey RJ Korst RJ Kris MG Rusch VW Long-term results of combined-modality therapy in resectable non-small cell lung cancer J Clin Oncol 2002 120 1989 1995 11956257 10.1200/JCO.2002.08.092
Schleiermacher G Delattre O Detection of micrometastases and circulating tumour cells using molecular biology techniques in solid tumors Bull Cancer 2001 88 561 570 11459702
Mehes G Luegmayr A Ambros IM Ladenstein R Ambros PF Combined automatic immunological and molecular cytogenetic analysis allows exact identification and quantification of tumor cells in the bone marrow Clin Canc Res 2001 7 1969 1975
Ghosh S Ahmed K Hopkinson DN Vaughan R Pulmonary adenocarcinoma is associated with poor long-term survival after surgical resection. Effect of allogeneic blood transfusion Cancer 2004 101 2058 2066 15455359 10.1002/cncr.20590
Boldrini L Cacinai A Sarnaritani E Pistolesi F Mussi A Lucchi M Angeletti CA Basolo F Fontanini G Tumor necrosis factor alpha and transforming growth factor beta are significantly associated with better prognosis in non small cell lung carcinoma: putative relation with BCL-2 mediated neovascularization Br J Cancer 2000 83 480 486 10945495 10.1054/bjoc.2000.1345
MacDonald IC Groom AC Chambers AF Cancer spread and micrometastases development: quantitative approaches for in vivo models Bioassays 2002 24 885 893 10.1002/bies.10156
| 15801980 | PMC1087511 | CC BY | 2021-01-04 16:39:04 | no | World J Surg Oncol. 2005 Mar 31; 3:18 | utf-8 | World J Surg Oncol | 2,005 | 10.1186/1477-7819-3-18 | oa_comm |
==== Front
AIDS Res TherAIDS Research and Therapy1742-6405BioMed Central London 1742-6405-2-21582630210.1186/1742-6405-2-2ResearchMonitoring processed, mature Human Immunodeficiency Virus type 1 particles immediately following treatment with a protease inhibitor-containing treatment regimen Baird Heather A [email protected] Andre J [email protected] Michael M [email protected] Alan [email protected] Donna [email protected] Daniel R [email protected] Harold A [email protected] Eric J [email protected] Division of Infectious Disease, Department of Medicine, Case School of Medicine and the Center for AIDS Research, Case Western Reserve University, University Hospitals of Cleveland, Cleveland, OH, USA2 Department of Pharmacology, Case School of Medicine, Case Western Reserve University, Cleveland, OH, USA3 Department of Immunology and Microbiology, Rush-Presbyterian-St. Luke's Medical Center, Chicago, IL, USA4 Division of Infectious Diseases, Beth Israel Medical Center, New York, NY, USA5 Harvard Medical School, Brigham and Women's Hospital, Boston, USA6 Section of Infectious Diseases, Department of Medicine, Rush-Presbyterian-St. Luke's Medical Center, Chicago, IL, USA2005 12 4 2005 2 2 2 28 2 2005 12 4 2005 Copyright © 2005 Baird et al; licensee BioMed Central Ltd.2005Baird 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.
Protease inhibitors (PIs) block HIV-1 maturation into an infectious virus particle by inhibiting the protease processing of gag and gag-pol precursor proteins. We have used a simple anti-HIV-1 p24 Western blot to monitor the processing of p55gag precursor into the mature p24 capsid immediately following the first dosage of a PI-containing treatment regimen. Evidence of PI activity was observed in plasma virus as early as 72 hours post treatment-initiation and was predictive of plasma viral RNA decrease at 4 weeks.
Protease inhibitorsHIV-1p24 antigen capture
==== Body
Background
Assembly and transport of the 55 kDa gag (p55gag) and 160 gag-pol (p160gag-pol) proteins to the inner plasma membrane is essential for the packaging of the viral genomic RNA, host tRNALys,3 primer, as well as for interactions with HIV-1 envelope glycoproteins [5]. Budding and virus release initiates the processing of the gag and gag-pol precursor proteins. This processing step likely requires the dimerization of two gag-pol precursors (at least in the region of protease) that permits a low-efficiency cleavage of the precursor proteins and release of fully active protease (PR) homodimers [16]. These enzymes then complete protein maturation to produce an infectious virus particle. Thus, protease inhibitors (PI) appear to be most active at blocking HIV-1 replication following budding of the immature virus particle [4,6]. In contrast other antiretroviral drugs (ARV) such as nucleoside reverse transcriptase inhibitors (NRTI) and non-nucleoside RT inhibitors (NNRTI), block reverse transcription during intracellular HIV-1 replication [3].
To date, the best method to monitor inhibition of HIV-1 replication is to evaluate virus concentrations in the plasma [10]. Several commercial, FDA-approved assay kits (HIV-1 Quantiplex (bDNA) assay, AMPLICOR HIV-1 MONITOR assay, NucliSens HIV-1 assay) involve measuring virus levels via reverse transcription-PCR amplification of genomic HIV-1 RNA [8]. It is important to recognize however, that these assays cannot monitor the pharmacodynamic properties of many antiretroviral agents immediately following treatment initiation. Protease inhibitors block HIV-1 protease processing following virus release from cells in contrast to NNRTIs or NRTIs that inhibit during an intracellular replication step, i.e. reverse transcription. The half life of plasma virus is estimated to be approximately 6 hrs [13,18]. but the half life of activated CD4+ cells infected with and producing HIV even in the presence of PIs is approximately 1.2 days [13,18,19] during phase I decay An assay measuring levels of viral RNA does not distinguish between the immature virus (processing blocked by PIs) and infectious virus, both of which encapsidate HIV-1 genomic RNA. The estimated time required for protease inhibitors to clear the majority of free virus particles from the circulation and activated cells (not latently infected or quiescent cells) is approximately 4 weeks. Thus, a viral RNA assay performed on plasma does not provide a complete assessment of PI activity for at least 1–4 weeks.
Methods
In this study, we tested the ability of three different assays to measure the quantity of both infectious virions and defective/immature virus particles in the plasma of HIV-infected patients who started treatment with a PI-containing regimen. The performance of three assays was validated in vitro utilizing HIV-1 infected cell lines in the presence or absence of PIs and other ARVs. These tests were followed by in vivo analyses using plasma samples from patients receiving a PI-based treatment regimen. The following provides a brief summary of the first two assays that could detect the effects of PI activity in vitro but failed in vivo.
The first assay involved measurement of infectious virus potential. We serially diluted cell-free culture supernatants from chronically HIV-infected U87.CD4.CCR5 cells treated with PIs. This diluted and undiluted plasma was then added to uninfected peripheral blood mononuclear cells (PBMC). Although this assay could be used to measure infectious potential of high titer viruses in tissue culture, only plasma containing extremely high viral loads (> 104 viral RNA copies/ml) could support any HIV-1 infection of PHA/IL-2 treated PBMC regardless of the patients treatment status (data not shown). Concentrating the virus by ultracentrifugation did little to increase infectious titer of virus from plasma.
The second assay involved PCR amplification of strong-stop viral DNA found in cell-free virus. Previous reports have shown that viral DNA is found HIV-1 particles [9,17] but that steric hindrance or the lack of dNTP substrates limit reverse transcription and presence of viral DNA to 1:1000 to 1:10,000 virions [1,2]. We have shown that a defective protease abolishes the synthesis of any HIV-1 DNA in virus particles [1,2]. HIV-1 strong-stop DNA was not detected by PCR amplification in virus produced from the chronically infected cells in the presence of PIs (data not shown). However, viral loads of >10,000 RNA copies/ml were required in patients to even detect the presence of HIV-1 DNA in plasma, which is consistent with previous findings. Thus, this assay was not effective for those patients starting PI therapy with lower viral loads (<103–4 RNA copies/ml).
In contrast to the assays described above, an anti-p24 Western blot was successful in measuring both in vitro and in vivo PI effects and was the simplest in design and application. To initially test this assay we infected U87.CD4.CXCR4 cells with a wild type HXB2 virus or the protease inhibitor resistant virus, RF (containing PR mutations V82F and I84V) [12]. Following established infection and stable virus production over three days (as measured by RT activity in the culture supernatant), cultures were treated with 0.2 and 20 nM lopinavir (LPV) or 2 and 200 nM nevirapine (NVP). The higher concentration of each drug was approximately 100-fold greater than the reported IC50 values (i.e. lower concentrations of each drug) [11,14,15]. Cell free culture supernatant (1 ml) was then harvested at 0, 4, 8, 24, and 72 h post drug addition. Virus was pelleted from the supernatant by ultracentrifugation (35,000 g for 1 h) and then resuspended in 50 μl of sodium-dodecyl sulfate (SDS) lysis buffer (1% SDS, 10% glycerol, 10% β-mercaptoethanol, 0.04 M Tris pH 6.8); of which 10 μl were heated to 95°C, separated on Tris-HCl-12.5% polyacrylamide precast gels (Bio-Rad), and transferred onto polyvinylidene difluoride membranes (Immobilon P; Millipore)by electroblotting (BioRad). Membranes were incubated with blocking reagent (5% milk-0.05% Tween in phosphate-bufferedsaline) for 1 h at room temperature then hybridized with a mouse anti-p24 monoclonal antibody (diluted 1:1,000; Fitzgerald Industries International, Inc.) overnight at 4°C. After washing, membranes were incubated with horseradish peroxidase-conjugated goat anti-mouse IgG1 antiserum (diluted 1:40,000; Pierce) for 3 hours. Immune complexes were visualized with the ECL system (Amersham) according to the manufacturer's instructions and films were analyzed using BioRad Quantity One software.
Results and discussion
Fig. 1, panels A and B show Western blot analyses and p55:p24 ratios of two viruses grown in tissue culture, RF (protease inhibitor resistant) and HXB2 (protease inhibitor sensitive) and treated with 20 nM LPV. Two major bands appeared on the film of the ECL blot. The faster migrating product was the processed capsid (CA) p24 and the slower migrating band was unprocessed p55gag. We observed minute amounts of partly cleaved gag product containing the matrix (MA) p17 and CA p24, i.e. 41 kDa. In these experiments, the HXB2 p55:p24 is increasing over the first 72 hours of 20 nM LPV treatment indicating an inhibition of gag processing. In contrast, there was no evidence of decreased gag precursor processing with the RF virus treated with 20 nM LPV as indicated by a constant p55:p24 ratio of 0.1 during the 72 h of treatment. Treatment with a lower concentration of LPV (0.2 nM) or with NVP (2 or 200 nM) did not result in a significant difference in the HXB2 or RF p55:p24 ratios over time (data not shown). In tissue culture, longer incubation times with LPV (20 nM) reduced virus production to levels at which the quantities of p55 and p24 products were difficult to detect by Western blot analyses.
Figure 1 Western blots for the HIV-1 gag proteins in HIV-1 produced in tissue culture following treatment with protease inhibitors. U87/CD4/CXCR4 cells were plated in 6 well plates at 80,000 cells/well and allowed to grow to confluence. The cells were infected with either HXB2 or RF/V82F/I84V (protease inhibitor resistant virus) and RT activity was monitored. On day 3 of culture, infected cells were treated with 20 nM LPV, and 1 ml of media was removed at 0, 4, 8, 24, and 72 hours post-drug treatment. The virus was pelleted, and the pellet was then lysed using sodium-dodecyl sulfate (SDS) lysis buffer and then run on a 10% SDS polyacrylamide gel. Following transfer to nylon membranes, blots were probed with the primary mouse anti-p24 antibody and the horseradish peroxidase-conjugated goat anti-mouseantiserum. Films were exposed following treatment with the ECL kit (panel A). Ratio of unprocessed p55gag to processed CA p24 over a 72 hour time course was determined by scanning the blots and quantifying the bands (panel B).
To test the utility of this simple Western blot assay to monitor initial PI treatment, nine patients were enrolled into the A5036s Substudy of the ACTG Clinical Trial Substudy A5014 [7]. All of the patients in A5014 were ARV treatment naïve and were randomized to receive either LPV+ the non-nucleoside RT inhibitor NVP or NVP + three nucleoside RT inhibitors: lamivudine (3TC) + stavudine (D4T) + abacavir (ABV). All participants and investigators in this study were blinded to the treatment arms. Ten ml of blood was drawn into Acid Citrate Dextrose (ACD) tubes prior to the first drug administration, then 4, 8, 12, 24, 72 hours, three days and four weeks post drug administration. Two aliquots of 3.5 ml of plasma were shipped on dry ice to CWRU and then stored at -70°C prior to analyses.
In addition to the Western blot analyses, the sub-study also called for a measure of infectious potential by HIV-1 found in plasma. For these tests we exposed HIV-negative peripheral blood mononuclear cells to plasma samples obtained prior to and immediately following treatment with the PI- or non PI-containing HAART regimen. Only plasma samples of one patient (of 9) resulted in productive infection of PHA/IL-2 treated PBMC cultures suggesting that this not a sensitive assay. Unfortunately, no assessment of PI activity could be evaluated using this infectious assay since this patient was treated with NVP and the three NRTIs. It is unlikely that viral levels in plasma is the sole factor contributing to the ability of plasma virus to infect PBMC cultures since all patients in this substudy had viral RNA loads approximately 104 copies/ml at initiation of treatment. The level of virus production or success of PBMC infections did not increase if the virus was concentrated from plasma by ultracentrifugation. This concentration step would also remove any residual drug in plasma that might affect infectivity of virus in plasma after the initial treatment.
Preliminary data indicated that plasma protein concentrations were too high to efficiently concentrate virus and resulted in excessive background on the Western blot for HIV-1 gag proteins. Thus, 1.5 ml of each plasma sample was diluted with 3.5 ml of phosphate-buffered saline (PBS) prior to concentrating the virus by ultracentrifugation. The procedures for the Western blot analyses are described above. Fig. 2A shows the Western blot results employing samples from patients A and B from the ACTG5014 clinical trial. It is important to note that this was a double-blinded trial and that all samples from each patient were analyzed prior to knowledge of treatment regimens [7]. Interestingly, the ratio of p55:p24 was greater than 1 in 8 of 9 patient samples prior to ARV treatment. High p55:p24 ratios suggest an increased proportion of noninfectious virus particles to infectious virions in the plasma. The ratio of p55:p24 in HIV-1 propagated in tissue culture is typically much less than one, suggestive of higher proportions of infectious to non-infectious virus in plasma. Most plasma proteins or free viral proteins were separated from the virus via centrifugation and pelleting of the virus. A 95% reduction of Coomassie blue staining of all proteins on the SDS PAG following transfer suggested that both the p55 and p24 proteins were efficiently electrotransferred to the nylon membranes. Increased ratios of p55:p24 was not due to selective antibody binding to the p55gag considering the anti-p24 antibody should bind at least as efficiently to CA p24 than to the unprocessed p55gag. It is also possible that background p55 is due to the rapid turnover, and therefore that nascent virions make up a large fraction of the total. These virions could be more infectious than the fully processed ones seen in cell culture, since they would be rapidly processed in the course of the assay.
Figure 2 Western blot analyses for the HIV-1 gag proteins in patient plasma prior to and following ARV treatment. Patient plasma was obtained at 0, 4, 8, 12, 24, 72 hours and 4 weeks following ARV treatment. Plasma was diluted with serum-free media and then centrifuged to pellet HIV-1 particles prior to the analyses (see Fig. 1). Panel A shows the Western blot results on plasma samples obtained from patient A who was treated with NVP+3TC+D4T+ABV and patient B who was treated with LPV+NVP. Panel B shows the ratios of unprocessed p55gag to processed CA p24 in patients treated with NVP+3TC+D4T+ABV (patient A) or NVP+LPV (patient B, C, and D). Each bar at each time point represents analyses from a separate Western blot. Panel C is a plot showing the changes in CD4 cell count (cells/mm3; open squares) and viral RNA load in plasma (copies/ml, filled diamonds) following treated with either treatment regimen.
We predicted that the p55:p24 ratio would increase during first three days of PI treatment with the possible dips in this ratio between PI dosages (every 12 h). Preliminary data with patients starting a PI-containing treatment regimen suggest a PI-mediated inhibition of p55 processing within 8–12 h of treatment (data not shown). However, these studies were performed with patients starting RIT+SAQ or IND-containing treatment regimens and not with patients treated with LPV. As indicated by the results of tissue culture infection experiments shown in Fig. 1, the p55:p24 ratio should remain stable in plasma samples obtained from patients receiving non-PI containing HAART regimens (i.e. NVP+3TC+D4T+ABV) since neither NRTI nor NNRTI inhibit processing of the gag or gag-pol precursors.
Although only one example of these analyses is shown (Fig. 2B, panel I and II), the Western blot results of plasmas from each of four patients treated with the NVP+3TC+D4T+ABV combination showed a constant ratio of p55:p24 following treatment. In patients randomized to receive the LPV+NVP regimen, the ratio of p55:p24 increased at 72 h following the initial dosing (Fig. 2B). This increase in the p55:p24 ratio was maintained after 4 weeks of PI treatment. Previous findings revealed that HAART resulted in a drop in RNA and plasma infectivity in one day [20], however, the efficacy of ARV treatment can be affected by factors such as drug concentrations, compliance, potency, and selection of ARV resistant quasispecies. Unfortunately, two of the patients randomized to receive the LPV+NVP combination dropped out of the 5036 sub-study prior to the 72 h sample collection, i.e. the time that is likely required to detect a LPV block on viral protein maturation. In one patient, the p24 band on the Western blot was below the limit of detection in all plasma samples. There was an apparent delay in LPV activity following treatment in vivo as compared to treatment in tissue culture (Figs. 1 and 2). A longer time was likely required to attain inhibitory concentrations in blood or other tissues whereas the effect of LPV on newly produced virus particles was immediate in tissue culture.
Nearly all antiretroviral drug-naïve patients recruited into AACTG 5014 demonstrate a drop in viral RNA load to undetectable levels after 8 weeks of treatment with either regimen. This viral load decrease was associated with an increase in CD4 cell counts. In patients B, C, and D, the drop in viral load was likely mediated by both NVP and LPV but the initial viral RNA load decrease (within one week) could be more of a measure of NVP than LPV inhibitory activity (Fig. 2C). Within three days, PI appeared to be blocking protease cleavage of precursor gag proteins in the virus particles found in plasma (Fig. 2B). This ratio increased only slightly during the next four weeks. Because this was a pilot study on a limited number of patients, it is difficult to ascertain what constitutes a significant change in p55:p24. However, it appears that there is a significant difference observed with the PI-containing regimen at 72 hours and 4 weeks. The minimal drop in viral load observed after three days post NVP+LPV treatment (<1 log) increased from a 13- to over a 100-fold decrease after four weeks of treatment (1.5 to 3 log decrease; panels II, III, IV in Fig. 2C and 3A). Interestingly, the relative decrease in viral load among these three LPV+NVP treated patients at four weeks also appeared to correspond to relative inhibition of protease cleavage at only three days post treatment (Fig. 3). Patient B showed a delayed and slower drop in viral load (at four weeks) and a higher increase in the p55:p24 ratio (at three days) than that observed in patient C and D (Fig. 3). A greater increase in p55:p24 ratio in patient B reflected the very low level of p55 detected. There was significant variation in the p55:p24 ratio (detected by Western blot) amongst all of the PI-naïve patients. Although difficult to test, this variation may be related to varying ratios of infectious virion:non-infectious virus particle production in HIV-infected individuals. These data suggest that the ratio of p55:p24 at three days following initiation of PI treatment may be predictive of the immediate HIV-1 inhibition by PIs in a patient. It should be noted that without the enrollment of more patients starting PI-based therapy, it is difficult to understand the relationship between the relative increase in the p55:p24 ratios and response to therapy except that the rapid increase in the ratio is strong indicator of anti-HIV PI activity in the patient.
Figure 3 Comparing the drop in viral RNA load to the increase in precursor gag protein following treatment with LPV and NVP. The fold increase in the p55:p24 ratio was calculated by dividing these ratios at 3 days and 4 weeks by the observed ratio prior to treatment (time 0) (panel A). Panel B shows the fold decrease in viral RNA load at 3 days and 4 weeks following LPV+NVP treatment. This calculation involved dividing the viral load at day 3 and week 4 by that at time 0.
Conclusion
In summary HIV protease inhibitors block the processing of p55gag and p160gag-pol precursor proteins during virus budding or following virus release. However, the protease inhibitor does not impede incorporation of genomic HIV-1 RNA into the virus particle. Thus, following PI treatment, viral load assays based on detection of viral RNA measure both noninfectious, immature virus particles and virions found in plasma. We have developed a method to measure the anti-HIV activity of a protease inhibitor using a simple approach. In three patients treated with LPV+NVP, the ratio of unprocessed p55:processed p24 increased at 72 hours and over the next four weeks of treatment. In contrast, the ratio of HIV-1 p55:p24 did not change over the four weeks of study in patients treated with an NNRTI-containing regimen (NVP+3TC+D4T+ABV). This pilot study, though limited in patient number, has provided evidence that an HIV-1 p24 Western blot can be used to immediately measure the antiviral activity of protease inhibitors. Preliminary in vitro data also suggests that inability of PIs to block PI-resistant HIV-1 in patients could be assessed within 3 days of treatment. Based on these findings we are now testing the utility of this assay in highly PI experienced patients to predict the success of a new PI-containing treatment regimen within 3 days of starting this therapy. In addition, this study indicates that western blot is an excellent tool for the evaluation of the activity of protease inhibitors in vitro, and may be useful in evaluating new drugs putatively active against isolates resistant to current agents, or to evaluate the activity of different combinations of protease inhibitors using a more insightful measure than viral infectivity.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
H.B. and A.J.M. performed the laboratory work presented in this paper. E.J.A. supervised the laboratory work. M.M.L. and A.L. were the PI's of the parent study A5014. D.M. was the statistician for A5014 and 5036. D.R.K. was the protocol virologist for A5014.
Acknowledgements
Research for this study was performed at Case Western Reserve University (E.J.A.) and was supported by funds from Social and Scientific Systems Inc. and the NIH/NIAID AIDS Clinical Trial Group. We thank the AACTG5014 and AACTG 5036 s team for their coordination of samples collection and support. Additional support was provided to EJA from the National Institute of Allergy and Infectious Diseases, NIH (AI49170). All virus work was performed in the Biosafety Level 2 and 3 facilities of the CWRU Center for AIDS Research (AI36219).
==== Refs
Arts EJ Mak J Kleiman L Wainberg MA DNA found in human immunodeficiency virus type 1 particles may not be required for infectivity J Gen Virol 1994 75 1605 1613 7517431
Arts EJ Mak J Kleiman L Wainberg MA Mature reverse transcriptase (p66/p51) is responsible for low levels of viral DNA found in human immunodeficiency virus type 1 (HIV-1) Leukemia 1994 8 S175 S178 7512179
Arts EJ Wainberg MA Mechanisms of nucleoside analog antiviral activity and resistance during human immunodeficiency virus reverse transcription Antimicrob Agents Chemother 1996 40 527 540 8851566
Ashorn P McQuade TJ Thaisrivongs S Tomasselli AG Tarpley WG Moss B An inhibitor of the protease blocks maturation of human and simian immunodeficiency viruses and spread of infection Proc Natl Acad Sci U S A 1990 87 7472 7476 2217178
Coffin JM Fields BN, Knipe DM, Howley PM Retroviridae: The viruses and thier replication Fundamental Virology 1996 763 843
Lambert DM Petteway SR JrMcDanal CE Hart TK Leary JJ Dreyer GB Meek TD Bugelski PJ Bolognesi DP Metcalf BW Human immunodeficiency virus type 1 protease inhibitors irreversibly block infectivity of purified virions from chronically infected cells Antimicrob Agents Chemother 1992 36 982 988 1510424
Landay AL Spritzler J Kessler H Mildvan D Pu M Fox L Kuritzkes D Ledeman MM Immune Reconstitution Is Comparable in Antiretroviral (ARV) Naive Subjects Following 1 Year of Therapy with a Nucleoside Reverse Transcriptase Inhibitor (NRTI) or Protease Inhibitor (PI) Containing ARV Regimen J Infect Dis 2003 188 1444 54 14624369 10.1086/379041
Liegler TJ Grant RM HIV Viral Load Assay. UCSF 2003
Lori F M di V de Vico AL Lusso P Reitz MS JrGallo RC Viral DNA carried by human immunodeficiency virus type 1 virions J Virol 1992 66 5067 5074 1378514
Mellors JW Rinaldo CR JrGupta P White RM Todd JA Kingsley LA Prognosis in HIV-1 infection predicted by the quantity of virus in plasma Science 1996 272 1167 1170 8638160
Merluzzi VJ Hargrave KD Labadia M Grozinger K Skoog M Wu JC Shih CK Eckner K Hattox S Adams J Inhibition of HIV-1 replication by a nonnucleoside reverse transcriptase inhibitor Science 1990 250 1411 1413 1701568
Otto MJ Garber S Winslow DL Reid CD Aldrich P Jadhav PK Patterson CE Hodge CN Cheng YS In vitro isolation and identification of human immunodeficiency virus (HIV) variants with reduced sensitivity to C-2 symmetrical inhibitors of HIV type 1 protease Proc Natl Acad Sci U S A 1993 90 7543 7547 8356053
Perelson AS Neumann AU Markowitz M Leonard JM Ho DD HIV-1 dynamics in vivo: virion clearance rate, infected cell life-span, and viral generation time Science 1996 271 1582 1586 8599114
Richman D Shih CK Lowy I Rose J Prodanovich P Goff S Griffin J Human immunodeficiency virus type 1 mutants resistant to nonnucleoside inhibitors of reverse transcriptase arise in tissue culture Proc Natl Acad Sci U S A 1991 88 11241 11245 1722324
Sham HL Kempf DJ Molla A Marsh KC Kumar GN CM Chen Kati W Stewart K Lal R Hsu A Betebenner D Korneyeva M Vasavanonda S McDonald E Saldivar A Wideburg N Chen X Niu P Park C Jayanti V Grabowski B Granneman GR Sun E Japour AJ Norbeck DW ABT-378, a highly potent inhibitor of the human immunodeficiency virus protease Antimicrob Agents Chemother 1998 42 3218 3224 9835517
Swanstrom R Wills JW Coffin JM, Hughes SH, Varmus HE Synthesis, Assembly, and Processing of Viral Proteins Retroviruses 1997 205 334
Trono D Partial reverse transcripts in virions from human immunodeficiency and murine leukemia viruses J Virol 1992 66 4893 4900 1378513
Wei X Ghosh SK Taylor ME Johnson VA Emini EA Deutsch P Lifson JD Bonhoeffer S Nowak MA Hahn BH Shaw G Viral dynamics in human immunodeficiency virus type 1 infection Nature 1995 373 117 122 7529365 10.1038/373117a0
Blankson JN Finzi D Pierson TC Sabundayo BP Chadwick K Margolick JB Quinn TC Siliciano RF Biphasic decay of latently infected CD4+ T cells in acute human immunodeficiency virus type 1 infection J Infect Dis 2000 182 1636 42 11069234 10.1086/317615
Perelson AS Essunger P Cao Y Vesanen M Hurley A Saksela K Markowitz M DHo D Decay characteristics of HIV-1-infected compartments during combination therapy Nature 1997 387 188 191 9144290 10.1038/387188a0
| 15826302 | PMC1087826 | CC BY | 2021-01-04 16:38:33 | no | AIDS Res Ther. 2005 Apr 12; 2:2 | utf-8 | AIDS Res Ther | 2,005 | 10.1186/1742-6405-2-2 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1021584017410.1186/1471-2105-6-102Research ArticleEvolutionary distance estimation and fidelity of pair wise sequence alignment Rosenberg Michael S [email protected] Center for Evolutionary Functional Genomics, The Biodesign Institute, and the School of Life Sciences, Arizona State University, Tempe, AZ 84287-4501, USA2005 19 4 2005 6 102 102 16 12 2004 19 4 2005 Copyright © 2005 Rosenberg; 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
Evolutionary distances are a critical measure in comparative genomics and molecular evolutionary biology. A simulation study was used to examine the effect of alignment accuracy of DNA sequences on evolutionary distance estimation.
Results
Under the studied conditions, distance estimation was relatively unaffected by alignment error (50% or more of the sites incorrectly aligned) as long as 50% or more of the sites were identical among the sequences (observed P-distance < 0.5). Beyond this threshold, the alignment procedure artificially inflates the apparent sequence identity, skewing distance estimates, and creating alignments that are essentially indistinguishable from random data. This general result was independent of substitution model, sequence length, and insertion and deletion size and rate.
Conclusion
Examination of the estimated sequence identity may yield some guidance as to the accuracy of the alignment. Inaccurate alignments are expected to have large effects on analyses dependent on site specificity, but analyses that depend on evolutionary distance may be somewhat robust to alignment error as long as fewer than half of the sites have diverged.
==== Body
Background
Evolutionary distance, the number of substitutions per site separating a pair of homologous sequences since they diverged from their common ancestral sequence, is an extremely important measure in molecular evolution and comparative genomics. It is used for a wide variety of purposes, ranging from phylogenetic analysis [1,2], to estimating times of divergence [3,4], the tempo and mode of evolutionary change [5], and functional constraints [6,7]. Evolutionary distance estimation is often one of the first steps in high-throughput sequence analysis; errors in these estimates may have wide-ranging consequences on downstream analyses and conclusions.
There are many ways to estimate evolutionary distance; accuracy of various methods tends to be dependent on proper specification of the substitution model and sequence length [8,9]. One factor that has not been well examined with respect to evolutionary distance estimation, however, is alignment (although see [10-12]). Sequence alignment is an extremely common analytical tool used in comparative genomics. The purpose of alignment is to identify positions in homologous sequences that are descended from a common ancestor. Because alignment is the first step in many complex, high-throughput studies [13], it is often forgotten that alignment algorithms produce a hypothesis of homology (just as a phylogenetic tree is a hypothesis of evolutionary history). As with other hypotheses, these alignments may contain more or less error depending on the nature of the data. While it is widely recognized that highly divergent sequences are more difficult to align and will contain more error than less divergent sequences (e.g., [14,15]), the nature of this error appears to be underappreciated and is usually ignored.
Little attention has been paid to how errors in sequence alignment affect downstream analysis. Individual studies have shown that error in alignment can have broad effects on computational approaches to discovering functional elements [16,17] and phylogenetic analysis (e.g., [18-24]); these studies have been based on specific data sets and generally show that different results are obtained by different alignments, rather than estimating the amount of error generated by incorrect alignment.
I performed a simulation study to examine the relationship between global alignment accuracy and evolutionary distance estimation of noncoding DNA sequences. It consists of a profile of the magnitude of error one expects to find in paired sequence alignment under the simulation conditions and the comparison of evolutionary distance estimates from correct and hypothesized alignments as the true divergence increases.
Results and discussion
Under the baseline simulation conditions, alignment accuracy (measured as the proportion of aligned sites that are truly homologous) is largely dependent on the proportion of homologous sites containing identical nucleotides. When sequence identity exceeded 80%, essentially all aligned sites (>99%) were truly homologous (Figure 1). As identity declined, the proportion of correctly aligned sites rapidly decreased. When identity reached 65%, about 90% of the aligned sites were still correct, but when identity reached 50% accuracy dropped to 30–65% (depending on the complexity of the substitution model). When fewer than 50% of truly homologous sites were identical, alignment accuracy becomes essentially zero.
A distinction needs to be made between the true identity of the sequences (the proportion of truly homologous sites in a pair of sequences containing identical nucleotides) and the aligned identity (the proportion of hypothesized homologous sites from an alignment that contain identical nucleotides). The nature of alignment algorithms is to predict homology by inserting gaps so that sites with identical nucleotides align. When true variation among sequences is large, algorithms can be quite efficient at incorrectly inferring identity. The theoretical minimum identity for a pair of sequences under the present simulation conditions is 25–26% (depending on the specific substitution model), yet Clustal yields sequences with a minimum identity of 44% (Figure 1), even for random data (similar results have been reported by others, e.g., [10,25]). The inflation in observed identity is predominantly found in sequences that truly differ by more than 50% of their sites; sequences with true identity of 50% or more have less than a 1% absolute increase in observed identity after alignment.
The results represented in Figure 1 describe the accuracy of pair wise alignment by Clustal under the specific simulation conditions and alignment parameters. While these exact profiles cannot be taken as representative of alignment accuracy for all sequences and algorithms, the shape of the curve probably does reflect a general pattern. Different evolutionary conditions and algorithms may lead to varying inflection points, but the general shape of the curve is likely to be constant (e.g., similar accuracy curves were found in [26]).
Up to a point, evolution distance estimation is somewhat robust to alignment error (Figures 2,3). The relative difference between evolutionary distances estimated from the true and hypothesized alignments (= |dtrue - dalign|/dtrue) is less than 10%, even when up to 50% of the sites are aligned incorrectly (Figure 3). Distance estimates from the hypothesized alignments begin to differ to a greater extent from the true alignment only when more than half of the sites are aligned incorrectly. For the JC and HKY substitution models, alignment inaccuracy did not have an effect on distance estimation for true distances less than 1.0 (Figure 2). For HKY + Γ, alignment inaccuracy produced little effect on distance estimation even when true distances were as large as 2.0. When scaled to percent identity (the proportion of aligned sites that contain the same nucleotides in the paired sequences), the curves for the different substitution models become congruent (Figure 3B). The robustness of these estimates appears to be related to the inflation of sequence identity (Figure 1C). As long as the true identity is greater than 50%, there is little inflation in the estimated identity due to alignment (even when the alignment is largely wrong). This translates to relatively little error in the estimation of distance, because distance estimates are based solely on the observed proportions of sites that differ among the sequences; JC distance is based on the overall count, Tamura-Nei distances partition the count into transversions and purine and pyrimidine transitions. Since these counts are being estimated reasonably accurately (even though the specific sites are wrong) the distance estimates are also reasonably accurate. The 50% barrier for distance estimate accuracy was also reported by [10] using a least-squares approach to estimating P-distance.
It may be possible to reduce the sequence identity inflation by changing the gap and mismatch penalties (or using more sophisticated methods of alignment); these changes would also alter the accuracy of the alignment. The purpose of this study was not to test the best possible way to construct alignments, but rather to examine the effects of typical alignment errors on evolutionary distance estimation from DNA sequences.
The effects of evolutionary parameters on alignment and distance estimation accuracy varied by parameter, but one general observation is that when the value of a specific parameter has an effect, the effect is amplified when the true distance between the pair of sequences is larger (Figure 4). While sequence length generally affects the accuracy of evolutionary distance estimates [9], there was no interaction between sequence length and the effect of alignment accuracy on evolutionary distance estimation. The mean alignment accuracy was unaffected by sequence length, although standard deviations were much reduced for longer sequences (Figure 4A). On the other hand, distance estimates became marginally better with longer sequences (Figure 4B). In contrast to sequence length, increasing both indel size and rate had large effects on alignment accuracy (Figures 4C, E). Evolutionary distance estimation, however, was unaffected by changes in these parameters (Figures 4D, F), showing a lack of association between alignment accuracy and evolutionary distance estimation (i.e., Figure 4 shows a 40% decrease in alignment accuracy with essentially no corresponding change in the accuracy of estimated evolutionary distances).
Changing the α parameter of the Γ-distributed rate variation had an easily predictable effect on alignment accuracy given the results in Figures 2,3. Decreasing α, increases the magnitude of intersite rate variation (an α of infinity indicates equal rates among all sites); thus, decreasing α will increase the proportion of identical sites found among the paired sequences since substitutions will occur at a fewer sites. As already shown, increased identity among the sequences leads to increased accuracy of alignment, a result confirmed in Figure 4G. Accuracy of distance estimates appear driven by the sensitivity of the results to proper specification of the site rate distribution [9], and a relationship with alignment accuracy is uncertain. Estimation of evolutionary distance with intersite rate variation substitution models requires user specification of the Γ-distribution shape parameter α. There are no established methods for estimating α from only a pair of sequences (all described methods require 3 or more sequences). The effect of alignment accuracy on distance estimation is dependent on the accuracy of α (results not shown). When α is underestimated (i.e., intersite rate variation is less than predicted), evolutionary distance will be underestimated for both the true and hypothesized alignments, but the difference between these estimates is reduced (relative to the correct specification of α). When α is overestimated (i.e., intersite rate variation is more than predicted), evolutionary distance will be overestimated for both the true and hypothesized alignments and the difference between these estimates is accentuated.
Not surprisingly, nucleotide frequency biases have a large effect on the accuracy of both alignment and distance estimation (Figures 4I–J). Alignment accuracy decreases with increasing nucleotide frequency bias due to the increased probability of false homology (Figure 4I). A corresponding increase in the error of distance estimation is seen (Figure 4J), but the lack of alignment accuracy cannot necessarily be considered causal. Note the contrast between this result and that for indel rate and size (Figures 4C–F). Increased indel rate and size show a similar magnitude of effect on alignment accuracy as nucleotide frequency, but without the corresponding change in distance estimation. This emphasizes the independence of distance estimation to alignment accuracy for moderate evolutionary divergences.
Overall, these results are somewhat encouraging, particularly when one considers that the more realistic substitution models (i.e., HKY + Γ in this study) are more robust to alignment error for much longer evolutionary distances. However, the robustness of these distance estimates is extremely context dependent. Whether a 10% error in distance estimate is large or small depends on the questions being asked as well as the relative distances of other sequence pairs being analyzed.
Some of the general results in this study have been reported previously [10,11,27], but the present study differs from these in the inclusion of more complicated substitution models (HKY + Γ vs. JC) and distance estimates (Tamura-Nei vs. P-distance), as well as a somewhat different approach to sequence alignment. Clustal is among the most commonly used alignment programs and implements a variation of the most commonly used pair wise alignment method, the Needleman-Wunsch algorithm [28]. Algorithms which make statistical estimates of alignment, either maximum likelihood or Bayesian [27,29-32], may also incorporate evolutionary distance estimation, sometimes estimating distances over the alignment probability landscape [12]. These methods may be more accurate than Clustal and thus the relationships between alignment accuracy and distance estimation may be very different for these approaches than those described within this study [10,11,27]. One goal of this study was to profile alignment and distance estimation errors as commonly used by the bioinformatics and genomics community; the methods I employed in the present work are much more commonly used that are the statistical alignment procedures.
The performed simulations represent a global alignment condition (there were no rearrangements which would change homology of the overall sequences) and thus focused on global alignment. Local alignment programs and algorithms, such as BlastZ [33] or Dialign [34], implicitly assume that subsections of the sequences are simply not homologous (or that homologous regions occur in different orders). By finding only conserved regions, local alignment algorithms essentially decrease the probability of false positives (aligned sites which are not truly homologous) while increasing the number of false negatives (unaligned sites which are truly homologous). Thus, within the aligned regions, local alignment may be expected to be more accurate than global alignment, but also may lead to underestimates of evolutionary distances since the poorly conserved homologous regions will likely be excluded from the alignment. The tradeoffs between local and global alignment with respect to distance estimation need to be explored in some depth.
Because alignment error appears to be somewhat underappreciated by the genomics community, the alignment error profiles are in-and-of themselves interesting. While it is generally known that sequences become difficult to align as they diverge (e.g., [14,15]), the precipitous decline in accuracy (Figure 1) has only recently been profiled through simulation [26]. Not surprisingly, the exact nature of these curves appears to be highly dependent on indel size and rate (Figure 4). To some extent, the alignment accuracy profiled in Figure 1 can be viewed as a best-case-scenario since the simulation parameters could be considered realistic, but otherwise low, values. As insertion and deletion events increase in size and rate, alignment accuracy, particularly for more divergent sequences will decline precipitously. It is certainly possible that additional accuracy may be recovered by use of different alignment algorithms or better optimization of the alignment parameters.
Conclusion
In this study, I've shown evolutionary distance estimation to be somewhat robust to errors in alignment for moderate divergences (>50% identity). Other uses of aligned data, including for example, identification of conserved sites relative to exploration of genetic disease [35,36], are likely to be highly dependent on the accuracy of alignment and even a tiny error may have a large effect on the results. Different alignments are known to lead to different hypotheses in phylogenetic analysis [18,19]; how various phylogenetic methods respond to alignment error is an open question in need of future study.
Methods
Three large sets of simulations were conducted, each differing by the nucleotide substitution model: Jukes-Cantor (JC) [37], Hasegawa-Kishino-Yano (HKY) [38], and HKY + Γ distributed site rate variation. A summary of all simulation conditions is found in Table 1. For JC, initial sequences consisted of 1000 random nucleotides, with the expected base composition equal for all nucleotides (i.e., 25% each). Initial sequences were replicated into a pair of independent lineages and allowed to evolve under the JC model of evolution to an expected fixed divergence (the realized number of substitutions were drawn from a Poisson distribution), ranging from 0.02 to 2.0. Insertions and deletions were also allowed to occur, with the expected rate of deletion events being one occurrence every 40 substitutions and the expected rate of insertion events being one occurrence every 100 substitutions (as observed in primates and rodents) [39]. Realized number of insertions and deletions were drawn from a Poisson distribution with mean equal to the expected value. The lengths of individual insertion and deletion events were also chosen from a truncated (so as not to include zero) Poisson distribution with a mean of 4 bases (as observed from primate and rodent lineages) [39,40]. Each simulation condition was replicated 1000 times.
The second set of simulations conducted was identical to the first, except using the HKY model of nucleotide substitution. For this model, initial and expected nucleotide frequencies were πC = πG = 0.3, πT = πA = 0.2, and the transition-transversion bias was set to that observed at neutral sites in mammals, κ = 3.6 [41]. The third set of simulations conducted was identical to the second, except allowing rate variation among sites within the sequence, modelled by a gamma distribution with a shape parameter of 1.0 [9].
Sequence length is known to play an important role in evolutionary distance estimation; to test for an interaction between sequence length and alignment accuracy, subsets of the HKY simulations were repeated with initial sequences of 100, 200, 300, 400, 500, 1500, 2000, 5000, and 10000 bases. To test the effect of rate and size of insertions and deletions on distance estimation, subsets of the HKY simulations were repeated with mean indel lengths of 2, 6, 8, and 10 bases (the original simulations had a mean of 4 bases) and with insertion and deletion rates of 1 every 200 (insertion) & 80 (deletion) substitutions (half the original rate), 150 & 60 substitutions (2/3 the original rate), 75 & 30 substitutions (4/3 the original rate), and 50 & 20 substitutions (double the original rate). The effects of nucleotide frequencies (G+C% = 60%, 70%, 80%, and 90%) and gamma-distributed rate variation (α = 0.25, 0.5, 1.0, and ∞) were similarly examined.
For every simulated data set, the fate of each of the original sites was tracked and an alignment representing the true homology was constructed for each data set (that is, the simulation program produced gapped sequences in which all aligned sites were truly homologous). The gaps were removed from the sequences and each data set was then aligned using Clustal W version 1.83 [42] with the default parameters, as is common in high-throughput analysis and comparative studies of this sort [26,34,43-45]. This produced a hypothesized alignment, just as one would obtain from analysis of real data. Clustal is one of the most widely used global alignment programs, particularly for high-throughput genomic analysis, and tends to be among the most accurate [26,46]. While it is possible that another program or algorithm or changing the default Clustal parameters might lead to more accurate alignments, the primary purpose of this study is not to highlight the accuracy of this (or any other) alignment program, but rather to examine the effects of alignment error on evolutionary distance estimation. One could purposefully misalign the sequences by hand, but using a common alignment program allows us to create errors consistent with those found in alignment of real data.
Evolutionary distances between sequence pairs were estimated for both the correct and hypothesized alignments using the Jukes-Cantor [37], Tamura-Nei, and Tamura-Nei + Γ formulas [47], as appropriate.
After the initial analyses, the order of the nucleotides in every simulated sequence was completely randomized to create random sequences with identical nucleotide content as the simulated sequences. The random sequences were also aligned using Clustal.
Authors' Contributions
MR designed, programmed, executed, and analyzed all parts of this study.
Acknowledgements
Many thanks to Sudhir Kumar, Sankar Subramanian, Arndt von Haeseler, and anonymous reviewers for comments on earlier versions of this manuscript.
Figures and Tables
Figure 1 Relationship of alignment accuracy and sequence identity. (A) Proportion of sites correctly aligned versus true percent identity among the sequences. (B) Proportion of sites correctly aligned versus observed percent identity after alignment. (C) Observed versus true percent identity. JC = Jukes-Cantor model [37]; HKY = Hasegawa-Kishino-Yano model [38]; HKY + Γ = Hasegawa-Kishino-Yano plus gamma-distributed rates model. All points represent the average of 1000 simulation replicates.
Figure 2 Relationships among true and estimated evolutionary distances. True distances were measured using the appropriate substitution model. (A) True P-distance versus true evolutionary distance. (B) Estimated evolutionary distance from aligned data versus true evolutionary distance.
Figure 3 Effect of alignment error on evolutionary distance estimation. Relative error in evolutionary distance is measured as the absolute difference between the distance estimated from the true alignment and the distance estimated from the observed alignment, divided by the distance estimate from the true alignment. (A) Relative error in evolutionary distance versus proportion of correctly aligned sites. (B) Relative error in evolutionary distance versus true percent identity.
Figure 4 Effects of parameter changes on alignment accuracy and relative error in evolutionary distance estimation. Ordinate axes are scaled to match Figure 3. Alignment accuracy (A, C, E, G, &I) and evolutionary distance estimation (B, D, F, H, &J). (A &B) Effect of initial sequence length. (C &D) Effect of mean insertion and deletion size. (E &F) Effect of insertion and deletion rate. (G &H) Effect of intersite rate variation. (I &J). Effect of nucleotide frequency bias (G+C content). Error bars represent ± one standard deviation. Black and white points represent HKY simulations with expected distances of 0.5 and 1.0, respectively.
Table 1 Summary of all simulation conditions
Model G+C% Initial # Sites κ α Insertion/Deletion Rate Mean Indel Size True distances simulated
JC 0.5 1000 n/a n/a 100/40 4 0.02, 0.04, 0.06, 0.08, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0, 1.2, 1.4, 1.6, 1.8, 2.0
HKY 0.6 1000 3.6 n/a 100/40 4 0.02, 0.04, 0.06, 0.08, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0, 1.2, 1.4, 1.6, 1.8, 2.0
HKY + Γ 0.6 1000 3.6 1.0 100/40 4 0.02, 0.04, 0.06, 0.08, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0, 1.2, 1.4, 1.6, 1.8, 2.0
HKY 0.6 100 3.6 n/a 100/40 4 0.5, 1.0
HKY 0.6 200 3.6 n/a 100/40 4 0.5, 1.0
HKY 0.6 300 3.6 n/a 100/40 4 0.5, 1.0
HKY 0.6 400 3.6 n/a 100/40 4 0.5, 1.0
HKY 0.6 500 3.6 n/a 100/40 4 0.5, 1.0
HKY 0.6 1500 3.6 n/a 100/40 4 0.5, 1.0
HKY 0.6 2000 3.6 n/a 100/40 4 0.5, 1.0
HKY 0.6 5000 3.6 n/a 100/40 4 0.5, 1.0
HKY 0.6 10000 3.6 n/a 100/40 4 0.5, 1.0
HKY 0.6 1000 3.6 n/a 100/40 2 0.5, 1.0
HKY 0.6 1000 3.6 n/a 100/40 6 0.5, 1.0
HKY 0.6 1000 3.6 n/a 100/40 8 0.5, 1.0
HKY 0.6 1000 3.6 n/a 100/40 10 0.5, 1.0
HKY 0.6 1000 3.6 n/a 200/80 4 0.5, 1.0
HKY 0.6 1000 3.6 n/a 150/60 4 0.5, 1.0
HKY 0.6 1000 3.6 n/a 75/30 4 0.5, 1.0
HKY 0.6 1000 3.6 n/a 50/20 4 0.5, 1.0
HKY 0.7 1000 3.6 n/a 100/40 4 0.5, 1.0
HKY 0.8 1000 3.6 n/a 100/40 4 0.5, 1.0
HKY 0.9 1000 3.6 n/a 100/40 4 0.5, 1.0
HKY + Γ 0.6 1000 3.6 0.25 100/40 4 0.5, 1.0
HKY + Γ 0.6 1000 3.6 0.5 100/40 4 0.5, 1.0
κ is the transition/transversion bias. α is the shape parameter for Γ-distributed intersite rate variation. Insertion/Deletion Rate is relative to the point mutation rate, i.e., a rate of 100/40 indicates 1 insertion every 100 point mutations and 1 deletion every 40 point mutations. Each simulation condition was replicated 1000 times.
==== Refs
Saitou N Nei M The neighbor-joining method: A new method for reconstructing phylogenetic trees Mol Biol Evol 1987 4 406 425 3447015
Edwards AWF Cavalli-Sforza LL The reconstruction of evolution Heredity 1963 18 553
Kumar S Hedges SB A molecular timescale for vertebrate evolution Nature 1998 392 917 920 9582070
Hedges SB Kumar S Genomic clocks and evolutionary timescales Trends Genet 2003 19 200 206 12683973
Clark AG Glanowski S Nielsen R Thomas PD Kejariwal A Todd MA Tanenbaum DM Civello D Lu F Murphy B Ferriera S Wang G Zheng X White TJ Sninsky JJ Adams MD Cargill M Inferring nonneutral evolution from human-chimp-mouse orthologous gene trios Science 2003 302 1960 1963 14671302
Hare MP Palumbi SR High intron sequence conservation across three mammalian orders suggests functional constraints Mol Biol Evol 2003 20 969 978 12716984
Thomas JW Touchman JW Blakesley RW Bouffard GG Beckstrom-Sternberg SM Margulies EH Blanchette M Siepel AC Thomas PJ McDowell JC Maskeri B Hansen NF Schwartz MS Weber RJ Kent WJ Karolchik D Bruen TC Bevan R Cutler DJ Schwartz S Elnitski L Idol JR Prasad AB Lee-Lin SQ Maduro VVB Summers TJ Portnoy ME Dietrich NL Akhter N Ayele K Benjamin B Cariaga K Brinkley CP Brooks SY Granite S Guan X Gupta J Haghighi P Ho SL Huang MC Karlins E Laric PL Legaspi R Lim MJ Maduro QL Masiello CA Mastrian SD McCloskey JC Pearson R Stantripop S Tiongson EE Tran JT Tsurgeon C Vogt JL Walker MA Wetherby KD Wiggins LS Young AC Zhang LH Osoegawa K Zhu B Zhao B Shu CL De Jong PJ Lawrence CE Smit AF Chakravarti A Haussler D Green P Miller W Green ED Comparative analyses of multi-species sequences from targeted genomic regions Nature 2003 424 788 793 12917688
Tamura K Kumar S Evolutionary distance estimation under heterogeneous substitution pattern among lineages Mol Biol Evol 2002 19 1727 1736 12270899
Nei M Kumar S Molecular Evolution and Phylogenetics 2000 Oxford: Oxford University Press
Fleißbner R Metzler D von Haeseler A Bornberg-Bauer E, Rost U, Stoye J, Vingron M Can one estimate distances from pairwise sequence alignments? Proceedings of the German Conference on Bioinformatics 2000 Berlin: Logos Verlag 89 95
Fleißner R Sequence alignment and phylogenetic inference PhD Thesis 2003 Düsseldorf: Heinrich-Heine-Universität Düsseldorf
Hein J Wiuf C Knudsen B Møller MB Wibling G Statistical alignment: Computational properties, homology testing and goodness-of-fit J Mol Biol 2000 302 265 279 10964574
Lecompte O Thompson JD Plewniak F Thierry J-C Poch O Multiple alignment of complete sequences (MACS) in the post-genomic era Gene 2001 270 17 30 11403999
Duret L Abdeddaim S Higgins D, Taylor W Multiple alignments for structrual, functional, or phylogenetic analyses of homologous sequences Bioinformatics: Sequence, Structure, and Databanks 2000 Oxford: Oxford University Press 51 76
Altschul SF Gish W Doolittle RF Local alignment statistics Methods in Enzymology: Computer Methods for Macromolecular Sequence Analysis 1996 266 San Diego: Academic Press 460 480
Frith MC Hansen U Spouge JL Weng Z Finding functional sequence elements by multiple local alignment Nucleic Acids Res 2004 32 189 200 14704356
Margulies EH Blanchette M Haussler D Green ED Identification and characterization of multi-species conserved sequences Genome Res 2003 13 2507 2518 14656959
Xia XH Xie Z Kjer KM 18S ribosomal RNA and tetrapod phylogeny Syst Biol 2003 52 283 295 12775520
Cammarano P Creti R Sanangelantoni AM Palm P The Archaea monophyly issue: A phylogeny of translational elongation factor G(2) sequences inferred from an optimized selection of alignment positions J Mol Evol 1999 49 524 537 10486009
Kjer KM Aligned 18S and insect phylogeny Syst Biol 2004 53 506 514 15503677
Kjer KM Use of rRNA secondary structure in phylogenetic studies to identify homologous positions: An example of alignment and data presentation fromt he frogs Mol Phylogenet Evol 1995 4 314 330 8845967
Titus T Frost DR Molecular homology assessment and phylogeny in the lizard family Opluridae (Squamata: Iguania) Mol Phylogenet Evol 1996 6 49 62 8812305
Morrison DA Ellis JT Effects of nucleotide sequence alignment on phylogeny estimation: A case study of 18S rDNAs of Acpiocomplexa Mol Biol Evol 1997 14 428 441 9100373
Hwang UW Kiim W Tautz D Friedrich M Molecular phylogenetics at the Felsenstein zone: Approaching the Strepsipera probelm using 5.8S and 28S rDNA sequences Mol Phylogenet Evol 1998 9 470 480 9667995
Shabalina S Kondrashov AS Pattern of selective constraint in C. elegans and C. briggsae genomes Genetical Research 1999 74 23 30 10505405
Pollard DA Bergman CM Stoye J Celniker SE Eisen MB Benchmarking tools for the alignment of functional noncoding DNA BMC Bioinformatics 2004 5 6 14736341
Metzler D Fleißbner R Wakolbinger A von Haeseler A Assessing variability by joint sampling of alignments and mutation rates J Mol Evol 2001 53 660 669 11677626
Needleman SB Wunsch CD A general method applicable to the search for similarities in the amino acid sequence of two proteins J Mol Biol 1970 48 443 453 5420325
Keightley PD Johnson T MCALIGN: Stochastic alignment of noncoding DNA sequences based on an evolutionary model of sequence evolution Genome Res 2004 14 442 450 14993209
Holmes I Bruno WJ Evolutionary HMMs: A Bayesian approach to multiple alignment Bioinformatics 2001 17 803 820 11590097
Thorne JL Kishino H Felsenstein J Inching toward reality: An improved likelihood model of sequence evolution J Mol Evol 1992 34 3 16 1556741
Thorne JL Kishino H Felsenstein J An evolutionary model for maximul likelihood alignment of DNA sequences J Mol Evol 1991 33 114 124 1920447
Schwartz S Kent WJ Smit A Zhang Z Baerstch R Hardison RC Haussler D Miller W Human-Mouse alignments with BLASTZ Genome Res 2003 13 103 107 12529312
Morgenstern B DIALIGN 2: Improvement of the segment-to-segment approach to multiple sequence alignment Bioinformatics 1999 15 211 218 10222408
Miller MP Kumar S Understanding human disease mutations through the use of interspecific genetic variation Human Molecular Genetics 2001 10 2319 2328 11689479
Miller MP Parker JD Rissing SW Kumar S Quantifying the intragenic distribution of disease mutations Annals of Human Genetics 2003 67 567 579 14641244
Jukes TH Cantor CR Munro HN Evolution of protein molecules Mammalian Protein Metabolism 1969 New York: Academic Press 21 132
Hasegawa M Kishino H Yano T Dating of the human-ape splitting by a molecular clock of mitochondrial DNA J Mol Evol 1985 22 160 174 3934395
Ophir R Graur D Patterns and rates of indel evolution in processed pseudogenes from humans and murids Gene 1997 205 191 202 9461394
Sundström H Webster MT Ellegren H Is the rate of insertion and deletion mutation male baised?: Molecular evolutionary analysis of avian and primate sex chromosome sequences Genetics 2003 164 259 268 12750337
Rosenberg MS Subramanian S Kumar S Patterns of transitional mutation biases within and among mammalian genomes Mol Biol Evol 2003 20 988 993 12716982
Thompson JD Higgins DG Gibson TJ CLUSTAL W: Improving the sensitivity of progressive multiple sequence alignment through sequence weighting, positions-specific gap penalties and weight matrix choice Nucleic Acids Res 1994 22 4673 4680 7984417
Thompson JD Plewniak F Poch O A comprehensive comparison of multiple sequence alignment programs Nucleic Acids Res 1999 27 2682 2690 10373585
Bray N Dubchak I Pachter L AVID: A global alignment program Genome Res 2003 13 97 102 12529311
Brudno M Do CB Cooper GM Kim MF Davydov E Green ED Sidow A Batzoglou S LAGAN and Multi-LAGAN: Efficient tools for large-scale multiple alignment of genomic DNA Genome Res 2003 13 721 731 12654723
Hickson RE Simon C Perrey SW The performance of several multiple-sequence alignment programs in relation to secondary-structure features for an rRNA sequence Mol Biol Evol 2000 17 530 539 10742045
Tamura K Nei M Estimation of the number of nucleotide substitutions in the control region of mitochondrial DNA in humans and chimpanzees Mol Biol Evol 1993 10 512 526 8336541
| 15840174 | PMC1087827 | CC BY | 2021-01-04 16:02:47 | no | BMC Bioinformatics. 2005 Apr 19; 6:102 | utf-8 | BMC Bioinformatics | 2,005 | 10.1186/1471-2105-6-102 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1041584768110.1186/1471-2105-6-104Research ArticleA method for the prediction of GPCRs coupling specificity to G-proteins using refined profile Hidden Markov Models Sgourakis Nikolaos G [email protected] Pantelis G [email protected] Panagiotis K [email protected] Stavros J [email protected] Department of Cell Biology and Biophysics, Faculty of Biology, University of Athens, Athens 157 01, Greece2005 22 4 2005 6 104 104 1 11 2004 22 4 2005 Copyright © 2005 Sgourakis 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
G- Protein coupled receptors (GPCRs) comprise the largest group of eukaryotic cell surface receptors with great pharmacological interest. A broad range of native ligands interact and activate GPCRs, leading to signal transduction within cells. Most of these responses are mediated through the interaction of GPCRs with heterotrimeric GTP-binding proteins (G-proteins). Due to the information explosion in biological sequence databases, the development of software algorithms that could predict properties of GPCRs is important. Experimental data reported in the literature suggest that heterotrimeric G-proteins interact with parts of the activated receptor at the transmembrane helix-intracellular loop interface. Utilizing this information and membrane topology information, we have developed an intensive exploratory approach to generate a refined library of statistical models (Hidden Markov Models) that predict the coupling preference of GPCRs to heterotrimeric G-proteins. The method predicts the coupling preferences of GPCRs to Gs, Gi/o and Gq/11, but not G12/13 subfamilies.
Results
Using a dataset of 282 GPCR sequences of known coupling preference to G-proteins and adopting a five-fold cross-validation procedure, the method yielded an 89.7% correct classification rate. In a validation set comprised of all receptor sequences that are species homologues to GPCRs with known coupling preferences, excluding the sequences used to train the models, our method yields a correct classification rate of 91.0%. Furthermore, promiscuous coupling properties were correctly predicted for 6 of the 24 GPCRs that are known to interact with more than one subfamily of G-proteins.
Conclusion
Our method demonstrates high correct classification rate. Unlike previously published methods performing the same task, it does not require any transmembrane topology prediction in a preceding step. A web-server for the prediction of GPCRs coupling specificity to G-proteins available for non-commercial users is located at .
==== Body
Background
G-protein coupled receptors are important receivers of information input to eukaryotic cells. They share a common fold of seven transmembrane helices arranged as a seven α-helix bundle, as confirmed by analysis of the crystal structure of Rhodopsin [1] that has been extensively used as template for homology-based modeling of GPCRs [2-4]. A collection of messages of extreme diversity including photons and native agonists, such as ions, odorants and pheromones, amino acids, nucleotides, peptides, biogenic amines, prostaglandines and glycoprotein hormones [5] interact with different extracellular and/or transmembrane domains of GPCRs, in order to convey their messages to the interior of the cell [2,6]. Based primarily on shared sequence motifs, six distinct families of GPCRs are traditionally defined: A, B, C, D, E and the frizzled/smoothened family, as summarized in the GPCRDB classification scheme [7]. Various methods have been deployed for higher-level classification of GPCRs including profile Hidden Markov Models [8,9], support vector machines [10] and Position Specific Scoring Matrices [11].
The physiological response of the interaction between a GPCR and one of its ligands is judged by the subset of the inactive heterotrimeric (αβγ) G-proteins within the cell that interact with the activated receptor complex, although many receptors mediate their actions through G-protein independent signaling pathways [2]. Different agonists may stabilize complexes of GPCRs with G-proteins belonging to different subfamilies (Gs, Gi/o, Gq/11 or G12/13) resulting in the activation of different signaling pathways [12].
G-proteins are heterotrimeric complexes, named after their α subunits. On a basis of sequence identity, at least 16 discrete α subunits have been identified and classified into four subfamilies: Gs and Gi/o, which stimulate and inhibit respectively adenylate cyclase, Gq/11 which stimulate phospholipase C, and the less characterized G12/13 subfamily that activate the Na+/H+ exchanger pathway [13-17]. We should mention at this point, that in the gpDB classification [18], the term "families" has been reserved for this level of hierarchy of G proteins, however hereinafter we will use the term "subfamilies" instead.
Agonist binding to GPCRs leads to association of the heterotrimeric G-protein with the receptor, which triggers the exchange of the guanosine diphosphate (GDP) bound on the α-subunit of the G-protein with guanosine triphosphate (GTP). These events promote the dissociation of the α subunit of the G-protein from the receptor and the βγ complex. The dissociated subunits can activate or inhibit several effector proteins, such as adenylyl cyclase 1–9, PLCβ 1–4, tyrosine kinases, ion channels and molecules of the mitogen-activated protein kinase pathway, resulting in a variety of cellular functions that depend on the biological specificity of the dissociated subunits [17,19]. G-protein α subunits possess an intrinsic GTPase activity, which enables them to act as time switches: Hydrolysis of the bound GTP to GDP promotes the re-association of the α subunit with the βγ dimer and renders the G-protein in an inactive form.
Due to the lack of structural data for activated GPCR complexes, several complementary approaches have been used to decipher the molecular events leading to G-protein activation, and to identify the regions that determine the coupling specificity of a GPCR to a subset of the pool of intracellular G-proteins. These biochemical approaches, that were focused mainly on A GPCRs, include site-directed mutagenesis studies [20], chimeric receptor engineering [21,22], the use of synthetic peptides to mimic the GPCR regions that activate G-proteins [23] and antibodies to neutralize GPCR binding sites on the G-proteins [24,25]. These studies revealed the major role of GPCR intracellular loops, especially the second and third, and the C-terminal region, as the main determinants of GPCR coupling specificity. Furthermore, structural data from high resolution X-ray diffraction of the light-sensing GPCR rhodopsin, as well as complementary methods (Nuclear Magnetic Resonance Spectroscopy, Electron Spin Resonance Spectroscopy, protein engineering, amino acid fluorescent replacement) [26-28] have indicated that ligand binding induces large conformational changes. These conformational changes reveal GPCR regions buried within the membrane which could interact with the G-protein [5]. Through a combination of entropy variability plots and correlated mutation analysis, key residues for a variety of GPCR functions, including coupling to G-proteins, can be identified and a mechanism of GPCR activation has been proposed [29-31].
Due to their role as information receivers of eukaryotic cells, GPCRs are involved in many pathophysiological responses. They comprise attractive drug targets for a variety of diseases, including cancer [32], Alzheimer's syndrome [33] and AIDS [34]. Indeed, over 50% of all prescribed drugs target on GPCRs [35]. Furthermore, the information explosion in biological sequence databases has resulted in many GPCR entries of unknown ligand binding properties, known as orphan receptors. In order to screen these orphan receptors with libraries of potential ligands, researchers must be able to assay the GPCR-ligand interaction through a downstream event. Such events are transcription of a reporter gene or rise in second messenger concentration, which is dependent on the interaction of the GPCR under study with members of a specific G-protein subfamily. Thus, knowing or being able to predict, the coupling specificity of orphan GPCRs to G-protein subfamilies, is essential for choosing the appropriate cell lines for heterologous expression and any further in vitro and in vivo studies of potential drug targets [36]. Meanwhile, a dataset of GPCRs of known coupling specificity exists [37], large enough to guide an in silico database mining approach that could aid further in vivo GPCR research. Furthermore, in a work published recently, many GPCRs and their interactions with G-proteins have been summarized in the gpDB system [18].
As in every biological interaction, the specificity of GPCR coupling to specific G-proteins is determined by structural components located on contact regions of the molecules. Since the three-dimensional architecture of a protein is encoded in protein sequence, GPCR coupling specificity could be defined by sequence alone. However, GPCRs with low sequence similarity may couple to members of the same subfamily of G-proteins, while members of the same GPCR subfamilies often couple to members of distinct G-protein subfamilies [38]. In addition, GPCR coupling is not a one-by-one function since many GPCRs, known as promiscuous GPCRs, have been proven to couple to members of more than one G-protein subfamilies. Due to these limitations, GPCR coupling specificity prediction in one step using sequence comparison methods such as the BLAST [39] or CLUSTALW [40] algorithms is insufficient [36]. However a weak sequence signal can be detected among receptor subfamilies where G- protein selectivity was a recent evolutionary process, such as the biogenic amines receptors [41].
Previous computational methods of GPCR coupling specificity to G-protein subfamilies have been applied on a priori selected intracellular regions of GPCR sequences. A Naive Bayes model [42] yields a 72% correct classification rate, while a data-mining approach that combined pattern discovery with membrane topology prediction [43] has also been applied in an effort to model GPCR regions that determine coupling specificity. However, previous approaches are either context-dependent on the a priori knowledge that GPCR coupling specificity is governed by the entire intracellular regions sequence or limited by the non-probabilistic nature and limited descriptive power of patterns as regular expressions, that cannot implement weights to different sequence variation. The approach of this study is exploratory regarding the length and localization of the coupling determining regions among the intracellular regions sequences and recruits profile Hidden Markov Models (pHMMs) as highly discriminative models of biological sequences that have a formal probabilistic basis [44]. The results obtained by this method, presented below, justify the chosen approach.
Results and discussion
Our primary aim was to develop a wide-range predictive system that can be applied with the same discriminative power globally, for all three main GPCR coupling groups, being also able to model promiscuous receptor coupling. Our method proved to be self-consistent: Using a set of 282 GPCR sequences of experimentally identified coupling properties, according to the Trends in Pharmacological Sciences nomenclature supplement of receptors and ion channels (TiPS) [37], that were used to train the models and adopting a five-fold cross-validation procedure, the methods yielded a 89.7% correct classification rate. When tested in 479 sequences of GPCRs (retrieved also from the UniProt database [45]) that are homologous to the sequences used to train the models and whose coupling properties are also summarized in [37], at a subtype level, our method yields a 91.0% correct classification rate (Table 1). Finally, the method predicts correctly the coupling specificity of 25 out of 30 GPCRs derived from the gpDB database [18] that were not included in [37] (Table 2).
In order to assess the efficiency of the same method trained on a smaller and non-redundant dataset, the same procedure was applied to a dataset containing only the human GPCRs in the original training set. Alternative pHMMs were generated and integrated into a second predictive system that proved to be also self-consistent. On this human-only dataset, correct classification rate in a five-fold cross-validation, is 86% (data not shown). When these models were applied to the 479 sequences of the validation set, the correct classification rate was 88.9%, showing an insignificant decrease, as one would expect for a non-overfitted method. Additionally, when the model that was trained on human sequences, was applied to the remaining 178 non-human sequences derived from [37], yields also a high correct classification rate of 88.8%.
Due to insufficient experimental data, resulting in uncertainty about whether or not most receptors that are known to couple with a specific G-protein group can couple with G-proteins of another subfamily under different physiological conditions, we cannot estimate whether all of the promiscuous predictions are correct or not. For instance, a GPCR that is reported to couple only to G-proteins members of Gi/o subfamily, may proved that couples also to members of Gs subfamily. It is also well-known that the same GPCR may also couple to different G-protein subfamilies in different heterogenous expression systems. Promiscuous coupling was correctly predicted for 6 out of 24 GPCRs of known promiscuous coupling properties according to information in [37], as one can observe in Table 3. We did not attempt to train any pHMMs from sequences that have been proven to be promiscuous, in order to avoid unnecessary complexity and unequal distribution of the training set to the three major coupling groups of GPCRs.
The main reason that no pHMMs have been constructed that indicate coupling to G12/13 proteins is the limited amount of data available for the coupling properties of this subfamily of G-proteins. For this reason, this feature is not provided by any of the already published methods that perform the same task. Furthermore, to the knowledge of the authors no promiscuous GPCRs are included in the training set (i.e. GPCRs that couple to members from multiple subfamilies of G-proteins), and no receptors that preferentially couple only to members of the G12/13 subfamily have been identified [2]. Therefore, constructing pHMMs that classify G12/13 coupled GPCRs with high discriminative power, at this moment, is practically impossible. Once larger datasets have been established in the future, promiscuous receptors could be included in the training set, allowing predictions for G12/13 coupled receptors.
Our exploratory approach resulted in the discovery of sub-regions within the intracellular GPCR domains that play a key role in determining GPCR coupling specificity to G-proteins. The contribution of these regions to the overall coupling scheme of GPCRs could arise through short-range protein-protein interactions with their structural counterparts in G-proteins, that is, through intermolecular stabilizing interactions that enable several regions of the GPCR molecule to interact with G-proteins. The conformation of the intracellular regions of GPCRs is regulated by intramolecular interactions between the intracellular segments [38]. Furthermore, each query against the refined library of pHMMs reveals regions of high identity to the profiles, if such exist in the target sequence. Residues in these identified intracellular regions could be targeted for site-directed mutagenesis approaches in order to elucidate the structural features of GPCR – G-protein coupling.
Our method can only predict the potential of interaction between a GPCR and a G-protein subfamily, since its only input is the GPCR sequence. Thus, common in vivo regulators of GPCR coupling specificity including mechanisms such as selective targeting of GPCRs to specific cell-membrane regions, post-translational modifications [46,47] or effects of accessory/scaffolding proteins interacting with GPCRs (reviewed in [2]) cannot be modeled by our prediction system. Also, GPCR homo- or hetero-dimerization, that appears to be a common feature of many GPCRs, necessary for G-protein activation [48-50] cannot be directly included in our prediction system.
pHMMs derived from this study have been trained to model sub-regions within GPCR intracellular domains rather than entire GPCR sequences. The a priori knowledge that a query sequence belongs to a GPCR would be valuable in enhancing the predictive power of the method. When the method is applied to the non-GPCR receptor and the globular protein non-redundant test sets, it produces false positives with a rate 19.2% and 6.4% respectively. However filtering the query sequences, by using 7-transmembrane domain pHMMs derived from the Pfam database Version 14.0 [51] in a preceding step, diminishes completely the above false positive results without affecting the overall sensitivity of the method. All six pHMMs for 7-transmembrane receptors contained in the Pfam database Version 14.0 have been integrated into our publicly available method. In conclusion, the method could effectively be used in combination with existing 7-transmembrane receptor predictive systems for genome-wide applications.
Compared to other previously published methods, performing the same task, our method does not only perform significantly better in terms of overall accuracy, but also employs additional superior features. Firstly, it does not rely on the identification of intracellular loops as does the Naive Bayes method in [42]. Our method was trained using the annotations for the transmembrane regions (which in most cases come from prediction methods) but in the testing phase no such information is required, thus it operates using as input solely the sequence. Compared to the pattern discovery method of [43], our method uses a more sophisticated scheme for whole-sequence scoring that has a formal probabilistic interpretation. We should note, however, that most of the patterns discovered by [43] were captured by our pHMMs (Figure 1), but in a more mathematically sound and exploitable manner. In addition, in [43] no overall measures of accuracy were reported in order to assess a fair comparison. Finally, our method is the only method reported until now, that is publicly available through a web-server. At the URL: , the user may submit a sequence in Fasta format, and receive the prediction. The method is rather fast, producing a self-explanatory output, and, thus it may be used by both molecular biologists requesting information for a single GPCR, and by bioinformaticians performing large-scale computational analyses.
At the final stages of preparation of this manuscript, another method developed independently by Sreekumar and coworkers, was published [52], which uses also pHMMs. However, the method of Sreekumar and coworkers, does not treat the multiple intracellular loops of a given GPCR independently, but instead it concatenates them into a single sequence. These concatenated sequences are then used to build pHMMs with the HMMER package. Although, the method performs very well as reported by the authors (they claim a 99% correct classification rate in a cross validation test), there are some severe disadvantages arising from the aforementioned strategy: With this method, to test a newly found protein, one has to perform predictions on the GPCR regarding its transmembrane topology, extract the intracellular loops and concatenate them into a single sequence. This adds another source of error, originating from the prediction errors of the transmembrane topology prediction algorithm. Having in mind, that to date, even the best topology prediction algorithms, predict correctly the full topology of a protein with an accuracy of no more than 75% [53,54], this will further reduce the performance of the method. We should also note that regarding GPCRs, the most accurate predictors fail to even predict seven transmembrane segments for more than 15% of the presented examples [54]. Furthermore, the method does not control appropriately the level of false positives, since it was not tested on non-GPCR sequences. On the contrary, the method proposed in this work, although it uses essentially the same principles in extracting the loop regions, it treats them independently using the Qfast algorithm, and thus, in the prediction phase, no a-priori knowledge of loops and transmembrane topology is needed. In addition to that, the rate of false positive predictions is controlled, providing us a confidence about the validity of the results. Last, and perhaps more important, our method is the only one until now that is fully automated and publicly available via a web-server.
Conclusion
We applied here, a data-mining exploratory approach combined with the high discriminative power of profile Hidden Markov Models (pHMMs), to generate a system that predicts GPCR coupling specificity to the three main subfamilies of G-proteins (Gi/o, Gq/11 and Gs), based solely on the information included in the protein sequence. We report superior correct classification rate compared to other previously published methods, and we have created a web-server, running the application, freely available for academic users (Commercial users should contact Professor S. J. Hamodrakas to obtain a licence). At present, this is the only web-based server for prediction of GPCRs coupling to G-proteins. Expanding this information to characterize the coupling properties for thousands of orphan GPCRs in large-scale proteome annotation studies, our understanding of receptor signaling pathways might improve and new targets for drug research may be uncovered. Future studies, utilizing larger representative training sets of GPCRs with known coupling specificity to G-proteins, and more advanced algorithmic techniques are needed in order to increase the accuracy of the prediction method, as also as to handle more efficiently the promiscuity in preferential coupling of GPCRs to G-proteins. This way, we may also be capable of predicting the coupling of GPCRs to G-proteins members of the G12/13 subfamily, a feature neither addressed in this study, nor in previously published methods.
Methods
Datasets
Our primary training dataset consists of 282 sequences of GPCRs of known coupling properties to G-proteins (120 Gi/o, 94 Gq/11 and 68 Gs) according to the Trends in Pharmacological Sciences 2000 Receptor and Ion Channel nomenclature supplemement [37]. All sequences in the training dataset were of GPCRs with non-promiscuous coupling according to [36] and were retrieved from the Uni Prot 1.10 database [45], excluding fragments. Based on their coupling preference, they were grouped into Gi/o, Gq/11 or Gs coupled receptors. The Uniprot Accession numbers of the proteins in the training set can be found in our web-page . Moreover, an alternative non-redundant dataset comprised only of the 104 human counterparts of GPCRs in the original dataset was used to train the method. This was done in order to investigate the effect of redundancy posed by homologous sequences. A validation set was also generated, including 479 GPCR species homologues of the receptor subtypes with known coupling specificity according to [37] (256 Gi/o, 102 Gq/11 and 121 Gs). Finally, the method was also validated on an independent set, composed of GPCRs, belonging to different subtypes with known coupling properties extracted from the gpDB database [18] that are not included in the training set [37].
As mentioned above, a sufficient amount of experimental data signifies the role of GPCR intracellular regions (the three intracellular loops and the carboxyl terminal region) and the membrane proximal intracellular extensions of transmembrane α-helices (approximately 1.5 turns) as the main regions of interaction between the G-proteins and the activated receptor complex [5]. Based on this experimentally derived information as well as membrane topology information derived from the UniProt annotation of each entry (in the "FT TRANSMEM" lines), we adapted our primary dataset, extracting sequence regions that corresponded to intracellular regions or transmembrane regions with intracellular proximity, spanning for approximately 7 residues within the cell membrane.
In order to investigate the performance of the method when applied to non-GPCR sequences, we used two alternative datasets. The first dataset includes a total of 1361 non-GPCR transmembrane receptors, whereas the second includes 1239 non-homologous globular proteins with structures known at atomic resolution [8].
Data mining: Generation and evaluation of HMMs
A multiple alignment was generated for each group of intracellular sequence regions derived from GPCRs with same coupling preference using the ClustalX package [55]. Pairwise alignment scoring parameters were set as: BLOSUM 30 substitution matrix, Gap-opening penalty of 10.00 and gap extension penalty of 0.10. Multiple alignment parameters were set as BLOSUM 30 series substitution matrix, gap-opening penalty of 10.00 and gap extension penalty of 0.20. Multiple sequence alignments were then scanned for low entropy regions of high scoring alignment rows. Thus, the training dataset was further diminished to low-entropy sequence regions with a sequence identity criterion. The resulting multiple alignment rows were then used to generate a library of HMMs in an explorative way: for a given multiple alignment low entropy block starting from every offset and with any window of seven or more alignment rows a HMM was constructed. Thus, a low entropy block of n alignment rows generates
potential alignments, where w is the window length. This analysis yielded a total of 6149 HMMs that were tested on the fly with the hmmsearch program against the training set, in order to compare the discriminative power among alternative HMMs. As an estimator of the discriminative power of HMMs we calculated the Coverage of the results, i.e. the percentage of positives scoring an e-value lower than the lowest e-value of the negatives. This critical e-value corresponds to the noise cutoff of Pfam entries [51]. In order to compare Coverage values between HMMs derived from different coupling groups, we calculated the p-value of Coverage as a random variable (probability of a model derived from alignment rows of random sequences scoring a Coverage greater or equal to a specific observed Coverage), as follows:
In a set of n sequences containing κ positives, the probability of choosing x positives before the first negative, without reset, equals to:
where f(x) is the probability function of the negative hypergeometric distribution. The cumulative probability function, i.e. the probability of choosing x or less positives before the first negative without reset is:
Thus, the probability of choosing × or more positives before the first negative (i.e. the p-value of the test) equals to 1 - F(x).
Based mainly on Coverage measurements and their p-values, we discovered HMMs from several sub-regions within the loops that show up to 12-fold increase of Coverage, in comparison to models derived from the entire loop sequences. Our exploratory approach resulted in 5 to 7 refined pHMMs for each one of the three groups of GPCRs in the dataset, as summarized in Table 4. The overall flow chart of the method, is presented in Figure 2.
Cutoff optimisation
Due to sequence similarity among intracellular regions of GPCRs of different coupling preferences, an hmmpfam query of the training set against the refined HMM library under a default e-value cutoff threshold, yielded artifact promiscuity in the predictions. HMMs derived from different intracellular regions show extensive variation in their discriminative power, as a consequence of different participation of different regions among the three groups of GPCRs in the overall coupling preference of the molecule. Our aim was not to exclude promiscuity from the predictive power of the method, so each HMM in the refined library was given a discrete cut off threshold resulting from ROC curve analysis after evaluation of the distribution of scores that corresponded to positive and negative targets of the training set. For each refined HMM of our library, we estimated and applied cutoff values that maximize the value:, where TP and FP are the percentages of positives and negatives respectively that score an e-value equal or less than a defined threshold at an hmmpfam query against the training set. This value is represented by the distance from the diagonal f(x) = x of a ROC curve, as presented in Figure 3. Once we had estimated the optimized cutoff thresholds for each HMM fingerprint, we combined the e-values of queries against HMMs that characterize the same GPCR coupling group using the QFAST algorithm [56]: , where p is the maximum product of n independent, uniform random variables and F(p) the combined p-value. In order to maximize the discriminative power of the combined predictions, a cutoff threshold was set for each one of the three e-value combinations, based mainly on the shape of the distribution of e-values scored by positives and negatives of the test set. A threshold was also set for the difference of combined e-values scored by the first and second match, expressed in logarithmic units, based on the combined e-values scored by HMMs characterizing different coupling groups in a hmmpfam search against a set of all 24 coupling-promiscuous GPCR sequences summarized in [37]. This estimation was based on the observation that the distributions of combined e-value differences, between the first two coupling predictions, from queries against promiscuous and non-promiscuous GPRCs are distinguishable when expressed in a logarithmic scale. Thus, alternative coupling groups are predicted as multiple combined e-value hits when querying the library against GPCR sequences of promiscuous coupling properties.
In order not to over-estimate the correct classification rate, a five-fold cross-validation procedure was adopted. Initially, the training set was randomly divided to five equally balanced sets. Afterwards, we trained a model, according to the above-mentioned procedure, using as training set the sequences in the four sub-sets, whereas the last sub-set was used for testing. This procedure was repeated five times, and the final results are the overall results obtained from the five sets.
Authors' contributions
NS performed the analysis, and implemented the algorithms and the web-interface. PB formulated the problem, collected the training and testing sets and designed the training scheme. PP implemented the Qfast algorithm and participated in the optimization procedure. SH coordinated and supervised the project suggesting innovative features. NS, PB and SH drafted the manuscript. All authors have read and accepted the manuscript.
Acknowledgements
PB was supported by a grant from the IRAKLEITOS fellowships program of the Greek Ministry of National Education, supporting basic research in the National and Kapodistrian University of Athens. We thank the University of Athens for financial support. The authors would like to thank the anonymous referee for his valuable criticism and comments on the manuscript.
Figures and Tables
Figure 1 Comparison of the methods. Coupling determining patterns discovered by Moller et al. [43], show redundancy in their targets, since different patterns may apply to the same sequences in an overlapping manner. In addition, there are loop sequences of major GPCR subfamilies not characterized by any pattern, possibly due to low sequence identity. In comparison to patterns, profile Hidden Markov models (pHMMs) provide a whole sequence scoring scheme. Sequence information contained in multiple patterns can be integrated in a single pHMM derived from a low entropy region of a multiple sequence alignment. Thus, every query sequence can be given a score that has a formal probabilistic interpretation.
Figure 2 Flow chart of the method. GPCR entries in UniProt of known, non-promiscuous coupling specificity to G-proteins summarized in [37] were used to extract intracellular regions sequences, based on membrane topology information of the UniProt annotation. ClustalX was used to generate multiple sequence alignments of intracellular regions from which low-entropy blocks were selected based on ClustalX row scores. For every begin and end row, within low-entropy regions, sub-alignments were extracted and profile Hidden Markov models (pHMMs) were built. The discriminative power of each pHMM was assayed, after an hmmsearch run against the training set. The most discriminative HMMs for each intracellular region were selected for each one of the three main coupling groups and appended in the refined library. E-value thresholds were then set for each pHMM included in the refined library. The reverse course is followed during a query against the library of refined pHMMs.
Figure 3 ROC curve analysis. ROC curve as an assay of discriminative power of a profile Hidden Markov models (pHMMs) that models a sub-region within the second intracellular loop of GPCRs that are known to couple with G-proteins of the Gs subfamily. The space under the ROC curve is a measure of the discriminative power of the model, while the distance of each point from the diagonal of the chart y = x is a measure of combined specificity and sensitivity of the model. The e-value that corresponds to the maximal distance from the diagonal spot is set as the threshold that discriminates positive from negative predictions in an hmmpfam run, regarding the selected pHMM. Similar charts were applied to optimize e-value cutoffs of all HMMs in the refined final library.
Table 1 Results of the cross-validation and independent set tests. A. Correct classification rate results obtained from the three main G-protein coupling groups, in a five-fold cross-validation procedure. The training set was randomly divided to five equally balanced sets. Afterwards, we trained a model using the sequences in the four sets whereas the last set was used for testing. This procedure was repeated five times. B. The library of refined profile Hidden Markov models (pHMMs) derived from the primary dataset of 282 GPCRs (see text) was tested against a validation set comprised of all GPCR sequences of subtypes with known coupling preference summarized in [37], excluding the sequences used to train the models (479 GPCRs in total). This independent test yielded 91% correct classification rate. Numbers in the diagonal of the charts represent true positive predictions. The total number of predictions for each group (row) is not equal to the total number of observations, since several GPCRs were not classified in any group.
A predicted
Five-fold cross-validation test Gi/o Gq/11 Gs total
observed Gi/o 109(90.8%) 1 1 120
Gq/11 2 78(82.9%) 2 94
Gs 0 0 66(97.1%) 68
111 79 69 253(89.7%)
B predicted
Validation test (479 GPCRs) Gi/o Gq/11 Gs total
observed Gi/o 233(91.4%) 16 4 256
Gq/11 9 90(88.2%) 2 102
Gs 6 2 113(93.4%) 121
248 108 119 436(91.0%)
Table 2 Prediction results for 30 different GPCR subtypes with known coupling properties extracted from the gpDB database [18] that are not included in the training set [37]. The annotation of GPCR coupling properties in gpDB is based on data from the scientific literature. According to the gpDB classification scheme, Ggust and Gt G-proteins belong to the Gi/o subfamily and Golf to the Gs subfamily. The GPCR sequences were extracted from gpDB and parsed into our prediction server .
GPCR Subtype Uniprot AC observed predicted
5-OXO-ETE receptor TG1019 Q8TDS5 Gi/o Gq/11
Allatostatin receptor Q9W4R0 Gi/o Gi/o Gq/11
Apelin receptor P35414 Gi/o Gi/o
Gonadotropin releasing hormone I receptor Q92644 Gq/11 Gi/o
Gonadotropin releasing hormone II receptor Q8TCX8 Gq/11 Gq/11
Gustatory receptor 43 Q9JHE2 Ggust Gi/o
Gustatory receptor GUST27 P34987 Gi/o Gi/o
Neuromedin U1 receptor Q9JIB2 Gq/11 Gq/11 Gi/o
Neuromedin U2 receptor Q9NRA6 Gq/11 Gq/11 Gi/o
Nicotinic Acid receptor Q8TDS4 Gi/o Gi/o
Olfactory receptor 10A7 Q96R19 Golf Gs
Blue opsin Q13877 Gt Gi/o
Orexin 1 receptor O43613 Gq/11 Gi/o Gq/11
Orexin 2 receptor O43614 Gq/11 Gi/o Gq/11
Platelet activating factor receptor P25105 Gi/o Gq/11 Gi/o
Prolactin-releasing peptide receptor O75194 Gq/11 Gi/o
Trace amine receptor TAR Q9P1P4 Gs Gs
Trace amine receptor TAR-1 Q96RJ0 Gs Gq/11 Gs Gi/o
Trace amine receptor TAR-2 Q923Y7 Gs Gs Gi/o
Trace amine receptor TAR-3 Q96RI9 Gs Gi/o Gs
Urotensin II receptor Q9UKP6 Gq/11 Gi/o
Gastric inhibitory peptide receptor Q9UPI1 Gs Gs
Glucagon Receptor P47871 Gs Gs
Glucagon like peptide receptor 2 O95838 Gs Gs
Ghrelin receptor Q96RJ7 Gq/11 Gq/11
Parathyroid hormone receptor 1 P49190 Gs Gs
Parathyroid hormone receptor 2 Q03431 Gs Gs
Secretin receptor Q13213 Gs Gs
Calcium-sensing receptor P41180 Gi/o Gq/11 Gi/o
Taste receptor TASTE_TB334 Q62942 Ggust Gi/o
Table 3 Promiscuous validation test. All 24 sequences of promiscuous coupled GPCR subtypes (as summarized in [37] and other sources [38]) were parsed in an hmmpfam query against our refined library of profile Hidden Markov models (pHMMs). The query sequences were not included in the training set. Promiscuous coupling was correctly predicted for 6 queries.
GPCR Subtype Uniprot AC observed predicted
Alpha-2A adrenergic receptor P08913 Gq/11 Gi/o Gs Gq/11 Gi/o Gs
AT1 angiotensin receptor P30556 Gq/11 Gi/o Gq/11 Gi/o
Beta-3 adrenenergic receptor P25962 Gi/o Gs Gq/11 Gio Gs
C3A anaphylatoxin chemotactic receptor Q16581 Gq/11 Gi/o Gi/o
Calcitonin receptor P30988 Gq/11 Gs Gs
Cholecystokinin type 1 receptor P32238 Gq/11 Gs Gq/11 Gi/o
Sphingosine 1-phosphate receptor 3 (EDG3) Q99500 Gq/11 Gi/o Gi/o
Lysophosphatidic acid receptor 2 (EDG4) Q9HBW0 Gq/11 Gi/o Gi/o
Endothelin B receptor O60883 Gq/11 Gi/o Gi/o
Endothelin A receptor P25101 Gq/11 Gs Gq/11
Follicle-stimulating hormone receptor Q95179 Gi/o Gs Gs
Galanin receptor type 2 O43603 Gq/11 Gi/o Gq/11 Gi/o
Leukotriene B4 receptor Q15722 Gq/11 Gi/o Gq/11 Gi/o
Lutropin-choriogonadotropic hormone receptor P22888 Gq/11 Gi/o Gs Gs
Neuromedin K receptor (NK-3) P29371 Gq/11 Gs Gq/11 Gi/o
Neuropeptide Y receptor type 1 O02813 Gq/11 Gi/o Gi/o
Oxytocin receptor P30559 Gq/11 Gi/o Gq/11 Gi/o
P2Y purinoceptor 11 Q96G91 Gq/11 Gs Gi/o
Platelet-activating factor receptor Q62035 Gq/11 Gi/o Gq/11 Gi/o
Prostaglandin E2 receptor EP3 P43115 Gq/11 Gi/o Gs Gi/o
Substance-K receptor (NK-2) P21452 Gq/11 Gs Gq/11 Gi/o
Substance-P receptor (NK-1) P25103 Gq/11 Gs Gq/11 Gi/o
Proteinase activated receptor 1 (thrombin receptor) P25116 Gq/11 Gi/o Gi/o
Thyrotropin receptor P47750 Gq/11 Gi/o Gs Gi/o Gs
Table 4 Fingerprint discovery results. Coverage values of profile Hidden Markov models (pHMMs) given after each alignment of entire loop sequence regions in comparison to maximized Coverage values of highly discriminative pHMMs derived from selected sub-alignments within low-entropy regions. pHMMs derived from sub-alignments showed up to 12-fold increase in their discriminative power, as measured by Coverage, in comparison to pHMMs that characterize the entire loop sequence alignments. Alternative alignment regions with the same discriminative power are separated by commas, double dots separate beginning and ending of sub-alignments used to generate HMMs. In the case of discovery of non-overlapping sequence fragments with high discriminative power (e.g. Gs loop1), separate pHMMs were generated and appended in the refined library of that group.
Coupling preference Gi/o Gq/11 Gs
Intracellular regions loop1 loop1 loop1
Whole sequence region 1..39/2.65 1..28/16.05 1..59/58.01
Non-overlapping most discriminative regions 3..24/19.86 13..25/30.53 7..19/63.35,43..51/48.09
loop2 loop2 loop2
1..50/19.20 1..42/0.70 1..39/12.21
10..18/32.78 14..34/19.85 18..25/54.19
loop3a loop3a loop3a
1..29/19.54 1..17/16.79 1..15/27.48
2..18/27.15 5..17/19.08 6..15/42.74
loop3b loop3b loop3b
1..26/6.63 1..21/3.82 1..15/20.61
14..21/23.51 8..19,9..20/12.21 1..15/20.61
loop3c loop3c loop3c
1..29/8.28 1..17/3.81 1..22/0.76
8..21/36.09 2..12/48.85 8..21,13..21/10.68
C-terminal C-terminal C-terminal
1..29/8.28 1..24/12.97 1..29/29.77
8..21/36.09 5..19/35.87 8..16/57.25
==== Refs
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
Kristiansen K Molecular mechanisms of ligand binding, signaling, and regulation within the superfamily of G-protein-coupled receptors: molecular modeling and mutagenesis approaches to receptor structure and function Pharmacol Ther 2004 103 21 80 15251227 10.1016/j.pharmthera.2004.05.002
Orry AJ Wallace BA Modeling and docking the endothelin G-protein-coupled receptor Biophys J 2000 79 3083 3094 11106614
Nikiforovich GV Galaktionov S Balodis J Marshall GR Novel approach to computer modeling of seven-helical transmembrane proteins: current progress in the test case of bacteriorhodopsin Acta Biochim Pol 2001 48 53 64 11440183
Gether U Uncovering molecular mechanisms involved in activation of G protein-coupled receptors Endocr Rev 2000 21 90 113 10696571 10.1210/er.21.1.90
Schwartz TW Locating ligand-binding sites in 7TM receptors by protein engineering Curr Opin Biotechnol 1994 5 434 444 7765177 10.1016/0958-1669(94)90054-X
Horn F Bettler E Oliveira L Campagne F Cohen FE Vriend G GPCRDB information system for G protein-coupled receptors Nucleic Acids Res 2003 31 294 297 12520006 10.1093/nar/gkg103
Papasaikas PK Bagos PG Litou ZI Hamodrakas SJ A novel method for GPCR recognition and family classification from sequence alone using signatures derived from profile hidden Markov models SAR QSAR Environ Res 2003 14 413 420 14758984 10.1080/10629360310001623999
Papasaikas PK Bagos PG Litou ZI Promponas VJ Hamodrakas SJ PRED-GPCR: GPCR recognition and family classification server Nucleic Acids Res 2004 32 W380 2 15215415
Karchin R Karplus K Haussler D Classifying G-protein coupled receptors with support vector machines Bioinformatics 2002 18 147 159 11836223 10.1093/bioinformatics/18.1.147
Attwood TK Croning MD Flower DR Lewis AP Mabey JE Scordis P Selley JN Wright W PRINTS-S: the database formerly known as PRINTS Nucleic Acids Res 2000 28 225 227 10592232 10.1093/nar/28.1.225
Kenakin T Ligand-selective receptor conformations revisited: the promise and the problem Trends Pharmacol Sci 2003 24 346 354 12871667 10.1016/S0165-6147(03)00167-6
Exton JH Role of G proteins in activation of phosphoinositide phospholipase C Adv Second Messenger Phosphoprotein Res 1993 28 65 72 8398419
Benjamin DR Markby DW Bourne HR Kuntz ID Solution structure of the GTPase activating domain of alpha s J Mol Biol 1995 254 681 691 7500342 10.1006/jmbi.1995.0647
Johnston CA Watts VJ Sensitization of adenylate cyclase: a general mechanism of neuroadaptation to persistent activation of Galpha(i/o)-coupled receptors? Life Sci 2003 73 2913 2925 14519441 10.1016/S0024-3205(03)00703-3
Kurose H Galpha12 and Galpha13 as key regulatory mediator in signal transduction Life Sci 2003 74 155 161 14607242 10.1016/j.lfs.2003.09.003
Cabrera-Vera TM Vanhauwe J Thomas TO Medkova M Preininger A Mazzoni MR Hamm HE Insights into G protein structure, function, and regulation Endocr Rev 2003 24 765 781 14671004 10.1210/er.2000-0026
Elefsinioti AL Bagos PG Spyropoulos IC Hamodrakas SJ A database for G proteins and their interaction with GPCRs BMC Bioinformatics 2004 5 208 15619328 10.1186/1471-2105-5-208
Pierce KL Premont RT Lefkowitz RJ Seven-transmembrane receptors Nat Rev Mol Cell Biol 2002 3 639 650 12209124 10.1038/nrm908
Blin N Yun J Wess J Mapping of single amino acid residues required for selective activation of Gq/11 by the m3 muscarinic acetylcholine receptor J Biol Chem 1995 270 17741 17748 7629074 10.1074/jbc.270.30.17741
Wess J Bonner TI Dorje F Brann MR Delineation of muscarinic receptor domains conferring selectivity of coupling to guanine nucleotide-binding proteins and second messengers Mol Pharmacol 1990 38 517 523 2172767
Hwa J Graham RM Perez DM Chimeras of alpha1-adrenergic receptor subtypes identify critical residues that modulate active state isomerization J Biol Chem 1996 271 7956 7964 8626475 10.1074/jbc.271.14.7956
Higashijima T Burnier J Ross EM Regulation of Gi and Go by mastoparan, related amphiphilic peptides, and hydrophobic amines. Mechanism and structural determinants of activity J Biol Chem 1990 265 14176 14186 2117607
Georgoussi Z Milligan G Zioudrou C Immunoprecipitation of opioid receptor-Go-protein complexes using specific GTP-binding-protein antisera Biochem J 1995 306 ( Pt 1) 71 75 7864831
Matesic DF Manning DR Luthin GR Tissue-dependent association of muscarinic acetylcholine receptors with guanine nucleotide-binding regulatory proteins Mol Pharmacol 1991 40 347 353 1716729
Farrens DL Altenbach C Yang K Hubbell WL Khorana HG Requirement of rigid-body motion of transmembrane helices for light activation of rhodopsin Science 1996 274 768 770 8864113 10.1126/science.274.5288.768
Turcatti G Nemeth K Edgerton MD Meseth U Talabot F Peitsch M Knowles J Vogel H Chollet A Probing the structure and function of the tachykinin neurokinin-2 receptor through biosynthetic incorporation of fluorescent amino acids at specific sites J Biol Chem 1996 271 19991 19998 8702716 10.1074/jbc.271.33.19991
Javitch JA Fu D Liapakis G Chen J Constitutive activation of the beta2 adrenergic receptor alters the orientation of its sixth membrane-spanning segment J Biol Chem 1997 272 18546 18549 9228019 10.1074/jbc.272.30.18546
Oliveira L Paiva AC Vriend G Correlated mutation analyses on very large sequence families Chembiochem 2002 3 1010 1017 12362367 10.1002/1439-7633(20021004)3:10<1010::AID-CBIC1010>3.0.CO;2-T
Oliveira L Paiva PB Paiva AC Vriend G Identification of functionally conserved residues with the use of entropy-variability plots Proteins 2003 52 544 552 12910454 10.1002/prot.10490
Oliveira L Paiva PB Paiva AC Vriend G Sequence analysis reveals how G protein-coupled receptors transduce the signal to the G protein Proteins 2003 52 553 560 12910455 10.1002/prot.10489
Dolezalova H Shankar G Huang MC Bikle DD Goetzl EJ Biochemical regulation of breast cancer cell expression of S1P2 (Edg-5) and S1P3 (Edg-3) G protein-coupled receptors for sphingosine 1-phosphate J Cell Biochem 2003 88 732 743 12577307 10.1002/jcb.10394
Lee HG Zhu X Ghanbari HA Ogawa O Raina AK O'Neill MJ Perry G Smith MA Differential regulation of glutamate receptors in Alzheimer's disease Neurosignals 2002 11 282 292 12566929 10.1159/000067427
Unutmaz D KewalRamani VN Littman DR G protein-coupled receptors in HIV and SIV entry: new perspectives on lentivirus-host interactions and on the utility of animal models Semin Immunol 1998 10 225 236 9653049 10.1006/smim.1998.0134
Hopkins AL Groom CR The druggable genome Nat Rev Drug Discov 2002 1 727 730 12209152 10.1038/nrd892
Wess J Molecular basis of receptor/G-protein-coupling selectivity Pharmacol Ther 1998 80 231 264 9888696 10.1016/S0163-7258(98)00030-8
Alexander SPH Peters JA TiPS Receptor and Ion channel nomenclature supplement Trends in Pharmacological Sciences 2000 11 Elsevier
Wong SK G protein selectivity is regulated by multiple intracellular regions of GPCRs Neurosignals 2003 12 1 12 12624524 10.1159/000068914
Altschul SF Gish W Miller W Myers EW Lipman DJ Basic local alignment search tool J Mol Biol 1990 215 403 410 2231712 10.1006/jmbi.1990.9999
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
Horn F van der Wenden EM Oliveira L AP IJ Vriend G Receptors coupling to G proteins: is there a signal behind the sequence? Proteins 2000 41 448 459 11056033 10.1002/1097-0134(20001201)41:4<448::AID-PROT30>3.0.CO;2-C
Cao J Panetta R Yue S Steyaert A Young-Bellido M Ahmad S A naive Bayes model to predict coupling between seven transmembrane domain receptors and G-proteins Bioinformatics 2003 19 234 240 12538244 10.1093/bioinformatics/19.2.234
Moller S Vilo J Croning MD Prediction of the coupling specificity of G protein coupled receptors to their G proteins Bioinformatics 2001 17 Suppl 1 S174 81 11473007
Eddy SR Mitchison G Durbin R Maximum discrimination hidden Markov models of sequence consensus J Comput Biol 1995 2 9 23 7497123
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 Database issue D115 9 14681372 10.1093/nar/gkh131
Daaka Y Luttrell LM Lefkowitz RJ Switching of the coupling of the beta2-adrenergic receptor to different G proteins by protein kinase A Nature 1997 390 88 91 9363896 10.1038/36362
Krupnick JG Benovic JL The role of receptor kinases and arrestins in G protein-coupled receptor regulation Annu Rev Pharmacol Toxicol 1998 38 289 319 9597157 10.1146/annurev.pharmtox.38.1.289
Kroeger KM Pfleger KD Eidne KA G-protein coupled receptor oligomerization in neuroendocrine pathways Front Neuroendocrinol 2003 24 254 278 14726257 10.1016/j.yfrne.2003.10.002
Breitwieser GE G protein-coupled receptor oligomerization: implications for G protein activation and cell signaling Circ Res 2004 94 17 27 14715532 10.1161/01.RES.0000110420.68526.19
George SR O'Dowd BF Lee SP G-protein-coupled receptor oligomerization and its potential for drug discovery Nat Rev Drug Discov 2002 1 808 820 12360258 10.1038/nrd913
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 Database issue D138 41 14681378 10.1093/nar/gkh121
Sreekumar KR Huang Y Pausch MH Gulukota K Predicting GPCR - G protein coupling using hidden Markov models Bioinformatics 2004
Viklund H Elofsson A Best alpha-helical transmembrane protein topology predictions are achieved using hidden Markov models and evolutionary information Protein Sci 2004 13 1908 1917 15215532 10.1110/ps.04625404
Moller S Croning MD Apweiler R Evaluation of methods for the prediction of membrane spanning regions Bioinformatics 2001 17 646 653 11448883 10.1093/bioinformatics/17.7.646
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
Bailey TL Gribskov M Combining evidence using p-values: application to sequence homology searches Bioinformatics 1998 14 48 54 9520501 10.1093/bioinformatics/14.1.48
| 15847681 | PMC1087828 | CC BY | 2021-01-04 16:02:48 | no | BMC Bioinformatics. 2005 Apr 22; 6:104 | utf-8 | BMC Bioinformatics | 2,005 | 10.1186/1471-2105-6-104 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-631578013710.1186/1471-2105-6-63Research ArticleEvolutionary models for insertions and deletions in a probabilistic modeling framework Rivas Elena [email protected] Department of Genetics, Washington University School of Medicine, 4444 Forest Park Blvd., Saint Louis, Missouri 63108 USA2005 21 3 2005 6 63 63 16 12 2004 21 3 2005 Copyright © 2005 Rivas; 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
Probabilistic models for sequence comparison (such as hidden Markov models and pair hidden Markov models for proteins and mRNAs, or their context-free grammar counterparts for structural RNAs) often assume a fixed degree of divergence. Ideally we would like these models to be conditional on evolutionary divergence time.
Probabilistic models of substitution events are well established, but there has not been a completely satisfactory theoretical framework for modeling insertion and deletion events.
Results
I have developed a method for extending standard Markov substitution models to include gap characters, and another method for the evolution of state transition probabilities in a probabilistic model. These methods use instantaneous rate matrices in a way that is more general than those used for substitution processes, and are sufficient to provide time-dependent models for standard linear and affine gap penalties, respectively.
Given a probabilistic model, we can make all of its emission probabilities (including gap characters) and all its transition probabilities conditional on a chosen divergence time. To do this, we only need to know the parameters of the model at one particular divergence time instance, as well as the parameters of the model at the two extremes of zero and infinite divergence.
I have implemented these methods in a new generation of the RNA genefinder QRNA (eQRNA).
Conclusion
These methods can be applied to incorporate evolutionary models of insertions and deletions into any hidden Markov model or stochastic context-free grammar, in a pair or profile form, for sequence modeling.
==== Body
Background
Probabilistic models are widely used for sequence analysis [1]. Hidden Markov models (HMMs) are a very large class of probabilistic models used for many problems in biological sequence analysis such as sequence homology searches [2-4], sequence alignment [5], or protein genefinding [6-8]. Stochastic context-free grammars (SCFGs) are another class of probabilistic models used for structural RNAs for problems such as RNA homology searches [9-13], RNA structure prediction [14,15], and RNA genefinding [16].
Sequence similarity methods based on HMMs or SCFGs can take the form of profile or pair models and are very important for comparative genomics. These probabilistic methods for sequence comparison assume a certain degree of sequence divergence. For instance, in profile models (either profile HMMs [2-4] or profile SCFGs [12,13]) a sequence is compared to a consensus model. Profile models must allow for the occurrence of insertions and deletions with respect to the consensus, and they do so by using state transition probabilities that assign some position-dependent penalties for modifying the consensus with insertions or deletions. Similarly, in pair probabilistic models [8,16] two related sequences are compared (aligned and/or scored). Pairwise alignments need to allow for substitution, insertion and deletion events between the two related sequences. Substitutions are taken care of by residue emission probabilities, while insertion and deletion events are generally taken care of by state transition probabilities as in the case of profile HMMs.
In the BLAST programs [17], the score of a pairwise alignment is determined using substitution matrices which measure the degree of similarity between two aligned residues. Similarly, in pair probabilistic models, residue emission probabilities are based on substitution matrices. The evolution of substitution matrices has been studied at large for many different kinds of processes: nucleotides, amino acids, codons, or RNA basepairs [18-23]. The evolution of emission probabilities using substitution matrices is easily integrated into probabilistic models both for HMMs [24-29] and for SCFGs [14].
In probabilistic models, insertion and deletion events (indels) are sometimes described by treating indels as an additional residue (gap characters) in a substitution matrix. More often they are described using additional hidden states, where transition probabilities into those states represent the cost of gap initiation and transitions within those states represent the cost of gap extension. If the cost of gap initiation and gap extension are identical, it is referred to as a linear gap cost model. Hidden states allow arbitrary costs for gap initiation and gap extension, which is traditionally referred to as an affine gap cost model. Treating gaps as an extra character in a substitution matrix is equivalent to assuming a linear gap cost model. The parameters that modulate those processes should be allowed to change as the divergence time for the sequences being compared is varied. It has been difficult to combine probabilistic models such as profile and pair HMMs or SCFGs with evolutionary models for insertion and deletions [30-33]. Methods to evolve transition probabilities are not as well developed as those describing substitution matrices, but significant effort is currently aimed at this problem [34-41]. Models incorporating the evolution of insertions and deletions in the context of probabilistic models such as profile HMMs or pair models are a very important goal in order to make those probabilistic models more realistic.
I encountered this problem in working on QRNA, a computational program to identify noncoding RNA genes de novo. QRNA uses probabilistic comparative methods to analyze the pattern of mutation present in a pairwise alignment in order to decide whether the compared nucleic acid sequences are more likely to be protein-coding, structural RNA encoding, or neither. Originally QRNA was parameterized at a fixed divergence time. Motivated by the goal of making QRNA a time dependent parametric family of models, I investigated the possibility of evolving the transition and emission probabilities associated with a given probabilistic model. Since I already had the model parameterized for a given time, I aimed to use that model as a generating point of the whole time-parameterized family of models.
Because QRNA includes both linear and affine gap models in different places, in this paper I propose algorithms to describe the evolution of indels as a (N + 1)-th character in a substitution matrix, and algorithms to describe the evolution of the transition probabilities associated with a probabilistic model.
The purpose of this paper is to describe the general theoretical framework behind these methods. A detailed description of the particular implementation of these algorithms in QRNA and a discussion of the results obtained with "evolutionary QRNA" (eQRNA) will appear in a complementary publication.
Results
Evolutionary models for emission probabilities
The evolution of emission probabilities without gaps
In order to introduce notation, I will start with a brief review of the current methods for calculating joint probabilities conditional on time, P(i, j|t), where i, j are two residues (for instance, nucleotides, amino acids, RNA basepairs, or codons). P(i, j|t) gives us the probability that residues i and j are observed at a homologous site in a pairwise alignment after a divergence time t. Pairwise sequence comparison methods score aligned residue pairs with these joint probabilities either explicitly or implicitly [17]. In explicit generative pair probabilistic models, like the pair-HMMs and pair-SCFG in QRNA, the P(i, j|t) terms are referred to as pair emission probabilities.
The evolution of joint probabilities is usually obtained by modeling the corresponding conditional probabilities P(j|i, t)as a substitution process in which residue i has been substituted by residue j over time t. Probabilistic models for nucleotide substitutions [18,19,42-44] assume that nucleotide substitution follows a model of evolution that depends on an instantaneous rate matrix,
Qt = etR, (1)
where t is the divergence time, R is the instantaneous rate matrix, and Qt is the substitution matrix of conditional probabilities, that is Qt(ij) ≡ P(j|i, t). This is a reasonable model used, for instance, to describe nucleotide substitutions in the Jukes-Cantor [42] or Kimura [43] models, or the more general REV model [44]; this is also the evolutionary model used for amino acid substitutions [18,19,21,45,46], codon to codon substitutions [20,47], and RNA basepair to basepair substitutions [14,22,23].
Throughout this paper, I will use the words "divergence time", "divergence", or "time" equivalently to describe the amount of dissimilarity between biological sequences measured as the number of mutations and gaps introduced in the alignment of the sequences. I will never refer to "time" as representing an actual number of years of divergence, since this number cannot be determined intrinsically from sequence data.
Thus, given a rate matrix R, Qt (and therefore the desired joint emission probabilities) can be inferred for any desired time using the Taylor expansion for the matrix exponential,
This Taylor series converges in all cases.
There are several ways in which the rate matrix R can be determined. One approach is to use analytically inferred rate matrices that depend on a small number of external parameters [42-44,48]. For instance, the HKY model for nucleotide substitutions [48] depends on six parameters: the four stationary nucleotide frequencies, a rate of transitions, and a rate of transversions, which have to be provided externally. Another type of approach uses maximum likelihood methods [21,49,50] in order to estimate a rate matrix numerically from a training set of sequence alignments.
A third approach arises naturally in cases where suitable joint probabilities have already been estimated for a pair model, and we wish to make that model conditional on evolutionary divergence time. This approach starts from the assumption that our point estimate represents sequences at a particular arbitrary divergence time t*. For example, a similar assumption was taken to construct the BLOSUM matrices [51], which were obtained as joint probabilities at discrete point estimates from clusters of aligned sequences.
In this third approach the parameters at the generating time t* will be used to construct a rate matrix for the process. This approach is motivated by the kind of situation in which we find ourselves with probabilistic methods based on homology such as QRNA: a model has been trained in one kind of data, and the resulting probabilities represent some effective but fixed divergence time, and we wish to extend that model to a time-dependent parameterization.
For residue substitution processes, the rate matrix R and Q*, defined as the substitution matrix at the generating time t* [Q* ≡ Qt*], convey exactly the same information. More explicitly, assuming the evolutionary model given in equation (1) we can calculate the rate matrix of the process as a function of Q* as
Kishino et al [52] introduced the idea of calculating the rate matrix starting from a given substitution matrix using equation (3) and an eigenvalue decomposition of Q*. It is worth noting that the matrix equation for the rate R can be expressed as a Taylor expansion of the form
which allows for a direct numerical calculation of the rate matrix. The convergence of this series requires only that for every (real or complex) eigenvalue λ of matrix Q*, then |λ - 1| < 1. In addition, for any valid substitution matrix the eigenvalues have to be real and |λ| ≤ 1 (see Appendix A). Under these two conditions, the above Taylor series converges so long as the eigenvalues of Q* are positive. Therefore the three properties required of Q* in order to be able to obtain a rate matrix using the Taylor expansion in equation (4) are that its eigenvalues are all smaller than one (but one that is strictly one), real, and positive. Complex or negative eigenvalues would correspond to oscillatory behaviors, which do not seem to reflect the biology. All the substitution processes I have tested so far for nucleotides, amino acids, and RNA basepairs correspond to real and positive eigenvalues for which the above method is applicable.
It is relevant to compare instantaneous rate matrix approaches to the approach used in the PAM amino-acid substitution matrices [53]. The PAM matrices were not generated by calculating a rate matrix, but by estimating from a collection of highly similar sequences the substitution matrix for the time of one substitution per site ≡ Qt = 0.01, and then calculating Qt at any other (integer t) time by multiplication. This is a discrete approximation that converges to the same answer given by the rate method for very small time units. PAM matrices have been criticized for not being able to capture the substitutions that are observed for more dissimilar sequences. BLOSUM matrices empirically outperform PAM in sequence homology searches, presumably because sequences at larger divergence times were used to calculate the BLOSUM matrices. However, the BLOSUM method is not a time dependent continuous model but a very coarse-grained discretization. There are ways of combining the best of both approaches (more divergent sequence for training and a continuous-time model) to generate rate matrices, for instance by using the resolvent method [54], or using maximum likelihood methods as in the WAG matrices [21]. However, it is also possible to take a discrete BLOSUM matrix, for instance BLOSUM62, and convert it to an underlying rate matrix. The BLOSUM62-generated rate matrix obtained using equation (4) is shown in Figure 1.
A rate matrix can also be derived from the PAM data by various methods. One exact method is to do an eigenvalue decomposition as presented in [52]. Recently, other methods have been proposed to calculate a rate matrix from the Dayhoff data [55]. These methods still assume that R ≈ ( - I) which corresponds to taking only the first term in the Taylor series for the logarithm in equation (4). This assumption is good only for very closely related sequences. Using the Taylor series allows one to estimate, using the same input data and avoiding the calculation of eigenvalues, the rate matrix to any desired level of precision, independent of the degree of similarity in the training set.
Notice that the rate matrix obtained using BLOSUM62 (Figure 1) has two off-diagonal negative entries (and if we use more divergent BLOSUM matrices we have more negative off-diagonals). Off-diagonal entries of the rate matrix have to be positive so that I + δtR can be interpreted as a substitution matrix for very small times δt. This problem is not unique to sequence data. The construction of rate matrices for a Markov process from empirical data using a generating time is also used in mathematical modeling of financial processes such as credit risk modeling [56,57]. In the world of mathematical finances the problem is referred to as the regularization problem. I will use one of following regularization algorithms presented in [57]. The QOG algorithm (quasi-optimization of the generator) regularizes the rate matrix. The QOM algorithm (quasi-optimization of the root matrix) leaves the rate matrix unchanged and regularizes the conditional matrix at a given time if any negative probability appears. Using the QOG algorithm we obtain a regularized version of the rate matrix using BLOSUM62, which is given in Figure 2.
Regularization algorithms
Here I reproduce the QOG and QOM regularization algorithms. The proofs for these algorithms can be found in [57]. The QOG algorithm regularizes each row of a rate matrix independently. Given a row in a rate matrix R,
r = (r1,...,rn) ≡ (R(i, 1) ..., R(i, n)), (5)
the QOG algorithm solves the problem of finding the vector at the minimal Euclidean distance from r such that the sum of all its elements is zero, and all elements but one are positive.
The steps of the QOG algorithm are:
1. Permute the row vector so that r1 = R(i, i).
2. Construct the vector w, such that wi = ri - λ, where .
3. Obtain the permutation , such that .
4. Construct , for k = 2, ..., n - 1.
5. Calculate kmin = mink = 2, ..., n - 1 {k such that Ck ≤ 0}.
6. Construct the vector
7. The regularized row is given by r ← P-1 (). Finally reverse the permutation of step (1).
The QOM algorithm regularizes each row of a conditional matrix independently. Given a row in a conditional matrix Qt
r = (r1,..., rn) ≡ (Qt(i, 1),..., Qt(i, n)), (6)
the QOM algorithm solves the problem of finding the vector at the minimal Euclidean distance from r such that the sum of all its elements is one, and all elements are positive.
The steps of the QOM algorithm are:
1. Construct the vector w, such that wi = ri - λ, where .
2. If all wi are non negative, r ← w is the new regularized row.
3. Otherwise, obtain the permutation , such that .
4. Construct , for k = 1,..., n.
5. Calculate kmax = maxk = 1,..., n {k such that Ck ≤ 1}.
6. Construct the vector
7. The regularized row is given by r ← P-1 ().
A 4 × 4 example starting from joint probabilities at a given generating time
As an review of these techniques, I will use a set of 4 × 4 single-nucleotide joint probabilities P(i, j|t*) for i, j = {a, c, g, t} at a particular generating time t* to construct the corresponding rate matrix.
In this example, the joint probabilities at the generating time using the matrix notation P*(ij) ≡ P(i, j|t*) are given by,
These 4 × 4 pair-nucleotide probabilities are taken from the program QRNA. They were calculated according to [16] by marginalizing codon-codon joint probabilities which were constructed from the BLOSUM62 matrix of amino acid substitutions. These 4 × 4 probabilities can be viewed as a particular example of the REV model [44]. Note that the sum of all elements of P* adds up to one, and the matrix is symmetric.
The marginal probabilities defined as pi = ∑j P(i, j|t*) can be calculated from the joint probabilities to be,
p = (pa, pc, pg, pt) = (0.2836, 0.2311, 0.2531, 0.2322). (8)
Similarly, the conditional probabilities P(j|i, t*) can be calculated from the previous joint and marginal probabilities using the relationship P(i, j|t*) = P(j|i, t*) pi. Using the matrix representation Q*(ij) ≡ P(j|i, t*) we have,
Notice how the sum of the elements in each row adds up to one. Notice also how Q* is quite different from the identity matrix, which means that we have started with a quite divergent generating time.
If we assume a homogeneous Markov substitution process, we can interpret the conditional probabilities Q* as the matrix of substitution probabilities at the generating time. Thus, we can characterize the underlying evolutionary process by its instantaneous rate of evolution, which can be calculated from Q* using equation (4). The resulting rate matrix R (up to an arbitrary scaling factor t*) is given by,
This rate matrix has all the good properties: (i)"Normalization": the sum of the elements of each row is zero. (ii) "Reversibility": pi Rij = pj Rji. The process is reversible by construction because we started with symmetric joint probabilities. (iii) "Saturation". The rate matrix converges at time infinity to the given marginal probabilities in equation (8). We can test saturation by using equation (2) and calculating the substitution matrix for a very large time. For instance, for t = 10t* we have
Saturation (or stationarity) of a Markov process is a necessary consequence of (i) normalization and (ii) reversibility. Appendix A shows a derivation of the previous statement which was useful for me (and hopefully for some readers) when studying the behavior at t = ∞ of more complicated evolutionary models. Therefore, starting from joint probabilities as in this example, we can always interpret the marginal probabilities as the stationary probabilities of the evolutionary process.
In summary, starting with a single set of joint probabilities at one particular generating divergence time t*, we calculate the joint probabilities at any other arbitrary time, assuming an exponential model of evolution. To that effect, given the particular set of joint probabilities (7) we have calculated the corresponding rate matrix (10) by Taylor expansion. Thus we can estimate the substitution matrix/conditional probabilities at any other arbitrary time, simply using equation (2), and reconstruct the joint probabilities at any other arbitrary time. For instance, for t = 0.3t* we obtain,
This method allows us to evolve pair emission probabilities corresponding to different processes (in addition to the 4 × 4 nucleotide emissions) for instance 20 × 20 amino acid-to-amino acid joint emission probabilities, 64 × 64 codon-to-codon joint emission probabilities, or 16 × 16 RNA basepair-to-basepair joint emission probabilities. Thus, this method is useful to be applied in combination with pair HMMs or pair SCFGs already parameterized at one fixed divergence time to make their emission probabilities a time-dependent family.
The evolution of emission probabilities with indels treated as an extra character
Substitution processes (even if describing multi-nucleotide events such as codon evolution or RNA basepair evolution) are not enough to describe the full evolutionary relationship between two biological sequences. We also need to consider indels, for which we need to introduce more complicated models of evolution than the one described so far.
Indels have traditionally been a problem for phylogenetic methods. Programs to construct phylogenetic trees from data such as PHYLIP [58], PAUP* [59], and other phylogeny packages [60-64] treat gaps as missing data. The theoretical description of the evolution of gaps in a probabilistic fashion reached a landmark with the Thorne/Kishino/Felsenstein (TKF) model [30,31]. The TKF model however is hard to implement in combination with a probabilistic model such as an HMM, although an active area of research exists in that direction [36,39,40]. A more direct attack to the problem of introducing phylogeny into existing probabilistic models originated with the concept of tree HMMs [34,35]. The tree HMM method models the evolution of the parsing of different sequences through an HMM. This approach is more related with the evolution of transition probabilities, and I will discuss it later on in this paper.
Here I am going to describe a method for the evolution of indels under the assumption that they behave like an additional residue added to a N × N residue substitution matrix. This is a simplification of the problem because it forces indels to have linear penalties (that is, the cost of opening an indel in an alignment or the cost of extending it with one more indel character is the same) and to behave independently of each other (that is, successive indel characters in one sequence will be treated as independent events, rather than as a single indel of n residues long). Despite its apparent simplicity, this approach poses interesting problems in parameterizing evolution.
Let us review some of the implications of insertion and deletion processes. The treatment that pair models give to pairwise alignments can be interpreted (if we assume reversibility, as is the case here) with all generality as if one of the sequences is the ancestor of the other one. For any two aligned residues we assume that they can be related by a substitution process. For a residue aligned to a gap we assume that either a residue in the ancestor was deleted in the descendent sequence, or that a residue not present in the ancestor appeared in the descendent sequence.
An stochastic insertion–deletion process also involves insertions followed by subsequent deletions. These events leave no trace in pairwise alignments because alignments usually do not retain gaps aligned to gaps. However, when we are treating indels as an extra character, we have to account for such events.
If we were given ideal alignments with all their gap-to-gap aligned columns we could estimate from data the (N + 1) × (N + 1) extended joint probabilities at a generating time, . Because that is not the case, we need to make some inference about . Let us represent with Δ, such that 0 ≤ Δ ≤ 1/2, the expected frequency of observed gaps with respect to the total number of residues in pairwise alignments at a particular time t*. The parameter Δ, can be estimated from data, or it could be estimated according to the TKF model [30] as
if we knew the values for the rate of insertions λ and the rate of deletions μ, such that 0 <λ <μ.
Let us represent with Δ' the expected frequency of missing gap-to-gap aligned columns in a pairwise alignment at a particular time t*. One can estimate Δ' as the expected length of insertions that were later deleted without leaving any trace in current sequences. The probability of a stretch of l gap-to-gap characters is given by the geometric distribution density ρ(l) = (1 - Δ2) Δ2l. Therefore Δ' is given by,
Using these two parameters and the joint probabilities in the absence of gaps at the generating time P*() we can construct the set of (N + 1) × (N + 1) extended joint probabilities at t* as
where we have assumed independence for the joint probability of a residue and a gap. The normalization factor Ω = 1/(1 + 2Δ + Δ') represents the fact that the observed Δ is different from the value we would have obtained had we known the complete alignment.
Another implication of insertion and deletions appears in the behavior of the marginal probabilities of single residues and indels. At t = 0 when sequences have not yet diverged, the marginal probability of finding a gap in an alignment should be zero. In the limit t = ∞, the pairwise alignment of two finite-length sequences is going to be dominated by gap-to-gap alignments, which implies that as the divergence time increases the marginal probability of a residue becomes negligible, while the marginal probability of a gap becomes one in the limit t = ∞. Our evolutionary model has to be able to accommodate such saturation frequencies.
A step-by-step description of the algorithm for the evolution of gaps as an extra residue
I will start by describing the steps to implement the method before explaining how to derive those steps. This method can be applied starting from two different situations: starting from a N × N set of joint probabilities at a generating time that need to be extended to allow indel characters and evolved with time; or starting from a given N × N rate matrix that needs to be extended to allow indel characters.
Suppose we start with a N × N set of joint probabilities P* at a generating time t*, where p stands for the marginal probabilities and Q* represents the set of conditional probabilities associated with P*.
1. Extend the joint probabilities at the generating time t* to a (N + 1) × (N + 1) matrix of joint probabilities of the form,
where Δ is a parameter which represents the expected frequency of gaps with respect to the total number of residues in an pairwise alignment at t*, and which satisfies the condition 0 ≤ Δ ≤ 1/2. The parameter Δ' is given in terms of Δ as , and the normalization constant is given by Ω = 1/(1 + 2Δ + Δ'). The indices with hats () stand for the N residues, and exclude the gap character, which I represent with the symbol -.
The (N + 1) × (N + 1) extended conditional probabilities at the generating time are given by,
2. Construct the (N + 1) × (N + 1) extended rate matrix Rε as
where
3. Calculate the exponential of the rate matrix using the Taylor expansion,
4. Construct the extended matrix of conditional probabilities at arbitrary time as
where the matrix Q0 is given by
where are the original marginal probabilities of P*, and the probability 0 <q0 ≤ 1 is given by . The function δ(ij) is a Kronecker delta which takes value one for i = j and zero otherwise. The case q0 = 1 corresponds to the extreme case in which the N + 1 gap residue does not evolve.
5. Construct the extended marginal probabilities as
where the probability of a gap at time t is given by,
I call this process "quasi-stationary" because the background frequencies () at any finite time are always proportional to the original N-dimensional background frequencies . This result is a concequence of the fact that the first N elements of the last row of Q0 are proportional to the N stationary frequencies . On the other hand, while remaining "quasi-stationary" the background frequencies evolve from (, 0) at time zero towards "all gaps" at time infinity, i.e. limt → ∞ Λt = 1. This behaviour at time infinity is the consequence of the particular value of q0 selected in the previous step.
6. Finally, construct the evolved (N + 1) × (N + 1) joint probabilities at arbitrary time as
The expression for Λt in equation (24) guarantees reversibility, that is, that the extended constructed according to the above expression are symmetric.
For the other starting situation, in which we have a N × N rate marix R, the procedure to generate a (N + 1) × (N + 1) quasi-stationary reversible evolutionary model is the following:
1. Construct the (N + 1) × (N + 1) extended rate matrix Rε as
where we have extended the N × N rate matrix R with the parameter β > 0.
The instantaneous rate is given by,
Thus β is the instantaneous rate of deletion of a character, while -β(1 - q0) is the rate of insertion of character . (More complicated models in which the rate of deletion is different for different characters are also possible.) Notice that q0 = 1 corresponds to the case in which the rate of insertions is zero.
2. Find analytically, if an analytic expression for Rε is given by solving the differential equation d () / dt = Rε , or numerically, proceeding as in step (3) of the previous procedure.
3. Proceed as in steps (4)-(6) of the previous procedure.
A 5 × 5 example starting from joint probabilities at a given generating time
We start with the generating joint probabilities P* in the 4 × 4 example in equation (7), which we want to extend to a 5 × 5 matrix by adding a gap character. For this example, I have selected the arbitrary value for the gap parameter Δ = 0.18.
The 4 × 4 joint probabilities in equation (7) augmented to a 5 × 5 matrix using the gap parameter Δ = 0.18 (which implies that Δ' = Δ2/(1 - Δ2) = 0.0335) is given by,
The conditional probabilities (ij) = Pε (j|i, t*) are given by,
The extended marginal probabilities at the particular time instance t* are given by,
= (pa, pc, pg, pt, p-) = (0.2402, 0.1957, 0.2143, 0.1966, 0.1532), (30)
which are quasi-stationary with respect to the 4 × 4 stationary probabilities p = (0.2836, 0.2311, 0.2532, 0.2322) we started with in equation (8).
The matrix of conditional probabilities at time zero using expression (22) is given by,
The rate matrix for this example, calculated using the Taylor expansion described in equation (18) takes the value,
One should not be concerned to see a whole row of zeros for this rate matrix. For this generalized model the instantaneous rate of evolution is not directly given by the rate matrix; instead, the instantaneous rate of evolution is given by,
In this example, the instantaneous rate of evolution takes the form,
One should not be concerned either by having some negative off diagonal components. For small times δt, the conditional matrix is given by,
Therefore, in order to have a proper matrix of conditional probabilities for sufficiently small δt, it is necessary to satisfy the following condition for each pair of indices i, j,
if Q0(ij) = 0 then (Q0Rε)(ij) > 0. (36)
In this case, the off-diagonal components of the last row of Q0 are non-zero, which allows us to have negative off-diagonal elements for that row in the instantaneous rate matrix Q0Rε.
With the 5 × 5 rate matrix in hand, we can apply steps (3) and (4) to obtain the conditional probabilities at any arbitrary time . For instance for t = 0.3t* we obtain the following evolved conditional probabilities:
The quasi-stationary marginal probabilities are constructed using the result = 0.0487, and the 4 × 4 stationary probabilities p = (0.2836, 0.2311, 0.2532, 0.2322), following step (5) of the algorithm as,
Finally, using equation (25), for t = 0.3t*, we obtain the following evolved joint probabilities
Notice that this matrix is symmetric, which is the result of having imposed reversibility for any arbitrary divergence time.
We can also see by calculating the conditional probabilities at large divergence times how these probabilities evolve towards their saturation values given by (0, 0, 0, 0, 1). For instance, for t = 30t* we have,
An example starting from a rate matrix: The Jukes-Cantor model extended to gaps
As an example of a situation in which we start with a rate matrix, let us consider the generalization of the Jukes-Cantor model [42] to a 5 × 5 evolutionary model with a gap character. The original Jukes-Cantor model assumes that all nucleotides mutate at the same rate α > 0 which is represented by the rate matrix
In this simple case the conditional matrix Qt = etR can be found analytically by solving the matrix differential equation . Because of the symmetries of the problem we can write
with the condition rt + 3st = 1. We then obtain the following differential equations
and the solutions are,
By taking the limit t = ∞ in the previous two equations, one can see that the saturation frequencies of the Jukes-Cantor model are pi = 0.25 for i = a, c, g, t.
The 5 × 5 extended Jukes-Cantor rate matrix Rε is constructed by adding a rate of mutation to a gap represented by the quantity β ≥ 0 which in principle we will assume is different from the rate of substitutions α,
We also introduce the matrix at time zero Q0 which depends on the probability parameter 1 ≥ q0 > 0,
where the particular case q0 = 1 is only allowed if simultaneously β = 0, and corresponds to a trivial extension of the original Jukes-Cantor model in which the gap character does not evolve.
The conditional matrix at arbitrary time is given by = Q0. The symmetries of the problem in this case allow us to parameterize as
with the conditions rt + 3st + γt = 1 and 4ξt + σt = 1.
Introducing the matrix Mt ≡ , we can parameterize
which implies that
σt = (1 - q0) γt + q0. (52)
The differential equation to calculate Mt takes the form = Rε Mt, which translates into the differential equations,
Which are satisfied by
γt = 1 - e-βt. (58)
And in addition we have
σt = 1 - (1 - q0) e-βt. (60)
In the limit case β = 0, the solutions for rt and st reduce to those of the original Jukes-Cantor model with the trivial additions of σt = 1, ξt = 0 and γt = 0, after setting q0 = 1.
The extended Jukes-Cantor model depends on three parameters: the rate of nucleotide substitution α > 0, the rate of nucleotide deletion β ≥ 0, and the parameter 1 ≥ q0 > 0. What is the meaning of q0 ? q0 controls the saturation frequencies (i.e. the background frequencies at time infinity), as well as the background frequencies at any other finite time. For β > 0 and 1 >q0 > 0, taking the limit t = ∞ in equations (56)-(60), one can see that the saturation probabilities are given by (0, 0, 0, 0, 1). At any other finite time, the background frequencies of the model are quasi-stationary with respect to the background frequencies of the original Jukes-Cantor model, and are given by
pt (-) = Λt. (62)
Imposing the reversibility condition in particular we obtain (1 - Λt)γt + Λtσt = Λt which implies,
Therefore q0 controls how fast the background frequencies approach the saturation probabilities (0, 0, 0, 0, 1) through the factor Λt. For a given β, the larger q0, the faster Λt approaches one. (Note that Λt always approaches one as t goes to infinity.)
At first glance, it looks like q0 could take any value including one in the solution for the extended Jukes-Cantor model. q0 = 1 would result in fixed background frequencies of the form (0, 0, 0, 0, 1), which is an undesirable result, and the value q0 = 1 would have to be excluded when β > 0. In fact, the limit to the ungapped Jukes-Cantor model has to be taken by setting β = 0 first, and then q0 = 1. In that way, Λt = 0 for all times, which is the correct result for the original Jukes-Cantor model.
Derivation of the algorithm for a (N + 1) × (N + 1) quasi-stationary and reversible evolutionary process
Unlike the ungapped N × N case in which the marginal probabilities are time independent, in the presence of gaps the marginal probabilities have to evolve with time. In fact, as I discussed earlier, the marginal probability of a gap (-) has to evolve from zero at time zero to one at time infinity. As a result of that observation, probabilistic evolutionary models with Q0 ≠ I are necessary in the presence of gaps in order to maintain reversibility. The reason for this requirement is the following: for an evolutionary model of the form etR, reversibility implies that there is some p* such that p*Q* = p* [see Appendix B, equation (202)]; it follows then that p*R = 0, and therefore p*etR = p* for arbitrary time t. Thus, under a reversible model of the form Qt = etR, marginal probabilities do not evolve with time. On the other hand, if Q0 ≠ I then the condition p*Q* = p* does not imply p*R = 0 for the rate matrix R, and therefore it does not impose p* as the marginal probabilities for arbitrary t. (See appendix B for more details on this point.)
Therefore, to model the evolution of gaps we need to generalize the evolutionary model to have the following form
= Q0. (64)
The matrix Q0 can be parameterized in following form,
This matrix depends on one additional parameter q0. The particular dependency Q0(-, ) ∝ , is necessary to obtain quasi-stationary reversibility of the marginal probablilities.
The rate matrix is now a function of Q0 and , and takes the form
where the matrix has the form,
where
Notice that Q0 may be inverted as long as 0 <q0 ≤ 1.
With respect to the marginal probabilities we have that at the generating time t* because of the way the extended probabilities P* were constructed we imposed a quasi-stationary behavior of the form,
where
The generalized conditional matrix in (64) also saturates at very large times, and the saturation probabilities (i.e. the marginal probabilities at infinity) are given by those of the rate matrix, that is limt→∞ = limt→∞ (see Corollary A.1). Because of the relationship in equation (66) between the rate matrix and the matrix , the saturation probabilities are given by the condition (see Appendix B),
Then using equation (71) we can see that the saturation probabilities maintain the quasi-stationary property that was imposed at the time instance t*, and are given by,
where
As I discussed before, it is reasonable to impose that at infinity all we find is gaps, i.e Λ∞ = 1.0, which implies η = 1 and
Notice that because the relationship given in equation (14) between Δ' and Δ, then 0 <q0 < 1.
For an arbitrary time we have the reversibility relationship
This equation is satisfied by construction in the N × N subspace. By inspecting the implications of the above equation for the gap index, we obtain an expression of Λt (the marginal probability of a gap) at arbitrary time that allows us to have quasi-stationary reversible evolution. The function Λt is given by,
Evolutionary model for transition probabilities
The standard way in which comparative probabilistic models allow for insertions and deletions is by introducing several additional states with their corresponding transition probabilities. For instance, in a pair-HMM for sequence alignment (Figure 3) the presence of gaps requires the introduction of two states ("X" and "Y") which emit a nucleotide in only one of the two sequences. The probabilities associated with transitioning in and out of those states control the "gappiness" of the alignment. Therefore the evolution of these parameters with time is necessary in order to model different degrees of sequence divergence.
There has been a continued effort on improving the accuracy of the evolution of emission probabilities (i.e. substitution matrices) such as allowing correlations between the rates at different sites [65,66], improvements in the derivation of rate matrices from sequence data [23,67], or estimating multiple nucleotide changes [68]. In comparison, the ideas to describe the evolution of transition parameters in probabilistic models are much less standardized [34-40].
The goal of this section is to describe the evolution of transition probabilities. For instance, in the pair-HMM of Figure 3 the transition probabilities from the "XY" state to the "X" or "Y" states describe the introduction of gaps in one of the two aligned sequences, using an affine penalty. These transitions should be zero when the sequences have not yet diverged (time zero), but they should be maximal at infinite divergence. In between these two extremes, it is desirable to model the transition probabilities changing with divergence time. These methods are termed "evolutionary" because the transition probabilities will be parameterized with time, using functions that are generalizations of the Markov process that probabilistic evolutionary models assume for substitutions. Unlike the TKF model [30,31] and other related evolutionary models [32,33,41], the approach presented here will not describe the actual underlying evolutionary process that may have generated one sequence from another.
The tree-HMM method [34,35,37] is possibly the method closest to what I develop here. A tree HMM tries to model the phylogenetic relationship between related sequences by modeling the parsings of different sequences through the model. In a tree HMM it is not the actual transition probabilities of the HMM, but the parsing of the different sequences through the models that are evolved using rate matrices that resemble the diagonal rate matrices introduced in the first of the methods described below. Here I want to generate pair or profile probabilistic models that when comparing two related sequences are able to accommodate to the degree of divergence observed between the two sequences, and I intend to do that in a continuous-in-time and probabilistic fashion, using the smallest possible number of free parameters. No evolutionary history of individual insertion/deletion events will be generated; only a posteriori would an evolutionary history be established by comparing sequences (in the case of a profile model) or alignments (in the case of pair models) generated by the model at different times.
I present two methods to evolve transition probabilities. One of the methods considers the evolution of a vector of transition probabilities. In this method, the value of the transition probabilities at time zero and time infinity are input parameters, which gives a relatively large number of free parameters. In the second, more restrictive, method the transitions associated with several states are assumed to evolve under the same evolutionary process. This condition constrains some of the free parameters, but does not fix them all completely. When the more restrictive conditions are used, both algorithms give the same results. These two algorithms are applicable to most pair and profile probabilistic models, be they HMMs or SCFGs, generalized or not. I present an example of the evolution of a vector probability vector for a pair HMM, and an example of the evolution of a matrix of transition substitutions for a profile HMM.
Evolution of a vector of transition probabilities
A step-by-step description of the algorithm
Let us start by providing the recipe to apply the algorithm:
1. Given a transition probability vector
2. Assume its set of values is known at the three particular time instances of t = 0, t = t*, and t = ∞, named q0, q*, and q∞. Assume each component i in these probability vectors satisfies one of the following three conditions,
q0(i) <q*(i) <q∞(i) or q0(i) >q*(i) >q∞(i) or q0(i) = q*(i) = q∞(i), for all i. (78)
3. If the three input vectors satisfy the condition,
where r > 0 is a real number independent of i, then calculate qt at an arbitrary time t (0 <t <t∞) as
Normalization of the vector qt is guaranteed by equation (79).
4. Otherwise qt is given by the following expression
where the function wt is given by
An example: evolution of the transition probabilities of a pair-HMM "XY" state
Consider the transition probability vector associated to the "XY" state of the pair-HMM given in Figure 3,
which describe the four possible transitions from a correlated emission of two nucleotides to another correlated emission in both sequences (); to a gap in sequence Y (); to a gap in sequence X (); or to end the alignment ().
Below are some arbitrary values for the transition vector at divergence times: t = 0, t = t*, and t = ∞ associated with state "XY", qXY,
The transition TXY→E = τ is related to the expected length of the alignments generated using the model. We typically want to keep that transition invariant through time, and correlated with the alignment length L: τ = 1/L. (This pair HMM produces sequences with a geometric length distribution of mean 1/τ.) The other three transitions change with time from a situation of no gaps at time zero, to a situation at time infinity in which all there is present is gaps, because no residue in either sequence has a homologous residue in the other.
Transition probabilities at t = 0 and t = ∞ can be stated from first principles. Transitions at the generating time t*, are estimated from data, at the same time that emission probabilities are estimated. The transition probabilities at any other time are given by applying the algorithm. Using equation (81) we obtain,
Similarly to this "XY" state case, all the other transition probabilities that appear in the pair model of Figure 3 could be continuously parameterized with the divergence time of the alignment being scored. This algorithm can be applied to any full set of transition probabilities emerging from a particular state in a given probabilistic model that must evolve with time.
Connection with a tree-HMM 2×2 match-transition matrix
In the original representation of a tree-HMM [35] the idea of a match-transition matrix is introduced. If one parse through the HMM generated a Match to Match (MM) transition, while another parse through the model generated a Match-to-Delete (MD) transition, one can consider the substitution of MM by MD similarly to a substitution of residues by the conditional probability P(MD|MM, t). This leads to the concept of a 2 × 2 match-transition matrix given by,
which in [35] is parameterized with two real numbers r ≥ 0, and 0 ≤ a ≤ 1 as
Tree-HMMs model the evolution of paths though the HMM. In contrast, the method proposed here models the evolution of the transition probabilities of the model themselves. However, one can see that the match-transition matrix is closely related in form to the model we have proposed here. Introduce the probability vectors,
For t = 0 and t = ∞ they have the following values,
It is easy to see that the match-transition matrix given in equation (87) can be rewritten as,
for a diagonal rate matrix . Such diagonal rate matrix does not require additional normalization because it corresponds to the case described in equation (80).
Derivation of the algorithm to evolve a vector of transition probabilities
To describe the evolution of transition probabilities, the simple exponential models used for substitution matrices are not sufficient. I propose to adopt a generalization of the exponential model of the form,
where I is the n × n identity matrix, r is a vector still to be identify, and a R is a n × n rate matrix.
This model simply adds to the exponential term a time independent vector a,
Because qt=0 = q0, then it is necessary that a = q0 - r, thus giving the expression in equation (92). Note that this is the most general solution of a differential equation of the form ∝ (qt - a). Until now it was always assumed that the constant term was zero, that is r = q0. The freedom added by including a term constant in time is that, while before the behavior at t = ∞ was solely controlled by etR, now the additional term also contributes to that limit.
An immediate consequence of this generalization is that the rate matrix is not now sufficient to determine the whole evolutionary process. In addition to the rate matrix, the probability vector must also be specified at time zero (no divergence) and at time infinity (all mutations have saturated) such that,
The exponential of the rate matrix R has the general form,
for some real eigenvalues . If conditions are restricted to the case in which ki > 0, ∀i, the immediate consequence of working with positive eigenvalues is that,
There is then a simple relationship between the vector r and the values of the probabilities at time zero and saturation,
q∞ = q0 - r. (97)
Therefore, we can write with all generality
However, for the given information (q0, q*, q∞), the time-parameterized vector qt in (98) is still underdetermined.
In order to reduce the amount of freedom, I assume that etR is diagonal (i.e. U = I). Diagonal rate matrices have been used in other contexts of generalized evolution such as the tree-HMM model [34,35]. Then we have,
At this time the known probabilities at the generating time t*, q*, have not yet been used. These are,
Thus we obtain
which can be solved for ki,
The condition ki > 0 translates into 0 < < 1, or
This condition has two solutions:
q0(i) <q*(i) <q∞ (i) or q0 (i) >q* (i) >q∞ (i). (104)
Even though this model was derived under the conditions of equation (104), it also extends to the degenerate case where for some i we have
q0(i) = q*(i) = q∞(i), (105)
since this simply corresponds to these parameters undergoing no evolution at all.
Therefore if the input column vectors satisfy one of the three previous conditions for each one of their elements, the parametric expression is
A normalization condition has not yet been imposed. Using the unity vector uT = (1,..., 1), normalization requires that
For an evolutionary model of the form qt ∝ etR normalization requires that etRu = u for arbitrary times. I refer to this property as the strong normalization condition. The normalization of a generalized evolutionary model of the form requires the weaker condition (q0 - q∞)TetRu = 0. This property is always true for a rate matrix that satisfies the strong normalization condition. I refer to this property as the weak normalization condition.
In order to obtain the strong normalization condition automatically it is necessary to have a rate matrix of the form R = U diag (0, - λ2,..., -λn) U-1 (see Appendix A, equation (193)). Such a type of rate matrix is not appropriate to describe the evolution of a probability vector, since such rate matrix cannot be uniquely inferred from the three input probability vectors q0, q*, q∞. For that reason, I have explored the use of rate matrices of the diagonal form R = diag (-k1,..., -kn), which can be inferred from the three input probability vectors q0, q*, q∞ using expression (102). Such diagonal rate matrices, however, in general do not satisfy the strong normalization condition, thus the weak normalization normalization condition must be obtained by other means.
Define wt:
If the input vectors satisfy the condition for all i,
for some real number r > 0, then the rate matrix has the particular form R = diag(-r,..., -r), the function wt = 0 for arbitrary times, and the weak normalization condition is satisfied automatically.
The previous condition is in general too restrictive. If the previous condition is not satisfied, by construction wt is zero at t = 0, t = t* and t = t∞, but in general wt ≠ 0 for n > 2. Normalization is then achieved (107) by modifying our definition of qt to
The final expression is
Evolution of a matrix of transition probabilities
In some cases, several states of a model correspond to a particular evolutionary event, and it seems natural to expect that their transitions would evolve under the control of the same rate matrix. For instance, in a profile HMM (Figure 4) I will consider the joint evolution of the transitions of three states associated with a given consensus position: Match (M), Insert (I), and Delete (D).
For a collection of m states S = {S1,..., Sm} that transition into a collection of n states E = {E1,..., En}, consider the set of all transition probabilities emerging from the m originating states S and ending in the n states E,
The set of S and E states do not have to be mutually exclusive, and some E states can also be part of the S set. The set of E states also has to be complete, in the sense that
On the other hand, not all E states need to be reached by a given state; some transitions may be forbidden by design. For instance, for the states associated with a consensus position in a profile HMM the set of originating states is S = {M, D, I}, and the set of ending states is E = {M', D', I}, where the prime index indicates the next position in the profile. The condition m ≤ n can be imposed with all generality.
We want to describe the evolution with time of the transition probabilities. For that purpose, I will use the m × n matrix of transitions Qt such that,
Note that an evolutionary model of the form Qt = Q0etR, like that used for evolving gaps as extra characters, is not sufficient. A model of the form Qt = Q0etR in the limit of infinite divergence would necessarily result in transitions that for a given end state are all identical and independent of the previous state. That is clearly too restrictive for most models, for instance in a profile HMM, in which some of the transitions are not evolved and are set to zero.
In order to allow for more general saturation properties of the transition probabilities, I propose the following model for the evolution of the matrix of transition probabilities,
Qt = Q0 + K (etR - I), (114)
where R is the n × n rate matrix, and the m × n matrix K is still to be determined. This extension (as in the vector model proposed before) corresponds to adding a constant term A = Q0 - K, and it is is the more general solution of a differential equation of the form ∝ (Qt - A).
We will see that K(etR - I) = (Q0 - Q∞)(etR - I), thus
Qt = Q0 + (Q0 - Q∞) (etR - In × n), (115)
= Q∞ + (Q0 - Q∞) etR. (116)
As in the previous case, a freedom provided by the additional constant-in-time term is that while the saturation behavior of Q0etR is controlled by the saturation probabilities of etR, the model given by equation (116) is independent of those saturation probabilities so that the probabilities at infinity can be set arbitrarily. That is, assuming that ψ is the n dimensional vector of saturation probabilities of etR,
Notice that while Qt, Q0, Q∞ and K are m × n matrices operating in the S × E space, the matrices R and etR are square n × n matrices operating in the E × E space. In fact, etR determines the change in time that a transition probability into one of the E states experiences and in which fashion that change is absorbed by the transition probabilities into any other E state.
A step-by-step description of the algorithm
The recipe to implement the algorithm is as follows:
1. Assume we know the m × n (m ≤ n) matrices of transition probabilities at time zero Q0 and at time infinity Q∞, such that the rank of Q0 - Q∞ is m.
2. If an analytic n × n rate matrix R is given, one can find the analytic expression for etR by solving the differential equation d(etR)/dt = RetR, and jump to step (6). For a numerical solution jump to step (5).
3. If the information given is the set of transition probabilities at a generating time t*, calculate the rate matrix R as,
The n × m matrix O is obtained by solving the set of linear equations
-(Q∞ - Q0) O + umvT = Im × m, (120)
where um is the m dimensional unity vector [i.e. ], and v is a m dimensional vector uniquely determined by the set of m independent linear equations,
vT (Q∞ - Q0) = 0, (121)
vT um = 1. (122)
The solution of equation (120) is not unique. In fact, equation (120) determines the matrix O up to a n dimensional probability vector ψ that satisfies the conditions ψT O = 0. This probability vector corresponds to the saturation probabilities of the matrix etR. While the rate matrix R and the matrix etR depend on the choice of the saturation probabilities ψ, the asymptotic behaviour of the matrix of transition probabilities is independent of ψ, as was shown in equation (118).
4. Impose the condition,
vT (Q* - Q0) = 0. (123)
This condition [necessary so that ψT R = 0] imposes constraints between the set of probabilities at time zero, at time t*, and at time infinity.
5. Calculate the exponential of the rate matrix etR using the corresponding Taylor expansion.
6. Finally, calculate the set of evolved transition probabilities as,
Qt = Q0 - (Q∞ - Q0)(etR - In × n) (124)
or
Qt = Q∞ + (Q0 - Q∞) etR. (125)
An example: Evolution of the transitions of a profile HMM given a rate matrix
To illustrate this method, consider the case of a profile HMM (Figure 4). There are three states associated with a given consensus position in the profile: the Match state (M), the Insert state (I) and the Delete state (D). These three states transition into states M' (the Match state at the next position in the profile), D' (the delete state at the next position in the profile), and I (the insert state between the two matched positions), therefore in this example m = 3 and n = 3.
Consider the following the transition matrix
This transition matrix, like a nucleotide substitution matrix, adds up to one by rows. We assume as in HMMER [4] that there are no transitions between the Insert and the Delete states, but the model could work under more general conditions.
The matrix at time zero is given by
The parameter 1/(1 - qI) is the average length of an insert in between two matched positions at very short times (if there were no deletions). The parameter 1/(1 - qD) is the average length of a deletion at very short times (if there were no insertions, and all position in the profile had the same parameters at time zero). For instance, one could set qI very close to zero, which implies that, for very small times when , the average length of a insertion would be very close to one.
At time infinity, one can parameterize the transition probabilities as
where mD and mD represent the probabilities of Match to Delete and Match to Insert at infinity, and dD and iI are the Delete to Delete and Insert to Insert probabilities at time infinity, (0 ≤ mD, mI, dD, iI ≤ 1).
Let us assume that the rate matrix is given by
for some parameter α > 0. This rate matrix assumes that the rate of change in the occurrence of state M' is similar to that of state D' and that of state I, and that this change reverts equally into the other two states. More realistic situations can be achieved using rate matrices depending on more parameters.
The instantaneous rate of transition change is given by,
This matrix gives the instantaneous change that a transition probability experiences under this model and describes how that change is transferred to the other allowed transition probabilities.
The matrix etR can be obtained analytically by solving the differential equation d(etR)/dt = RetR. This is a 3-dimensional Jukes-Cantor model which has as its solution,
with
Putting all together, we obtain the following evolved transition probabilities for a profile HMM under a Jukes-Cantor like assumption for the rate matrix:
Substituting the values for st given in equation (133), we have the following evolved transition probabilities for a profile HMM,
The evolution of different paths through the HMM
In a tree-HMM one assumes that the different paths through the model are the objects that are subject to evolution [34]. Here we have directly modeled the evolution of the transition probabilities of the HMM. We can get an intuition for the meaning of these evolved transition probabilities by estimating how these evolved transition probabilities induce the evolution of different paths through the model. A process that is similarly to that modeled by a one-branch tree-HMM.
Suppose that at time zero, we emitted residue a from state M, and residue b from state M'. The model assigns to such sequence a probability given by,
where pM(a) and pM'(b) represent the emission probabilities associated to the M and M' states respectively. Now suppose that at time t there has been an insertion of n residues in between the two matches a and b; the model assigns to such sequence a probability given by,
where pI(ii) represent the emission probabilities associated to the Insert state.
We can interpret that in time t the path through the model that generated ab has evolved into the path through the model that generated ai1...inb with probability given by
To get a better intuition of what this means, take as an example the case in which the time interval t is very small.
Then the probability that a path between two matches in the HMM inserts n residues in time t ≈ 0 is
This probability is proportional to 3αmI the rate of substituting a Match-to-Match transition for a Match-to-Insert transition, and to (1 - qI), which is the geometric factor associated to an insert of length n at time zero.
An example: Evolution of the transitions of a profile HMM given the transitions at a generating time
In this case, we maintain the same values for the transition probabilities at time zero Q0 and at time infinity Q∞, but the rate matrix will be obtained from a generating time for which we know the transition probabilities.
The set of linear equations in step (3) of this algorithm that determine the vector vT = (v1, v2, v3) are
mDv1 + (dD - qD)v2 = 0, (146)
mIv1 + (iI - qI)v3 = 0, (147)
v1 + v2 + v3 = 1. (148)
The solution of these linear equations is
v1 = (dD - qD)(iI - qI)/d, (149)
v2 = -mD(iI - qI)/d, (150)
v3 = -mI (dD - qD)/d, (151)
where d ≡ (dD - qD)(iI - qI) - mD(iI - qI) - mI(dD - qD).
Parameterize the matrix O in the form
with each row adding to zero. The set of linear equations in step (3) that determine the matrix O are
(dD - qD)(M1 - D1) + v1 = 0, (iI - qI)(M1 - I1) + v1 = 0, (153)
(dD - qD)(M2 - D2) + v2 = 1, (iI - qI)(M2 - I2) + v2 = 0, (154)
(dD - qD)(M3 - D3) + v3 = 0, (iI - qI)(M3 - I3) + v3 = 1. (155)
Solving Mi and Ii in terms of Di we have,
The matrix O is therefore determined up to the unitary vector (D1,D2, D3). The saturation probabilities ψT = (ψM', ψD', ψI')(ψTR = 0) are defined by the equations ψTO = 0, which imply
Substituting vector D with vector ψ we finally obtain the following expression for the matrix O in terms of the saturation probabilities ψ:
The condition in step (4) of the algorithm translates in this case into the following relationship of parameters,
This is an additional set of constraints that the "vector" algorithm does not impose.
To test the algorithm I have made up a toy HMM consensus state, which at the generating time t* is given by the matrix of transitions,
Selecting the particular values qD = qI = 0.1 and dD = iI = 0.6, using the constraints of equations (161) implies that mD = 0.33 and mI = 0.25. Using these values and the arbitrary values for the saturation probabilities ψ = (1/3, 1/3, 1/3), we obtain the following O matrix:
The rate matrix R constructed using equation (119) is given by
and an instantaneous rate matrix -(Q∞ - Q0)R is given by,
The evolved set transition probabilities at time t = 0.3 is given by,
Using the "vector" method, in which transition probability vectors evolve independently with the same set of parameters, we would have obtained the identical result,
The normalization function wt given in equation (82) is different from zero only for dimensions larger than two. The second and third row effectively have dimension two (since one of the elements is always zero), and do not require normalization. For the first row the normalization function takes the value w0.3 = 0.0020.
The vector method allows us to use more unrestricted sets of parameters than the matrix method since the conditions in equation (78) are independent for each row. In principle, however, the conditions in equation (161) seem to allow behaviors that the vector model does not allow such as >mD ≡ as long as, simultaneously, >dD ≡ . In practice when I have tested that kind of situation, the rate matrices obtained are always not real, and therefore they lack any biological interpretation.
Derivation of the algorithm to evolve a matrix of transition probabilities
We start with a model of the general form
Qt = Q0 + K (etR - In×n), (168)
where Q0 is the known m × n matrix of probabilities at time zero, and the m × n matrix K must still be determined.
Assume that we know the transition probabilities at time infinity, which we represent by the m × n matrix Q∞. Then, because of the asymptotic behavior of the exponential family etR, limt→∞etR = unψT for some n dimensional saturation probabilities, where ψT = (ψi,...,ψn), and the n dimensional unity vector = (1, ..., 1), we have
Q∞ = Q0 + K (unψT - In×n). (169)
This equation implies that
K = -(Q∞ - Q0)+ ψT, (170)
where is a m dimensional vector that represents the sums by rows of K, i.e. ∑jK(i, j) = which we impose to be different from zero.
Because for the exponential family etR we have the reversibility condition ψTetR = ψT for arbitrary time, introducing the expression for K in equation (170) in the equation (114) we have the general result,
Qt = Q0 - (Q∞ - Q0)(etR - In×n). (171)
This result proves point (6) of the previous algorithm description.
Therefore if given Q0, Q∞ and a n × n rate matrix R, which satisfy the reversibility conditions ψTR = 0, we can calculate the evolved transition probabilities using the equation (171).
In the case in which the information given is the set of transition probabilities at a generating time t*, designated by Q*, the calculation of the rate matrix R involves the following steps:
(a) The m × n matrix K is by construction invertible because we have imposed ≠ 0, for all rows i.
A little aside with respect to matrix inversions is in order here. The (unique) inverse of a matrix is defined only for square matrices. One can introduce a inverse-like matrix for a non-square matrix; these are called pseudoinverses [69]. The pseudoinverse of a non-square matrix is not unique and many pseudoinverses can be defined; one of the best known is the Moore-Penrose matrix inverse [70]. We will see how despite the fact that the pseudoinverse of K is not unique, we can still define Qt uniquely.
Therefore solving for R in equation (114) at the particular time t* we have
where K-1 is the n × m pseudoinverse of K defined by the conditions KK-1 = Im×m and K-1K = In×n.
(b) Because the final result for Qt in equation (116) does not depend on the values we can set them with all generality to the form = ρ ≠ 0. Therefore we have
K = -(Q∞ - Q0) + ρum ψT. (173)
Because K-1Kun = un and Kun = ρum, then we need that K-1um = ρ-1un. Therefore we propose that the n × m pseudoinverse matrix K-1 has the following form,
where the n × m matrix O, and the m dimensional vector v satisfy the conditions,
O um = 0, (175)
vTum = 1. (176)
(c) In order to satisfy K-1K = In×n we need to have,
vT(Q∞ - Q0) = 0, (177)
-O(Q∞ - Q0) + unψT = In×n. (178)
Equation (177) is a set of homogeneous linear equations that together with the normalization conditions in equation (176) uniquely determine the vector v.
On the other hand, in order to satisfy KK-1 = Im×m, the following must apply:
ψTO = 0, (179)
-(Q∞ - Q0)O + umvT = Im×m. (180)
Equation (180) is a set of linear equations which determines O aside from a dependence on an arbitrary probability vector. In particular we can find the expression of matrix O in terms of the vector ψ as we did in equation (160).
Once the matrix O has been obtained using equation (180) as a function of the vector ψ, one can verify that the set of equations describe by (178) is automatically satisfied for any vector ψ as long as it satisfies the condition ψTO = 0. This is the result presented in step (4) of the algorithm.
(d) Because ψ corresponds to the saturation probabilities of etR, then it is necessary that ψTR = 0. This condition is satisfied if,
Therefore it implies that,
vT(Q* - Q0) = 0, (182)
which is the condition imposed in step (4) of the algorithm. Under those conditions, it results for the rate matrix R,
Notice that the parameter ρ ≠ 0, which is necessary to be able to invert the matrix K to calculate the rate R, does not appear anywhere in the final result, either in the evolved transitions Qt or in the value of R. This results from the fact that in either equation the only relevant component is the projection of K (or K-1) into (etR - I). The same projection is what makes the vector ψ that appears in the pseudoinverse K-1 irrelevant. Even though limt→∞etR = uψT, it is also true that limt→∞ (Q∞ - Q0)etR = 0, so that the dependence on ψ disappears from the final expression of Q∞.
Reversibility and multiplicativity
For a given probabilistic model, imposing reversibility has different implications for its emission and transition probabilities. In pair models, we assume that the emission probabilities are reversible by imposing P(at, bt+t') = P(at+t',bt), which corresponds to using symmetric joint probabilities represented by the shorthand notation P(a, b|t'). If the emissions do not involve gaps, the marginal probabilities do not evolve, and the evolved joint probabilities are obtained from the evolved conditionals and the saturation probabilities. In the presence of gaps, I have described how to construct the evolved conditionals and the corresponding evolved marginals in a way that maintains reversibility for any arbitrary time, so that we can construct evolved symmetric joint probabilities.
For transition probabilities the situation is different. Mathematically, a matrix of transition probabilities is like a substitution matrix (i.e. conditional probabilities) but there is not the equivalent of "joint" probabilities for transitions. To maintain reversibility for the transitions of a probabilistic model, one has to build reversibility in the design of the model. In particular, one needs to be sure that the transition probabilities that involve gaps lack any directionality. For instance, in the pair-HMM of Figure 3 we need to impose that for arbitrary times. That is achieved by making sure that the input transition probabilities at time t*, zero and infinity do lack directionality.
Another property of probabilistic models of evolution for residue substitutions is multiplicativity. Multiplicativity is an immediate property for evolutionary models of the form etR. For residue-substitution evolutionary processes, multiplicativity implies that the transition from one given event (say residue a) to another event (say residue b) in a finite time, if it goes through any intermediate state, has to be of the form of any other possible substitution. In mathematical terms,
However, when allowing gaps, any intermediate evolutionary step can go through processes of deletions or insertions in addition to substitutions; therefore multiplicativity as described in the previous equation does not hold anymore. There is a natural explanation of why "substitutions-only multiplicativity" is modified when considering insertion and deletion events. Consider the evolution of gaps as single characters, which was introduced previously in this paper. The substitution matrix with gaps satisfies the relationship
Analyzing this matrix equation by components and using the expression for Q0 given in equation (22), the substitution of residue a into residue b in finite time t + t' has the following terms:
The first term corresponds to pure substitution events of the form , and it is identical to equation (184). The second term modulated by the coefficient 1/q0 (introduced in equation (65), which is part of the non trivial matrix Q0) represents the event in which . The third term (preceded by coefficient (1 - 1/q0)) represents the event in which . Note that this model would align at time t + t' residues which could have been derived by a gap intermediate. This is usually discouraged by evolutionary models that describe the evolutionary history of insertions and deletions, in which such event would be represented as . For the model at hand, the fact that a gap can revert into a residue is a consequence of treating gaps as an additional residue in a substitution matrix.
For the particular case of the generalized Jukes-Cantor model introduced before, it turns out that the two extra terms in equation (186) are independent of the particular substitutions and cancel, such that
Therefore the generalized Jukes-Cantor model preserves multiplicativity. This results from the extreme simplicity of the model and is not true for more complicated models. For instance, for the rate matrix created from a particular Q* in the other example presented in this paper (which is a particular case of the REV model [44]), the two extra terms in equation (186) are different for the different nucleotide substitutions, and do not cancel out.
A more complicated situation appears for probabilistic models that introduce gaps in an affine manner. A given residue-to-residue substitution process that occurred in a finite time could have appeared from a very large number of intermediate situations in which stretches of other nucleotides could have been added or removed. The simple one-to-one correspondence that models of substitutions maintain through evolution does not exist in the presence of insertion and deletion events. This does not mean that evolutionary models with gaps are inconsistent, however some traditional properties of phylogenetic trees of single residue evolution such as the pulley principle [71] cannot be applied under the transition probability evolution models.
Conclusion
Motivated by the goal of making QRNA (a comparative probabilistic method for RNA genefinding) an evolutionary model, I have introduced several probabilistic methods to describe the evolution of insertion and deletion events. The methods introduced here have a larger scope than this program alone, and they can be applied to other pair probabilistic models and to profile HMMs and SCFGs as well.
I described an algorithm which addresses the evolution of gaps as an extra residue in a (N + 1) × (N + 1) substitution matrix. This method can be applied to the joint emission probabilities of pair models. This method allows us to maintain a stationary N-dimensional background distribution, while the actual (N + 1)-dimensional background frequencies evolve towards all gaps at time infinity. I call this process quasi-stationary. As an example, I showed an analytic solution for the Jukes-Cantor model extended to gaps.
I also presented two methods for the evolution of transition probabilities in a profile or pair HMM or SCFG, that are applicable to any probabilistic model that uses transitions between states to model insertions and deletions. In the first algorithm, the transition probabilities associated with one state in the model are evolved as a vector independently of the transition probabilities associated to any other state in the model. I also presented a second algorithm in which the transition probabilities associated with a given set of states co-evolve under the control of a single rate matrix. I presented an example of the application of these methods to a pair-HMM and to a profile HMM.
I have applied these methods to the program QRNA, which was the motivation for the development of the algorithms in the first place. QRNA contains three probabilistic models (the oth, cod, and rna models) that analyze the pattern of mutation of a given pairwise alignment to decide which of the three models best classifies the alignment. These models are a combination of generalized pair-HMMs and a pair-SCFG. Originally, this program assumed a fixed divergence time, and all the emission probabilities of the different models were tied to those of BLOSUM62. That produced a QRNA parameterized for highly diverse sequences, which in turn produced a large number of false positives for highly similar sequences. In the new program eQRNA, all emission and transition probabilities are a continuous time-dependent family able to match any possible degree of sequence divergence.
The three models of QRNA (the OTH, COD, and RNA models) need to be at approximately the same evolutionary distance, so that when a pairwise alignment is analyzed, the differences in scores of the models result from observing a different pattern of mutations (coding, RNA, or none in particular) rather than because one model favors more closely related sequences than the other. This model synchronization requires a number of QRNA-specific design elements which are tangential to the implementation of the evolutionary models for indels and transition probabilities presented in this paper. For reasons of clarity, I leave for another paper a detailed description of the particular implementation designs that went into eQRNA in order to make it fully evolutionary. In a nutshell, the transition probabilities of the OTH and COD models are evolved according to the algorithm to evolve vectors of transition probabilities, while the emission probabilities of those two models were evolved using the original QRNA parameters as the generating time of the respective rate matrix. In the RNA model, for the context-free grammar component of the model, the transitions are fixed, and the evolution of gaps is accommodated by treating gaps as extra characters according to the method presented here for that purpose. The HMM component of the RNA model is parameterized with time similarly to the OTH and COD models. Preliminary results show an important improvement compared with the previous fixed-time implementation. The application of these evolutionary methods for other probabilistic models for sequence comparison beyond eQRNA should be tractable.
So far the methods presented here have been introduced only in profile and pair models. They could also be applied to probabilistic models where, instead of aligning two contemporary sequences, one aligns a sequence to an ancestor. The only difference with respect to an evolutionary pair model is that, in this case, the emission probabilities will be the substitution (conditional) matrices themselves instead of joint conditional-on-time probabilities. One important limitation of the methods presented here is that, in general, they lack the property of multiplicativity. In consequence, in order to extend the methods presented here to more than two sequences related by a phylogenetic tree, one would have to work with rooted trees. A future challenge is to incorporate these evolutionary methods into multiple sequence probabilistic models that explicitly describe the phylogenetic relationship between the sequences.
Availability
The different models presented in this paper have been implemented in several small ANSI C programs. These are not fully developed software applications, but demonstrations (for those who want to avoid the mathematical descriptions) of how the different algorithms work. The programs are freely available at .
Methods
Appendix A. Conditions for the saturation of a generalized substitution matrix
In this appendix I provide the conditions for saturation of a generalized evolutionary model of the form Qt = Q0etR. Saturation can be described as
for the unitary vector u, and a set of saturation frequencies at time infinity denoted by q∞, such that .
Here I show that saturation of Qt = Q0etR is a necessary condition of two properties of the matrix Q = {Q(ij)}, normalization and positivity. I also show that the saturation probabilities of Qt are the same as those of etR.
Proposition A.1. Consider first the simplest case Qt = etR. Normalization, i.e. ∑jQ(ij) = 1, together with positivity, i.e. Q(ij) > 0 ∀i, j, imply that a substitution matrix of the form Qt = etR saturates to a set of probabilities at time infinity.
Proof. Normalization of the rate matrix, ∑j Q(ij) = 1 implies that
That is, λ = 1 is an eigenvalue of Q. It also has implications for the norm of Q, defined as the largest row sum of absolute values
Therefore, because of the spectral theorem [72], the spectral radius σ(Q), defined as the largest absolute value of any eigenvalue of Q, is bounded by,
σ(Q) ≤ ||Q|| = 1. (191)
On the other hand there is an eigenvalue λ = 1 therefore
σ(Q) = 1. (192)
In consequence, Q has one eigenvalue, λ = 1, and all other eigenvalues are smaller than one.
Therefore because the substitution matrix is of the form Qt = etR, it implies that the instantaneous rate matrix R has one null eigenvalue, and all the other are negative. If we assume that the null eigenvalue is not degenerate and that the negative eigenvalues are real, we can write with all generality,
for some matrix U, and such that λi > 0 for i = 2, ..., n.
Therefore Qt = etR can be cast into the form,
In the limit,
for = (1, 0, ..., 0).
On the other hand using equation (194) we obtain
which implies that UΨ0 is the eigenvector of Q corresponding to the eigenvalue λ = 1. According to (189) that is,
Substituting in equation (195) we finally obtain,
This is the saturation condition (188) for some saturation probabilities defined by q∞ = U-1.
Corollary A.1. For a generalized evolutionary model of the form Qt = Q0etR, Qt also saturates at infinity, and the saturation probabilities of Qt are given by those of etR, that is,
Proof. Note that by construction Q0 has to have the same normalization and positivity conditions as Qt. It can be shown that under those conditions, Qt = etR also has to add up to one, summing by rows, and all its elements have to be positive. Therefore, using the result of Proposition A.1,
Therefore
which proves saturation for an evolutionary probabilistic process of the form Qt = Q0 etR.
Appendix B. Implications of reversibility on a generalized evolutionary process
In this appendix I discuss the implications that reversibility imposes on a generalized evolutionary model. I show that for an evolutionary model of the form Qt = etR, the marginal probabilities with respect to which Qt is reversible have to be stationary, and therefore coincide with the saturation probabilities. I also show that for an evolutionary model of the general form Qt = Q0etR, the marginal probabilities with respect to which Qt is reversible can change with time. In this way we decouple the "reversibility" frequencies from the saturation frequencies. I also demonstrate how to calculate the saturation probabilities, given Q0 and Q* at one particular time t*. This system sets the ground for the quasi-stationary model of evolution with gaps as an extra indel.
Lemma B.1. Consider a given matrix of conditional probabilities Q*, [∑j Q*(ij) = 1 ∀i] which is reversible with respect to a set of marginal probabilities p*,
p*(i)Q*(ij) = p*(j)Q*(ji). (202)
Then one can see that reversibility implies
Proof. Summing one of the indices in the reversibility conditions and taking into account the normalization condition for the Q* matrix results in,
which in vectorial notation takes the form
Lemma B.2. If R = log Q* then the reversibility condition (202) for Q* implies that
Proof. If R = log Q* then because of the Taylor series we have
Because of the reversibility condition for Q* (202) it is also true that
p*(i) (Q* - I)n (ij) = p*(j)(Q* - I)n (ji), (207)
for n ≥ 1. Therefore it follows that
p*(i) R(ij) = p*(j)R(ji). (208)
In addition we can also see by inspecting equation (206) that the normalization condition for Q* translates into ∑j R(ij) = 0 ∀i, which implies that
Lemma B.3. If Q* is a conditional matrix that satisfies the reversibility condition (202) and R = log Q* then the saturation probabilities of R are given by the p* vector in (202), that is,
Proof. Taking from Lemma B.2., we have ; therefore for n ≥ 0, and because of the relationship
it results that
for arbitrary t. Therefore, it also holds in the limit of very large time that
Additionally, Appendix A shows that . Combining those two equations together we have
This proves that the saturation probabilities are p*.
Proposition B.1. For a reversible evolutionary model of the form Qt = etR, it results that the associated marginal probabilities with respect to which the parametric family Qt is reversible have to be stationary (i.e. time independent).
Proof. From the parametric family Qt select one particular instance t*, and consider . Suppose that the marginal probabilities at this time are given by p*, that is: . Because of the relationship R = log Q*, it follows from Lemma B.3 that the whole parametric family etR has p* as the corresponding marginal probabilities, therefore the marginal probabilities do not evolve with time (stationary).
Proposition B.2. For a reversible generalized evolutionary model of the form Qt = Q0etR, the associated marginal probabilities with respect of which Qt is reversible can be evolved with time.
proof. In order to prove that this is the case, we just need to find an example in which that statement is true. Consider again one particular instance with its corresponding marginal probabilities p*. Because the model is reversible for arbitrary divergence times, in particular there should be some p0 probabilities such that . For this generalized model, the rate matrix is given by . Therefore it follow by Lemma B.3 that the saturation probabilities of R are given by the condition
Therefore the saturation probabilities p∞ are different from p* as long as p0 ≠ p*.
Therefore, we have constructed a parametric family, Qt = Q0etR, in which the marginal probabilities for reversibility are p0 at time zero, p* at t*, and p∞ at time infinity, with p0 ≠ p* ≠ p∞. Therefore if there is reversibility at arbitrary time, the marginals have to be time dependent,
pt (i)Qt(ij) = pt(j)Qt (ji). (216)
In particular in the Section "The evolution of emission probabilities with indels treated as an extra character" we have constructed a system in which the time-dependent reversibility condition (216) is satisfied by marginal probabilities that are quasi-stationary with respect to some (n - 1) p0 probabilities,
pt(i) = p0(i) (1 - Λt), for i = 1,...n - 1 (217)
pt(n) = Λt. (218)
Acknowledgements
Thanks to Sean Eddy for numerous discussions. Thanks to Matt Visser for insights into matrix logarithms. This work was supported by the NIH National Human Genome Research Institute. I would like to acknowledge the Benasque Center for Science in which part of this work took shape at an ESF and NIH funded workshop on computational RNA biology in the summer of 2003.
Figures and Tables
Figure 1 Rate matrix generated from BLOSUM62. Rate matrix obtained from the amino-acid substitution matrix BLOSUM62, rescaled to have an average number of one substitution per amino acid. Notice in bold the two off diagonal negative entries.
Figure 2 Regularized rate matrix generated from BLOSUM62 Regularized rate matrix generated from BLOSUM62 after the QOG algorithm has been applied. The matrix has been rescaled to have an average number of one substitution per amino acid. In this simple case in which there was at most one negative off-diagonal entry per row, the regularization process requires the negative off-diagonal value to be set to zero (represented in bold in this Figure), and to shift the rest of the elements in that row by the corresponding amount so that the sum of all elements is zero. Rows without any off-diagonal negative values remain unchanged from the values obtained in Figure 1.
Figure 3 A pair-HMM model. Description of a pair-HMM model. The three states: Emit-a-pair (XY), Emit-X (X), and Emit-Y (Y) have four possible transitions each, which we are going to make time-dependent functions. This a geometric model, in which the expected length is given by 1/τ. In order to generate alignments with the same expected length at all times, we will leave the parameter τ (the transition of each of the three states into the exit state) unchanged with time. The figure shows the transition probabilities having the following properties: , , , . These properties guarantee that the model is reversible.
Figure 4 Part of a profile HMM model. For a profile HMM we depict the transition probabilities associated with the states of a given consensus position in the profile: Match (M), Insert (I), and Delete (D). The three states corresponding to the next position in the profile are referred to with primes. The Match state has three transitions (into I, M', and D'), while the Insert and Delete states have two transitions each (into I and M', and into D' and M' respectively).
==== Refs
Durbin R Eddy SR Krogh A Mitchison GJ Biological Sequence Analysis: Probabilistic Models of Proteins and Nucleic Acids 1998 Cambridge UK: Cambridge University Press
Krogh A Brown M Mian IS Sjolander K Haussler D Hidden Markov models in computational biology: Applications to protein modeling J Mol Biol 1994 235 1501 1531 8107089 10.1006/jmbi.1994.1104
Karplus K Barrett C Hughey R Hidden Markov models for detecting remote protein homologies Bioinformatics 1998 14 846 856 9927713 10.1093/bioinformatics/14.10.846
Eddy SR Profile hidden Markov models Bioinformatics 1998 14 755 763 9918945 10.1093/bioinformatics/14.9.755
Eddy SR Rawlings C, Clark D, Altman R, Hunter L, Lengauer T, Wodak S Multiple Alignment Using Hidden Markov Models Proc Third Int Conf Intelligent Systems for Molecular Biology 1995 Menlo Park, CA: AAAI Press 114 120
Burge CB Karlin S Finding the Genes in Genomic DNA COSB 1998 8 346 354 10.1016/S0959-440X(98)80069-9
Cawley SL Pachter L HMM sampling and applications to gene finding and alternative splicing Bioinformatics 2003 II36 II41 ii36–ii41 14534169
Meyer IM Durbin R Gene structure conservation aids similarity based gene prediction Nucl Acids Res 2004 32 776 783 14764925 10.1093/nar/gkh211
Sakakibara Y Brown M Hughey R Mian IS Sjolander K Underwood RC Haussler D Stochastic Context-Free Grammars for tRNA Modeling Nucl Acids Res 1994 22 5112 5120 7800507
Eddy SR Durbin R RNA Sequence Analysis Using Covariance Models Nucl Acids Res 1994 22 2079 2088 8029015
Lowe TM Eddy SE tRNAscan-SE: A Program for Improved Detection of Transfer RNA Genes in Genomic Sequence Nucl Acids Res 1997 25 955 964 9023104 10.1093/nar/25.5.955
Eddy SR A Memory-Efficient Dynamic Programming Algorithm for Optimal Alignment of a Sequence to an RNA Secondary Structure BMC Bioinformatics 2002 3 18 12095421 10.1186/1471-2105-3-18
Klein RJ Eddy SR RSEARCH: Finding homologs of single structured RNA sequences BMC Bioinformatics 2003 4 44 14499004 10.1186/1471-2105-4-44
Knudsen B Hein J RNA secondary structure prediction using stochastic context-free grammars and evolutionary history Bioinformatics 1999 15 446 454 10383470 10.1093/bioinformatics/15.6.446
Dowell RD Eddy SR Evaluation of Several Lightweight Stochastic Context-Free Grammars for RNA Secondary Structure Prediction BMC Bioinformatics 2004 5 71 15180907 10.1186/1471-2105-5-71
Rivas E Eddy SR Noncoding RNA gene detection using comparative sequence analysis BMC Bioinformatics 2001 2 8 11801179 10.1186/1471-2105-2-8
Altschul S Gish W Miller W Myers EW Lipman DJ Basic Local Alignment Search Tool Jour Mol Biol 1990 215 403 410 2231712 10.1006/jmbi.1990.9999
Yang Z Estimating the pattern of nucleotide substitution J Mol Evol 1994 39 105 111 8064867
Goldman N Thorne JL Jones DT Using evolutionary trees in protein secondary structure prediction and other comparative sequence analyses J Mol Biol 1996 263 196 208 8913301 10.1006/jmbi.1996.0569
Muse SV Estimating synonymous and nonsynonymous substitution rates Mol Biol Evol 1996 13 105 114 8583885
Whelan S Goldman N general empirical model of protein evolution derived from multiple protein families using a maximum likelihood approach Mol Biol Evol 2001 18 691 699 11319253
Smith AD Lui TW Tillier ER Empirical models for substitution in ribosomal RNA Mol Biol Evol 2004 21 419 421 14660689 10.1093/molbev/msh029
Knudsen B Andersen ES Damgaard C Kjems J Gorodkin J Evolutionary rate variation and RNA secondary structure prediction Comput Biol Chem 2004 28 219 226 15261152 10.1016/j.compbiolchem.2004.04.001
Yang Z A space-time process model for the evolution of DNA sequences Genetics 1995 139 993 1005 7713447
Felsenstein J Churchill GA Hidden Markov Model approach to variation among sites in rate of evolution Mol Biol Evol 1996 13 93 104 8583911
Gribskov M Veretnik S Identification of sequence pattern with profile analysis Methods Enzymol 1996 266 198 212 8743686
Coin L Durbin R Improved techniques for the identification of pseudogenes Bioinformatics 2004 I94 I100 15262786 10.1093/bioinformatics/bth942
McAuliffe JD Pachter L Jordan MI Multiple-sequence functional annotation and the generalized hidden Markov phylogeny Bioinformatics 2004 20 1850 1860 14988105 10.1093/bioinformatics/bth153
Siepel A Haussler D Combining phylogenetic and hidden Markov models in biosequence analysis J Comput Biol 2004 11 413 428 15285899 10.1089/1066527041410472
Thorne JL Kishino H Felsenstein J An evolutionary model for maximum likelihood alignment of DNA sequences J Mol Evol 1991 33 114 124 1920447
Thorne JL Kishino H Felsenstein J Inching toward reality: an improved likelihood model of sequence evolution J Mol Evol 1992 34 3 16 1556741 10.1007/BF00163848
Metzler D Statistical alignment based on fragment insertion and deletion models Bioinformatics 2003 19 490 499 12611804 10.1093/bioinformatics/btg026
Miklos I Lunter GA Holmes I "Long Indel" model for evolutionary sequence alignment Mol Biol Evol 2004 21 529 540 14694074 10.1093/molbev/msh043
Mitchison GJ Durbin RM Tree-based maximal likelihood substitutions matrices and hidden Markov models J Mol Evol 1995 41 1139 11351 10.1007/BF00173195
Mitchison GJ probabilistic treatment of phylogeny and sequence alignment J Mol Evol 1999 49 11 22 10368429
Holmes I Bruno W Evolutionary HMMs: a bayesian approach to multiple alignment Bioinformatics 2001 17 803 820 11590097 10.1093/bioinformatics/17.9.803
Qian B Goldstein RA Detecting distant homologs using phylogenetic tree-based HMMs Proteins 2003 52 446 453 12866055 10.1002/prot.10373
Holmes I Using guide trees to construct multiple-sequence evolutionary HMMs Bioinformatics 2003 Suppl 1 147 157 10.1093/bioinformatics/btg1019
Knudsen B Miyamoto MM Sequence alignments and pair hidden Markov models using evolutionary history J Mol Biol 2003 333 453 460 14529629 10.1016/j.jmb.2003.08.015
Pedersen JS Hein J Gene finding with a hidden Markov model of genome structure and evolution Bioinformatics 2003 19 219 227 12538242 10.1093/bioinformatics/19.2.219
Holmes I A probabilistic model for the evolution of RNA structure BMC Bioinformatics 2004 5 166 15507142 10.1186/1471-2105-5-166
Jukes TH Cantor C Evolution of protein molecules Mamm Prot Met 1965 Academic Press 21 132
Kimura M A simple method for estimating evolutionary rates of base substitutions through comparative studies of nucleotide sequences J Mol Evol 1980 16 111 120 7463489
Tavaré S Some probabilistic and statistical problems in the analysis of DNA sequences Lectures on Mathematics in the Life Sciences 1986 17 57 86
Yang Z Nielsen R Hasegawa M Models of amino acid substitution and applications to mitochondrial protein evolution Mol Biol Evol 1998 15 1600 1611 9866196
Kosiol C Goldman N Buttimore NH new criterion and method for amino acid classification J Theor Biol 2004 228 97 106 15064085 10.1016/j.jtbi.2003.12.010
Yang Z Nielsen R Goldman N Pedersen A Codon-substitution models for heterogeneous selection pressure at amino acid sites Genetics 2000 155 431 449 10790415
Hasegawa M Kishino H Yano T Dating of the human-ape splitting by a molecular clock of mitochondrial DNA J Mol Evol 1985 21 160 174
Holmes I Rubin GM An expectation maximization algorithm for training hidden substition models J Mol Biol 2002 317 757 768 10.1006/jmbi.2002.5405
Müller T Spang R Vingron M Estimating amino acid substitution models: a comparison of Dayhoff's estimator, the resolvent approach and a maximum likelihood methods Mol Biol Evol 2002 19 8 13 11752185
Henikoff S Henikoff JG Amino acid substitution matrices from protein blocks Proc Natl Acad Sci 1992 89 10915 10919 1438297
Kishino H Miyata T Hasegawa M Maximum likelihood inference of protein phylogeny and the origin of chloroplasts J Mol Evol 1990 31 151 160
Dayhoff M Schwartz R Orcutt B model of evolutionary change in protein Atlas Prot Seq Struct 1978 5 345 352
Müller T Vingron M Modeling amino acid replacement J Comp Biol 2000 7 761 776 10.1089/10665270050514918
Kosiol C Goldman N Different Versions of the Dayhoff Rate Matrix Mol Biol Evol 2004 22 193 199 15483331 10.1093/molbev/msi005
Israel RB Rosenthal JS Wei JZ Finding generators for Markov chains via empirical transition matrices, with applications to credit rating Mathematical Finance 2001 11 245 265 10.1111/1467-9965.00114
Kreinin A Sidelnikova M Regularization algorithms for transition matrices Algo Res Quartely 2001 4 23 40
Felsenstein J PHYLIP (Phylogeny Inference Package) version 3.6. Distributed by the author Department of Genome Sciences, University of Washington, Seattle 2004
Swofford DL PAUP*. Phylogenetic Analysis Using Parsimony (*and Other Methods). Version 4 Sinauer Associates, Sunderland, Massachusetts 2003
Adachi J Hasegawa M MOLPHY programs for molecular phylogenetics version 2.3 Institute of Statistical Mathematics, Tokyo 1995
Yang Z PAML: a program package for phylogenetic analysis by maximum likelihood Comput Appl Biosci 1997 13 555 556 9367129
Liò P Goldman N Thorne JL Jones3 DT PASSML: combining evolutionary inference and protein secondary structure prediction Bioinformatics 1998 14 726 733 9789098 10.1093/bioinformatics/14.8.726
Ronquist F Huelsenbeck JP MRBAYES: Bayesian inference of phylogenetic trees Bioinformatics 2001 17 754 755 11524383 10.1093/bioinformatics/17.8.754
Cai W Pei J Grishin NV Reconstruction of ancestral protein sequences and its applications BMC Evol Biol 2004 4 33 15377393 10.1186/1471-2148-4-33
Siepel A Haussler D Phylogenetic estimation of context-dependent substitution rates by maximum likelihood Mol Biol Evol 2004 21 468 488 14660683 10.1093/molbev/msh039
Lunter G Hein J A nucleotide substitution model with nearest-neighbour interactions Bioinformatics 2004 I216 I223 15262802 10.1093/bioinformatics/bth901
Goldman N Whelan S A novel use of equilibrium frequencies in models of sequence evolution Mol Biol Evol 2002 19 1821 1831 12411592
Whelan S Goldman N Estimating the frequency of events that cause multiple-nucleotide changes Genetics 2004 167 2027 2043 15342538 10.1534/genetics.103.023226
Campbell SL Meyer CDJ Generalized Inverses of Linear Transformations 1991 New York: Dover
Jodár L Law AG Rezazadeh A Watson JH Wu G Computations for the Moore-Penrose and Other Generalized Inverses Congress Numer 1991 80 57 64
Felsenstein J Evolutionary Trees from DNA Sequences: A Maximum Likelihood Approach J Mol Evol 1981 17 368 376 7288891
Bronson R Matrix operations 1973 New York: McGraw-Hill
| 15780137 | PMC1087829 | CC BY | 2021-01-04 16:02:47 | no | BMC Bioinformatics. 2005 Mar 21; 6:63 | utf-8 | BMC Bioinformatics | 2,005 | 10.1186/1471-2105-6-63 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-661578413910.1186/1471-2105-6-66Research ArticleDIALIGN-T: An improved algorithm for segment-based multiple sequence alignment Subramanian Amarendran R [email protected] Jan [email protected] Michael [email protected] Burkhard [email protected] University of Tübingen, Wilhelm-Schickard-Institut für Informatik, Sand 13, 72076 Tübingen, Germany2 University of Göttingen, Institute of Microbiology and Genetics, Goldschmidtstr. 1, 37077 Göttingen, Germany2005 22 3 2005 6 66 66 1 11 2004 22 3 2005 Copyright © 2005 Subramanian 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 present a complete re-implementation of the segment-based approach to multiple protein alignment that contains a number of improvements compared to the previous version 2.2 of DIALIGN. This previous version is superior to Needleman-Wunsch-based multi-alignment programs on locally related sequence sets. However, it is often outperformed by these methods on data sets with global but weak similarity at the primary-sequence level.
Results
In the present paper, we discuss strengths and weaknesses of DIALIGN in view of the underlying objective function. Based on these results, we propose several heuristics to improve the segment-based alignment approach. For pairwise alignment, we implemented a fragment-chaining algorithm that favours chains of low-scoring local alignments over isolated high-scoring fragments. For multiple alignment, we use an improved greedy procedure that is less sensitive to spurious local sequence similarities. To evaluate our method on globally related protein families, we used the well-known database BAliBASE. For benchmarking tests on locally related sequences, we created a new reference database called IRMBASE which consists of simulated conserved motifs implanted into non-related random sequences.
Conclusion
On BAliBASE, our new program performs significantly better than the previous version of DIALIGN and is comparable to the standard global aligner CLUSTAL W, though it is outperformed by some newly developed programs that focus on global alignment. On the locally related test sets in IRMBASE, our method outperforms all other programs that we evaluated.
==== Body
Background
Traditional approaches to multiple sequence alignment are either global or local methods. Global methods align sequences from the beginning to the end [4,24,9]. Based on the Needleman-Wunsch objective function [18], these algorithms define the score of an alignment by adding up scores of individual residue pairs and by imposing gap penalties; they try to find an alignment with maximum total score in the sense of this definition. By contrast, most local methods try to find one or several conserved motifs shared by all of the input sequences [29,12,5].
During the last years, a number of hybrid methods have been developed that combine global and local alignment features [17,19,2,8]. One of these methods is the segment-based approach to multiple alignment [17] where alignments are composed from pairwise local sequence similarities. Altogether, these similarities may cover the entire input sequences – in which case a global alignment is produced – but they may as well be restricted to local motifs if no global homology is detectable. Thus, this approach can return global or local alignments – or a combination of both – depending on the extent of similarity among the input sequences.
Instead of comparing single residue pairs, the segment-based approach compares entire substrings of the input sequences to each other. The basic building-blocks for pairwise and multiple alignment are un-gapped pairwise local alignments involving two of the sequences under consideration. Such local alignments are called fragment alignments or fragments; they may have any length up to a certain maximum length M. Thus, a fragment f corresponds to a pair of equal-length substrings of two of the input sequences. Pair-wise or multiple alignments are composed of such fragments; the algorithm constructs a suitable collection A of fragments that is consistent in the sense that all fragments from A can be represented simultaneously in one output multiple alignment.
Note that, since multiple alignments are composed of local pairwise alignments, conserved motifs are not required to involve all of the input sequences. Unlike standard algorithms for local multiple alignment, the segment-based approach is therefore able to detect homologies shared by only two of the aligned sequences. With its capability to deal with both, globally and locally related sequence sets and with its ability to detect local similarities involving only a subset of the input sequences, the segment approach is far more flexible than standard methods for multiple alignment. It can be applied to sequence families that are not alignable by those standard methods; this is the main advantage of segment-based alignment compared to more traditional alignment algorithms. The previous implementation of the segment-based multi-alignment approach is DIALIGN 2.2 [16].
During recent years, systematic studies have been carried out on real and artificial benchmark data sets to evaluate the accuracy of multi-alignment programs [26,11,20]. These studies concluded that DIALIGN is superior to other programs if sequence sets with local homologlis are to be aligned. On sequences with weak but global homology, however, the previous implementation of the program is often out-performed by purely global methods such as CLUSTAL W [24], by hybrid medthods like T-COFFEE [19] or POA [13], or by the recently developed programs MUSCLE [8] and PROBCONS [6] that are currently the best-performing methods for global multiple protein alignment. In the next section, we show that the inferiority of DIALIGN 2.2 on weakly but globally related sequence sets is due to the objective function used by the program. If the program can choose between (a) a global pairwise alignment consisting of many fragments with low individual fragment scores and (6) an alternative local alignment consisting of only a few isolated fragments with higher individual scores, it tends to prefer the second type of alignment over the first one. Consequently, for sequences with weak but global similarity, DIALIGN is vulnerable to spurious random-similarities.
In this paper, we describe a complete re-implementation of the DIALIGN algorithm that overcomes some of the shortcomings of the previous program version 2.2. The paper is organised as follows: in the next section, we discuss the objective function that DIALIGN uses to assess the quality of different alignments for a given input data set. We show that this objective function systematically over-estimates isolated local alignments compared with alternative alignments that would extend over the entire length of the sequences. Next, we introduce two heuristics for pairwise and multiple alignment, respectively, to counter-balance this bias towards isolated local similarities. Then we describe additional features of our new implementation, and in in the section Results and discussion, we evaluate our software tool and compare it to the previous implementation of DIALIGN and to other standard multi-alignment programs.
Objective functions for sequence alignment
From a computer scientist's point-of-view, sequence alignment is an optimisation problem. Most alignment algorithms are – explicitly or implicitly – based on an objective function, i.e. on some kind of scoring scheme assigning a quality score to every possible alignment of a given input sequence set. Based on such a scoring scheme, different optimisation algorithms are used to find optimal or near-optimal alignments. For multiple alignment, a variety of optimisation techniques have been proposed. These algorithms differ substantially from each other in view of their computational complexity and in view of their ability to find or approximate numerically optimal alignments. However, the most important feature of an alignment program is not the optimisation algorithm that it uses, but rather the underlying objective function that is used to score possible output alignments. If the objective function is biologically wrong by assigning high scores to biologically meaningless alignments, then even the most efficient optimisation algorithms are only efficient in finding mathematically high-scoring nonsense alignments. With a more realistic objective function, however, even simple-minded heuristics may lead to biologically plausible alignments.
The objective function that we use in the segment-based approach is defined as follows: each possible fragment (segment pair) f is assigned a weight score w(f) depending on the probability P(f) of random occurrence of such a fragment. More precisely, the program uses a similarity function s assigning a score s(a, b) to each possible pair (a, b) of residues. For protein alignment, one of the usual substitution matrices can be used; for alignment of DNA or RNA sequences, the program simply distingues between matches and mismatches. For a fragment f, its Needleman-Wunsch score NW [f] is calculated which is defined as the sum of similarity values of aligned nucleotides or amino acid residues (note again, that fragments do not contain gaps). To define the weight score w(f) of f, we consider the probability P(f) of finding a fragment f' of the same length as f and with a Needleman-Wunsch score NW [f'] ≥ NW [f] in random sequences of the same length as the input sequences. w(f) is then defined as the negative logarithm of this probability; see [14] for more details. The total score of a – pairwise or multiple – alignment is defined as the sum of weight scores of the fragments it is composed of; gaps are not penalised. The idea is that the less likely a given fragment collection is to occur just by chance, the more likely it is to be biologically relevant so the higher its score should be. Thus, while standard alignment approaches try to find an alignment that is most likely under the assumption that the input sequences are related by common ancestry [7], we try to find an alignment that is most unlikely under the assumption that the sequences are not related. A pairwise alignment in the sense of the above definition corresponds to a chain of fragments, and an alignment with maximum total weight score can be found using a recursive fragment-chaining procedure [15]; for multiple alignment, a greedy heuristics is used [1,14].
As explained above, DIALIGN defines the score S(A)of an alignment A = {f1,..., fk} as the sum of weight scores w(fi) of its constituent fragments, and these weight scores are, in turn, defined as negative logarithms of probabilities P(fi) of their random occurrence. Thus, the score S(A)is calculated as
and searching for an alignment with maximal score is equivalent to searching for a consistent collection of fragments A = {f1,..., fk} with minimal product of probabilities ∏f∈A P(f). But considering the product of fragment probabilities means to consider the probability of their joint occurrence under the assumption that these events are independent of each other. This would be reasonable if we would search for an arbitrary fragment collection with low probability of random occurrence. In our approach, however, we require a fragment collection to be consistent, so the set of allowed combinations of fragments is drastically reduced. The probability of finding a consistent set of fragments is consequently far smaller than the product of the probabilities of finding all of the corresponding individual fragments. Thus, by using the product ∏f∈A P(f), DIALIGN generally over-estimates the probability P(A) of random occurrence of an alignment A.
In our context, the crucial point is that the probabilities P(A) – and therefore the scores S(A) – are not uniformly over-estimated – or under-estimated, respectively – for all possible alignments, but there is wide difference between global and local alignments. For a global alignment Ag that covers most of the sequences, the discrepancy between the real probability P(Ag) of its random occurrence and the approximation P(f) used by DIALIGN is far more significant than for a local alignment Al. This is because a global alignment corresponds to a dense collection of fragment, so here the consistency constraints are much tighter than in a local alignment consisting of only a few isolated fragments. As a result, DIALIGN relatively over-estimates the probability P(Ag) of a global aligment Ag compared with an alternative local alignment Al, so it under-estimates the score S(Ag) compared with the score S(Al).
Reducing the influence of isolated local similarities
In the previous section, we explained why the objective function used in DIALIGN systematically prefers local alignments over alternative global alignments of the same data set. An improved objective function that would use a better approximation to the probability P(A)of random occurrence of an alignment A would have to take into account the combinatorial constraints given by our consistency condition. Defining such an objective function would be mathematically challenging. For our new program, we therefore use the objective function that has been used in previous versions of DIALIGN. However, we introduce two heuristics to counterbalance the bias in this objective function towards isolated local alignments.
Excluding low-scoring sub-fragments
The pairwise alignment algorithm that we are using is a modification of the space-efficient fragment-chaining algorithm described in [15]. At each position (i, j) in the comparison matrix, this algorithm considers all fragments (= segment pairs) starting at (i, j) up to a certain maximum length M. For protein alignment, the previous program DIALIGN 2.2 uses a default value of M = 40; M can be reduced to speed-up the program, but this may result in decreased alignment quality. Initially, the length limitation for fragments has been introduced to reduce the program running time; this way the time complexity of the pairwise fragment-chaining algorithm is reduced from O(l3) to O(l2) where l is the maximum length of the two sequences. One might think that increasing the maximum fragment length M would result in improved alignment quality. In fact, we observed that with slightly increased values for M, better alignments were obtained, but with values M > 50, the quality of the produced alignments decreased dramatically.
In systematic test runs, we observed that for large values of M, output alignments often contain long fragments involving a mixture of high-scoring and low-scoring sub-fragments. With an ideal objective function, a single long fragment f containing low-scoring sub-fragments would automatically receive a lower score than the chain of short fragments that would be obtained from f by removing those low-scoring sub-fragments. As a result, output alignments would tend to consist of shorter fragments rather than of longer fragments with low-scoring sub-regions. For reasons explained in the previous section, however, the scoring scheme used by DIALIGN over-estimates single long fragments compared with chains of smaller fragments that would be obtained by removing low-scoring regions from those long fragments.
In our new approach, we use the following heuristics to prevent the algorithm from selecting long fragments with low-scoring sub-regions. We define a length threshold L for low-quality sub-fragments. Sub-fragments of length ≥ L with negative Needleman-Wunsch score are allowed within short fragments but are excluded in fragments of length ≥ T where T <M is a parameter that can be adjusted bu the user. For a pair of input sequences S1 and S2 and given values for the parameters T, M and L, our new algorithm proceeds as follows. Let f(i, j, k) denote the fragment of length k that starts at position i in sequence S1and at position j in sequence S2, respectively. By S1 [k], we denote the k-th character in sequence Si. As in the original DIALIGN algorithm, we traverse the comparison matrix for S1 and S2, and at every position (i, j), we consider fragments starting at this position; suitable fragments are then added to a growing set F of candidate fragments from which the algorithm selects a fragment chain with maximum total score with respect to the underlying objective function [15]. If a region of low quality occurs, the maximum fragment length M(i, j) for fragments starting at (i, j) is reduced from M to T. More formally, we perform the following steps for fragments starting at a fixed position (i, j):
1. Initially, the maximum length for fragments starting at (i, j) is M(i, j) = M.
2. We start with length k = 1, i.e. we consider the fragment f(i, j, 1).
3. If the current fragment length k exceeds M(i, j) then the procedure stops and we continue with fragments starting at (i, j + 1).
4. If the similarity score s(S1 [i + k - 1], S2 [j + k - 1]) of the last residue pair in f(i, j, k) is not negative, we take the fragment f(i, j, k) into account by adding it to the set F and continue with step 7. Otherwise we detected the potential beginning of a low-quality sub-fragment starting at positions i + k - 1 and j + k - 1, respectively.
5. In this case we do a lookahead and calculate the NW-score of the potential low-scoring fragment f(i + k - 1, j + k - 1, L) which is defined as
6. If NW [f(i + k - 1, j + k - 1, L)] < 0, we actually detected a low-quality sub-fragment. If k >T, the procedure stops and no further increasing of k is being considered, otherwise we set M(i, j) = T.
7. The length k is incremented by 1 and we continue with the step 3.
By default, our program uses a length threshold for low-quality sub-fragments of L = 4 and the maximum length of fragments containing such regions of low quality is T = 40. These values have been determined based on systematic test runs on BAliBASE. At this point, we want to mention the impact of the parameters L and T on the quality of the produced output alignments. For example, with values L = 3 or L = 5 the alignment quality is dramatically worsened compared with the default value L = 4.
Our stop criterion for low-scoring sub-fragments not only improves the quality of the resulting alignments but also reduces the program running time. The runtime of our pairwise algorithm is proportional to the number of fragments that are considered for alignment. Thus, the worst-case time complexity is O(l1·l2·M) where l1 and l2 are the lengths of the input sequences. By excluding long fragments with low-scoring sub-fragments, we ignore a large number of fragments that would have been considered for alignment in previous program versions. Therefore, our new heuristics allows us to increase the maximum possible fragment length from M = 40 to M = 100 without excessively increasing the total number of fragments that are to be looked at. A further extension of M is prohibited due to numerical instabilities. Altogether, the resulting alignments can reflect the extension of existing homologies more realistically than the previous version of DIALIGN with only a moderate increase in program running time.
Weight score factors
As mentioned above, DIALIGN uses a greedy optimisation procedure for multiple alignment. The order in which fragments are included into the multiple alignment is determined based on their weight scores. A general problem with this greedy approach is that if a wrong fragment is accepted for multiple alignment, it cannot be removed later on. Note that even a single wrong choice in the greedy procedure can impair the quality of the resulting alignment dramatically. Thus, special care has to be taken to prioritise fragments for the greedy algorithm. We observed that in many cases spurious but high scoring fragments from pairwise alignments are inconsistent with a good overall multiple alignment. Due to their weight scores, however, such fragments may be incorporated into the multiple alignment by the original DIALIGN, thereby leading to output alignments of lower quality.
As explained in the previous section, the weight score of a fragment depends on the probability of its random occurrence in sequences the length of the input sequences. Thus, weight scores are purely based on intrinsic properties of fragments – and on the length of the input sequences – but they do not take into account the context of a fragment within the pairwise alignment. In reality, however, the context of a fragment is crucial to assess its reliability. If a fragment f is part of a high-scoring pairwise alignment, then f is, of course, far more likely to be biologically significant than if the same fragment f would be found isolated in otherwise un-related sequences. Therefore the overall similarity among two sequences should be taken into account if fragments are ranked prior to the greedy procedure.
In our new program, we adopt the following approach: we multiply the weight score of each fragment by the square of the total weight score of the respective sequence pair divided by the overall weight score of all pairwise alignments. Let S1,..., Sn be the input sequences and let f be a fragment involving sequences Si and Sj. Next, let w(Si, Sj) denote the total weight score of the pairwise alignment for Si and Sj – i.e. the sum of weight scores of an optimal chain of fragments – and let W be the total sum of weight scores of all pairwise alignments. That is, we define
We then define the adjusted weight score
and in our greedy algorithm, fragments are sorted according to their adjusted scores w'(f). This way, we prefer fragments belonging to sequence pairs of high similarity over those from weakly related sequence pairs. Altogether, this weight adjustment respects the similarity of the sequence pairs better than the previous method and hence may keep the greedy procedure from adding isolated spurious fragments that would have led to a lower-scoring and biologically less meaningful output alignment. The sorted list of fragments from the optimal pairwise alignments are kept in a binary heap structure that can be updated efficiently when inconsistent fragments are to be removed or modified as explained in the next section.
Further program features
Dealing with inconsistent fragments
In the original DIALIGN approach, an inconsistent fragment f is completely discarded in the greedy procedure – even if just a few residue pairs are inconsistent with the current alignment. In such a situation, it would be of course more sensible to remove only those inconsistent residue pairs from f and to give the remaining sub-fragments a second chance in the greedy selection process. It is easy to see that a fragment f is consistent with an existing alignment A if and only if each pair of aligned residues in f is consistent with A. In our new implementation, we use the following procedure for non-consistent fragments. An inconsistent fragment f is processed from left to right. Starting with the left-most residue pair, we remove all inconsistent residue pairs until we find the first consistent pair p. Next, we consider all consistent residue pairs starting with p until we find again an inconsistent residue pair. This way, we obtain a consistent sub-fragment f' of f for which we calculate the weight score w(f'). By construction, f' is consistent with the existing alignment and could, in principle, be added to the list of accepted fragments.
However, we do not immediately include f' into the growing multiple alignment since the score w(f') might be smaller than the original score w(f). Instead, we insert f' at the appropriate position in our sorted list of fragments depending on its adjusted weight score w'(f'). With the binary heap structure mentioned in the previous section, consistent sub-fragments of inconsistent fragments can be efficiently re-positioned according to their newly calculated adjusted weights. The remainder of f is treated accordingly, i.e. inconsistent residue pairs are removed and the remaining consistent sub-fragments are inserted at appropriate positions in the list of candidate fragments. Note that with our weighting function w, the weight score w(f') of a sub-fragment f' contained in a fragment f can in general be larger than the weight w(f). In the above situation, however, we have necessarily w(f') ≤ w(f') [and therefore w'(f') ≤ w'(f)] since we assumed that f is part of the optimal pairwise alignment of two sequences. If the score w(f') of a sub-fragment of f exceeded w(f), then f' would have been selected for the optimal pairwise alignment instead of f.
Probability estimates
The previous implementation DIALIGN 2.2 uses pre-calculated probability tables to calculate fragment weight scores; these tables are based on the BLOSUM 62 substitution matrix. They have been calculated years ago and are difficult to re-calculate if a user wants to employ another similarity matrix. It is therefore not possible to run DIALIGN 2.2 with substitution matrices other than BLOSUM 62. In our new implementation, we use a rather efficient way to estimate the probabilities that are used for our weight score calculations. We pre-calculated probability tables for a variety of substitution matrices. In addition, the user can re-calculate these tables 'on the fly' for arbitrary matrices with a moderate increase in program running time.
As explained in section Objective functions for sequence alignment, we define the weight score of a fragment f involving sequences Si and Sj as
w(f) = - logP(f)
where P(f) denotes the the probability for the occurence of a fragment f' of the same length as f and with Needleman-Wunsch score NW [f'] ≥ NW [f] in random sequences of the same length as Si and Sj. By random sequences we mean independent identically distributed (iid) sequences where each residue occurs at any position with probability 1/4 for nucleic acid sequences and 1/20 for protein sequences, respectively. In the following, we outline how our program approximates the probabilities P(f).
In a first step, we estimate the probability of finding a fragment f' of length n and with Needleman-Wunsch score NW [f'] ≥ s in random sequences of length 2·n. Note that depends on the underlying substitution matrix but not on the length or composition of the input sequences Si and Sj. The numerical values are estimated as follows:
1. Random experiments are performed to obtain a preliminary estimates for . The experimental values should approximate with sufficient accuracy for values of n and s where enough experimental data are avail able. This is the case if is not too small.
2. For small values of , we first compute the probabilitys P1(s, n) for a single random fragment f' of length n to have a Needleman-Wunsch score NW [f'] ≥ s. P1(s, n) can be easily calculated as a sum of convolution products. Similar as in [14], small values of are estimated using the approximation formula
3. All in all, we define for a given value s by first considering the trivial case n = 1 and then defining for n = 2,..., M:
The described procedure to estimate is computationally demanding. Since the values do not depend on the input sequences, we pre-calculated these probabilities for several standard substitution matrices and stored their values in auxiliary files from where they are retrieved during the program run.
In a second step, we use to estimate the probability P(s, n) for finding a fragment f' of length n with Needleman-Wunsch score NW [f'] ≥ s in sequences the length of the input sequences. This step is computationally less expensive and can therefore be carried out during the program run. Let li and lj be the lengths of the input sequences Si and Sj, respectively. Similar as in [14], we compute P(s, n) as
where PT is a threshold parameter. During a program run, the values P(s, n) are calculated for all possible values of n and s before the pairwise alignment of sequences Si and Sj is carried out. The negative logarithms -log P(s, n) are stored in a look-up table where they are retrieved during the pairwise alignment to define the fragment scores.
We pre-calculated the probabilities for several substitution matrices of the BLOSUM family. To determine the experimental probability values Pexp(s, n), we carried out 106 random experiments for each relevant pair of parameters (s, n). Here, we considered values for n betwen 1 and a maximum fragment length M = 100. Files with the resulting values of values are distributed together with our software package. To calculate P(f), we use a threshold probability PT = 10-8. Our program can also be used to calculate the values for arbitrary user-defined substitution matrices. Calculating these values using 105 random experiments for each value of n and s takes around 20 minutes on a Linux workstation (RedHat 8.0) with an 1.5 Ghz Pentium 4 processor and 512 MB Ram. In our experience, 105 random experiments are sufficient to obtain high-quality probability estimates.
Results and discussion
We evaluated the performance of our program and compared it to alternative multi-alignment software tools using a wide variety of benchmark sequences. As a first set of reference data, we used the well-known BAliBASE 2.1 [25]. BAliBASE has been used in numerous studies to test the accuracy of multiple-protein-alignment software. It should be mentioned that, although some of the reference sequences in BAliBASE contain insertions and deletions of moderate size, BAliBASE is heavily biased towards globally related protein families. All BAliBASE sequences contain homologous core blocks with verified 3D structure; alignment programs are evaluated according to their ability to correctly align these blocks. According to the BAliBASE authors, these core blocks cover 58 % of the residues in the database. However, sequence similarity is clearly not restricted to those regions of verified 3D structure so, in reality, far more than 58 % of the total sequence length are homologous to other sequences in the respective sequence families. Also, the sequences in BAliBASE are not realistic full-length sequences, but they have been truncated by the BAliBASE developers in order to remove non-related parts of the sequences. As a result, BAliBASE consists almost entirely of globally related sequence sets; this is why global alignment programs such as CLUSTAL W perform best on these benchmark data.
To study the performance of alignment programs on locally related sequence sets, Lassmann and Sonnhammer used artificial random sequences with implanted conserved motifs [11]. Random sequences are frequently used to evaluate computational sequence analysis tools; they are particularly useful to study the specificity of a tool, see e.g. [23,10,20]. Unfortunately, the benchmark data by Lassmann and Sonnhammer are not publicly available. Therefore, we set up our own benchmark database for local multiple protein alignment that we called IRMBASE (Implanted Rose Motifs Base).
As Lassmann and Sonnhammer did in their previous study, we produced groups of artificial conserved sequence motifs using the ROSE software tool [23]. ROSE simulates the process of molecular evolution. A set of 'phylogenetically' related sequences is created from a user-defined 'ancestor' sequence according to a phylogenetic tree. During this process sequence characters are randomly inserted, deleted and substituted under a pre-defined stochastic model. This way, a sequence family with known 'evolution' is obtained, so the 'correct' multiple alignment of these sequences is known. Note that these alignments contain mismatches as well as gaps. We inserted families of conserved motifs created by ROSE at randomly chosen positions into non-related random sequences. This way, we produced three reference sets ref1, ref2 and ref3, of artificial protein sequences. Sequences from ref1 contain one motif each and sequences from ref2 and ref3 contain two and three motifs each, respectively. Each reference set consists of 60 sequence families, 30 of which contain ROSE motifs of length 30 while the remaining 30 families contain motifs of length 60. 20 sequence families in each of the reference sets consist of 4 sequences each, another 20 families consist of 8 sequences while the remaining 20 families consist of 16 sequences. In ref1, random sequences of length 400 are added to the conserved ROSE motif while for ref2 and ref3, random seqences of length 500 are added.
For both BAliBASE and IRMBASE, we used two different criteria to evaluate multi-alignment software tools. We used the sum-of-pair score where the percentage of correctly aligned pairs of residues is taken as a quality measure for alignments. In addition, we used the column score where the percentage of correct columns in an alignment is the criterion for alignment quality. Both scoring schemes were restricted to core blocks within the reference sequences where the 'true' alignment is known. For IRMBASE, the core blocks are defined as the conserved ROSE motifs. In general, the sum-of-pairs score is more appropriate than the column score because this latter score ignores all correctly aligned residues in an alignment column if one single residue in this column is mis-aligned. However, there are situations where the column score is more meaningful than the sum-of-pairs score. This is the case, for example, for BAliBASE reference sets containing 'orphan sequences'.
To compare the output of different programs to the respective benchmark alignments, we used C. Notredame's program aln_compare [19]. Tables 1 and 2 summarise the performance of DIALIGN-T, DIALIGN 2.2, CLUSTAL W, MUSCLE, PROBCONS, T-COFFEE and POA on IRMBASE while Tables 3 and 4 show their accuracy on BAliBASE. In addition, Tables 5, 6, 7 and 8 contain the percentage of sequence sets where DIALIGN-T is outperformed by the other programs that we tested. Tables 1 and 2 show that, on locally related sequence families, DIALIGN-T is significantly superior to the algorithms DIALIGN 2.2, T-COFEE, MUSCLE, POA and CLUSTAL W. Only DIALIGN-T, DIALIGN 2.2, T-COFFEE and (in a very reduced way) PROBCONS produced reasonable results on IRMBASE 1.0. However, DIALIGN-T is the fastest and most accurate amongst all methods that we looked at. We would like to emphasize that the performance of multi-alignment methods on simulated data only roughly reflects their performance on real data. Nevertheless, in the absence of real-world benchmark data for local multiple alignment, the results on IRMBASE can give us an idea of how different algorithms deal with locally conserved motifs.
For globally related sequence families, Tables 3 and 4 show that, on average, DIALIGN-T outperforms DIALIGN 2.2 and POA on BAliBASE 2.1 while its performance is similar to CLUSTAL W. By contrast, the previous version DIALIGN 2.2 is clearly outperformed by CLUSTAL W on these data sets. Finally, DIALIGN-T is still outperformed on many of the BAliBASE test sequences by T-COFFEE, MUSCLE and PROBCONS; the latter program is currently the best-performing multiple aligner on BAliBASE. The superiority of our new approach compared to DIALIGN 2.2 and POA is clearly statistically significant according to the Wilcoxon Matched Pairs Signed Rank Test. On BAliBASE reference sets ref1, ref2 and ref3 where sequences contain only small insertions and deletions, the performance of DIALIGN-T is roughly comparable to CLUSTAL W, but yet still significantly worse than T-COFFEE, PROBCONS or MUSCLE. Our program is statistically significantly superior or equal to all tested methods, except MUSCLE and PROBCONS, on the sequence sets with larger insertions or deletions (ref4 and ref5 of BAliBASE).
Overall, the relative performance of the different alignment tools is similar under the two alternative evaluation criteria that we used (sum of pairs and column score) – although, the absolute values of the column scores are, of course, lower than the sum-of-pairs scores. Maybe surprisingly, both versions of DIALIGN are superior to all other programs in our study on the locally related sequences from IRMBASE – while on the other hand, DIALIGN was outperformed by alternative methods on reference sets 4 and 5 of BAliBASE. These sequence sets are also considered locally related because they contain larger insertions and deletions then other BAliBASE sequences. The reason for this apparent discrepancy is that the ref4 and ref5 sequence sets in BAliBASE are not truly locally related, but they still show some similarity outside the conserved core blocks. In IRMBASE, by contrast, sequence similarity is strictly limited to the conserved motifs.
Since we re-implemented the DIALIGN algorithm from scratch and used a variety of novel program features, it is not possible to tell exactly to what extend each of these features contributed to the improved program performance. Systematic test runs with varying parameters indicate, however, that the superiority of our new program compared to the previous program DIALGIN 2.2 on locally as well as on globaly related sequence families is mainly due to the program features explained in the third section. The improvement that we achieved with these heuristics is statistically significant. The features explained in section Further program features also improved the program accuracy, though here the improvement was not statistically significant.
Table 9 shows the program running time for the seven programs that we tested in our study. DIALIGN-T is around 6 % slower than the previous implementation DIALIGN 2.2 on BAliBASE 2.1, but on IRMBASE, DIALIGN-T is approximately 30 % faster than DIALIGN 2.2. In DIALIGN, the CPU time for multiple alignment is mainly spent on pairwise alignments that are performed before fragments are included into the multiple alignment. As explained in section Excluding low-scoring sub-fragments, the runtime for pairwise alignment is roughly proportional to the number of fragments that are considered for alignment and, for sequences of length l1 and l2 and a maximum fragment length M, up to l1 × l2 × M fragments are to be considered. In our new program DIALIGN-T, the maximum fragment length M is increased to 100 compared to 40 for the original DIALIGN program. Nevertheless, the program running time is only slightly increased for the globally related protein families from BAliBASE and considerably decreased for the locally conserved sequences from IRMBASE. This is due to the heuristic stop criterion for fragments introduced above. The slowest program in our comparison was T-COFFEE which is more than eleven times slower than DIALIGN-T on IRMBASE and more than five times slower on BAliBASE. POA was the fastest method. On BAliBASE, the program PROBCONS produces the best results in terms of alignment accuracy. The program is, however, the second slowest program after T-COFFEE on both BAliBASE and IRMBASE. MUSCLE provides so far the best tradeoff between running time and quality on globally related sequence families, but when it comes to local alignments both running time and alignment quality decrease drastically. The memory consumption of our method has been improved compared to DIALIGN 2.2.
With the development of DIALIGN-T, we significantly improved the segment-based approach to multiple protein alignment on both local and global benchmark data. The new heuristics that we introduced, generally favour consistent groups of low-scoring fragments over isolated higher-scoring fragments. This way, we improved the program performance on globally related sequence sets where the segment approach was previously inferior to programs such as CLUSTAL W and POA. On these data sets, our new method is significantly more accurate but only slightly slower than DIALIGN 2.2. On BAliBASE, the performance of our approach is now comparable to the popular global alignment program CLUSTAL W. For locally related protein families, DIALIGN-T performs significantly better and is also considerably faster than the previous DIALIGN 2.2 which was, so far, the best available method on locally related protein families. In addition to these improvements, it is now possible to use arbitrary user-defined substitution matrices which was not possible for the original DIALIGN program. To further enhance the performance of our method, we are planning to improve the greedy algorithm that DIALIGN uses for multiple alignment. Rather than focusing on pairwise fragment alignments, we will develop heuristics that are able to integrate multiple local alignments into the final multiple alignment. This approach should further improve the sensitivity of our methods for locally conserved motifs.
Finally, we would like to make some general remarks on parameter tuning and program evaluation in multiple sequence alignment. As mentioned above, we identified suitable values for our parameters T and L based on test runs with BAliBASE, and we assume that this is how the program parameters for most multiple protein aligners have been tuned during the last years. Therefore, the question has been raised if current protein alignment programs are overfitted with respect to BAliBASE. Parameter overfitting is a serious problem for many Bioinformatics algorithms. For example, many gene-prediction programs have a large number of parameters to adjust, so it is easy to tune these programs to perform well on a given set of training data. For such programs it is therefore absolutely necessary to clearly separate training data that are used for parameter tuning from test data that are used to evaluate the program. The situation is totally different in multiple alignment. Most multi aligners have only a very small number of parameters to adjust. For our algorithm, for example, the only important parameters to tune are T and L. BAliBASE, on the other hand, comprises a large variety of test sequences for global multiple alignment. It consists of 139 sequence sets, each of which contains several core blocks, so there is a total of several hundred core blocks that are used to test alignment quality. It is absolutely impossible to tune a small number of parameters in such a way that they work well only on BAliBASE but not on other globally related protein sequences. Thus, if an alignment program performs well on BAliBASE, one can safely assume that it also works well on other globally related protein sequences, even if BAliBASE has been used to adjust its parameter values. In fact, it turned out that the parameters that we tuned on BAliBASE work not only well for these global test data but also on the totally different artificial local test sequences from IRMBASE.
The real problem with BAliBASE is its heavy bias towards globally related sequence sets. This does not only refer to the selection of protein families that are included into BAliBASE. As mentioned above, many protein sequences in the current release of BAliBASE are not real-world protein sequences, but have been artificially truncated by the developers of BAliBASE in order to make them globally related. With these non-realistic global test sequences, the BAliBASE authors carried out a systematic program evaluation and – not surprisingly – found out that Global alignment programs generally performed better than local methods [26]. The picture could have been totally different if realistic full-length proteins had been used instead of truncated sequences. To counterbalance the bias towards global test sets in BAliBASE, we created an additional benchmark data set consisting of simulated conserved domains embedded in non-related random sequences. The performance of alignment programs on artificial sequences should not be over-estimated as the design of such data sets is necessarily somewhat arbitrary. Nevertheless, our test runs on these simulated data give a rough impression of how different alignment methods perform on locally related data sets. Further systematic studies should be carried out to evaluate the performance of multiple-protein aligners under varying conditions using, for example, the full-length BAliBASE sequences or newly developed benchmark databases such as SABmark [27,28], Prefab [8] or Oxbench [21].
As a concluding remark, we would like to address a fundamental limitation of most multi-alignment methods, including the one presented in this paper: these methods implicitly assume that homologies and conserved motifs occur in the same relative order within the input sequences. There are two major reasons for making this assumption. First, an order-preserving multiple alignment that represents homologies by inserting gap characters into the input sequences provides a convenient visualisation of existing homologies. Second – and more importantly -, the order-preservation constraint greatly reduces the noise created by random similarities. A program that would return all detectable local or global similarities among the input sequences without the above ordering constraints would necessarily return many spurious random similarities. To reduce this noise, arbitrary threshold parameters would have to be applied which, in turn, could prevent a program from detecting some of the real homologies. With the ordering constraint that is implicitly imposed by most alignment programs, weak homologies can be detected, provided they are order-consistent with other detected similarities, i.e. if they fit into one single output alignment. Many evolutionary events such as insertions, deletions and substitutions preserve the relative ordering among sequence homologies. In this situation order-respecting alignment methods are, in principle, able to represent all true biological homologies in one multiple alignment. Nevertheless, for distantly related protein families, non-order-preserving events such as duplications or translocations need to be taken into account. Such events play an important role in comparative analysis of genomic sequences which became an important area of research during the last years [20]. Recently, some promising algorithms for multiple alignment of genomic sequences have been proposed that are able to deal with non-order-conserving evolutionary events [22,3]. Further progress in multiple protein alignment can be expected if these ideas are applied to protein alignment algorithms.
Program availability
DIALIGN-T is available at Göttingen Bioinformatics Compute Server (GOB-ICS):
• Project name: DIALIGN-T
• Project home page:
• Operating system(s): UNIX and LINUX
• Programming language: C
• Other requirements: none
• License: GNU LGPL
• Any restrictions to use by non-academics: none
A program package with functionalities to compute alignments of nucleic acid sequences will be available soon.
Authors' contributions
ARS conceived the new heuristics, implemented the program, constructed IRMBASE, did most of the evaluation and wrote minor parts of the manuscript; JWM participated in program evaluation; MK provided resources; BM supervised the work, provided resources and wrote most of the manuscript.
Acknowledgements
We would like to thank C. Notredame for providing his software tool aln_compare and R. Steinkamp for helping us with the web server at GOBICS. K. Hartwich contributed to the program evaluation. Three unknown referees made very useful comments and helped to improve the manuscript. The work was supported by DFG grant MO 1048/1-1 to BM.
Figures and Tables
Figure 1 Exclusion of low-scoring regions from alignment fragments. The scoring scheme used in DIALIGN gives relatively high weight scores to single fragments with high Needleman-Wunsch scores (a). In our new approach, we exclude low-scoring sub-regions within long fragments by applying a stop criterion for fragment extension. This can result in the replacement of a long fragment f by multiple sub-fragments (b) or in a completely different alignment (c).
Table 1 Performance of seven protein multi-alignment programs on the IRMBASE 1.0 database of benchmark alignments. Percentage values are sum-of-pairs scores, i.e. the percentage of correctly aligned residue pairs of ROSE motifs contained in the IRMBASE sequence families.
Method ref1 ref2 ref3 Total
DIALIGN-T 94.07% 92.69% 92.68% 93.14%
DIALIGN 2.2 92.26% 92.72% 91.87% 92.28%
T-COFFEE 1.37 91.18% 85.61% 87.81% 88.20%
PROBCONS 1.09 66.74% 68.30% 77.92% 70.98%
POA V2 90.26% 43.61% 36.85% 56.91%
MUSCLE 3.5 36.16% 37.84% 52.30% 42.10%
CLUSTAL W 1.83 8.02% 12.69% 20.16% 13.62%
Table 2 Performance of seven protein multi-alignment programs on IRMBASE using column scores as quality criterion. Thus, percentage values denote the percentage of correct alignment columns of the ROSE motifs in IRMBASE
Method ref1 ref2 ref3 Total
DIALIGN-T 82.28% 78.36% 79.71% 80.12%
DIALIGN 2.2 79.46% 77.82% 78.24% 78.51%
T-COFFEE 1.37 75.35% 66.60% 69.21% 70.19%
PROBCONS 1.09 33.13% 37.95% 51.26% 40.78%
POA V2 73.00% 12.46% 07.45% 30.97%
MUSCLE 3.5 09.41% 10.89% 22.37% 14.22%
CLUSTAL W 1.83 00.00% 00.83% 05.14% 01.92%
Table 3 Performance of seven protein multi-alignment programs on the BAliBASE benchmark database using sum-of-pairs scores as evaluation criterion.
Method ref1 ref2 ref3 ref4 ref5 Total
DIALIGN-T 82.76% 91.28% 75.34% 86.43% 93.30% 84.69%
DIALIGN 2.2 81.40% 89.56% 68.93% 91.24% 94.14% 83.59%
T-COFFEE 1.37 84.67% 93.24% 80.32% 75.80% 96.20% 85.95%
PROBCONS 1.09 90.37% 94.61% 84.34% 89.20% 98.07% 91.11%
POA V2 74.66% 88.32% 63.14% 82.62% 76.71% 76.76%
MUSCLE 3.5 88.25% 93.59% 82.36% 85.62% 97.80% 89.21%
CLUSTAL W 1.83 86.43% 93.22% 75.79% 81.09% 86.10% 86.15%
Table 4 Performance of seven protein multi-alignment programs on BAliBASE using column scores.
Method ref1 ref2 ref3 ref4 ref5 Total
DIALIGN-T 73.22% 43.43% 44.69% 66.13% 77.05% 65.65%
DIALIGN 2.2 71.49% 37.42% 35.03% 81.88% 84.47% 64.82%
T-COFFEE 1.37 75.32% 53.44% 52.20% 45.09% 86.96% 68.20%
PROBCONS 1.09 83.21% 59.76% 61.34% 71.09% 91.86% 77.23%
POA V2 63.21% 39.02% 25.57% 57.22% 47.18% 54.18%
MUSCLE 3.5 80.79% 56.37% 56.74% 62.65% 91.57% 74.13%
CLUSTAL W 1.83 78.39% 56.24% 48.87% 50.44% 63.89% 68.48%
Table 5 Percentage of sequence families where DIALIGN-T is outperformed on IRMBASE 1.0 by alternative methods according to the sum-of-pairs score. The symbol + denotes statistically significant superiority, - statistically significant inferiority and 0 non-significant superiority or inferiority of DIALIGN-T, respectively. Significance has been calculated according to the Wilcoxon Matched Pairs Signed Rank Test with p ≤ 0.05).
Method ref1 ref2 ref3 Total
DIALIGN 2.2 20.00%+ 23.33%0 23.33%+ 22.22%+
T-COFFEE 1.37 40.00%0 31.67%+ 41.67%+ 37.78%+
PROBCONS 1.09 20.00%+ 15.00%+ 21.67%+ 18.89%+
POA V2 16.67%+ 0.00%+ 0.00%+ 5.55%+
MUSCLE 3.5 5.00%+ 5.00%+ 0.00%+ 3.33%+
CLUSTAL W 1.83 0.00%+ 0.00%0 0.00%0 0.0%+
Table 6 Percentage of sequence families where DIALIGN-T is outperformed on IRMBASE 1.0 by other methods according to the column score. Notation is as in Table 5.
Method ref1 ref2 ref3 Total
DIALIGN 2.2 11.67%+ 21.67%0 23.33%+ 18.89%+
T-COFFEE 1.37 36.67%0 30.00%+ 26.67%+ 31.11%+
PROBCONS 1.09 18.33%+ 01.67%+ 16.67%+ 16.67%+
POA V2 15.00%+ 00.00%+ 00.00%+ 05.00%+
MUSCLE 3.5 05.00%+ 05.00%+ 00.00%+ 03.33%+
CLUSTAL W 1.83 00.00%+ 00.00%+ 00.00%+ 00.00%+
Table 7 Percentage of sequence families where DIALIGN-T is outperformed on BAliBASE 2.1 by other methods according to the sum-of-pairs score. Notation is as in Table 5.
Method ref1 ref2 ref3 ref4 ref5 Total
DIALIGN 2.2 28.05%+ 21.74%+ 16.67%+ 16.67%0 41.67%0 26.24%+
T-COFFEE 1.37 58.54%- 86.96%- 75.00%- 25.00%0 50.00%0 60.99%-
PROBCONS 1.09 71.95%- 82.61%- 100.00%- 33.33%0 75.00%- 80.14%-
POA V2 20.73%+ 34.78%+ 16.67%+ 33.33%0 0.00%+ 21.99%+
MUSCLE 3.5 71.95%- 73.91%- 83.33%- 25.00%0 75.00%- 69.50%-
CLUSTAL W 1.83 53.66%- 56.52%0 58.33%0 16.67%0 8.33%+ 47.52%0
Table 8 Percentage of sequence families where DIALIGN-T is outperformed on BAliBASE 2.1 by other methods according to the column score. Notation as in Table 5.
Method ref1 ref2 ref3 ref4 ref5 Total
DIALIGN 2.2 26.83%+ 13.04%+ 16.67%+ 16.67%0 50.00%0 24.82%+
T-COFFEE 1.37 56.10%- 73.91%- 66.67%0 25.00%0 50.00%0 56.74%-
PROBCONS 1.09 80.49%- 82.61%- 75.00%- 25.00%0 66.67%- 74.47%-
POA V2 20.73%+ 26.09%0 08.33%+ 16.67%0 00.00%+ 18.44%+
MUSCLE 3.5 73.17%- 73.91%- 83.33%- 16.67%0 66.67%- 68.79%-
CLUSTAL W 1.83 52.44%- 69.57%- 50.00%0 16.67%0 08.33%+ 48.23%0
Table 9 Average running time (in seconds) per multiple alignment for the 180 sequence families of IRMBASE and for 141 sequence families in BAliBASE 2.1. Program runs were performed on a Linux workstation (RedHat 8.0) with an 3.2 GHz Pentium 4 processor and 2 GB Ram.
Method Average runtime on IRMBASE 1.0 Average runtime on BAliBASE 2.1
DIALIGN-T 2.36 1.38
DIALIGN 2.2 3.33 1.30
T-COFFEE 1.37 27.54 7.64
PROBCONS 1.09 12.37 2.66
POA V2 1.44 0.58
MUSCLE 3.5 9.37 0.60
CLUSTAL W 1.83 1.41 0.47
==== Refs
Abdeddaïm S Morgenstern B Speeding up the DIALIGN multiple alignment program by using the 'greedy alignment of biological sequences library' (GABIOS-LIB) Lecture Notes in Computer Science 2001 2066 1 11
Brudno M Chapman M Göttgens B Batzoglou S Morgenstern B Fast and sensitive multiple alignment of large genomic sequences BMC Bioinformatics 2003 4 66 14693042 10.1186/1471-2105-4-66
Brudno M Malde S Poliakov A Do CB Couronne O Dubchak I Batzoglou S Glocal alignment: finding rearrangements during alignment Bioinformatics 2003 i54 i62 12855437 10.1093/bioinformatics/btg1005
Corpet F Multiple sequence alignment with hierarchical clustering Nucleic Acids Res 1988 16 10881 10890 2849754
Depiereux E Feytmans E Match-box: a fundamentally new algorithm for the simultaneous alignment of several protein sequences CABIOS 1992 8 501 509 1422884
Do C Brudno M Batzoglou S ProbCons: probabilistic consistency-based multiple alignment of amino acid sequences Proceedings Nineteenth National Conference on Artificial Intelligence 2004 703 708
Durbin R Eddy SR Krogh A Mitchison G Biological sequence analysis 1998 Cambridge University Press, Cambridge, UK
Edgar R MUSCLE: Multiple sequence alignment with high score accuracy and high throughput Nuc Acids Res 2004 32 1792 1797 10.1093/nar/gkh340
Gotoh O Significant improvement in accuracy of multiple protein sequence alignments by iterative refinement as assessed by reference to structural alignments J Mol Biol 1996 264 823 838 8980688 10.1006/jmbi.1996.0679
Guigó R Agarwal P Abril JF Burset M Fickett JW An assessment of gene prediction accuracy in large DNA sequences Genome Research 2002 10 1631 1642 10.1101/gr.122800
Lassmann T Sonnhammer EL Quality assessment of multiple alignment programs FEBS Letters 2002 529 126 130 12354624 10.1016/S0014-5793(02)03189-7
Lawrence CE Altschul SF Boguski MS Liu JS Neuwald AF Wootton JC Detecting subtle sequence signals: a gibbs sampling strategy for multiple alignment Science 1993 262 208 14 8211139
Lee C Grasso C Sharlow MF Multiple sequence alignment using partial order graphs Bioinformatics 2002 18 452 464 11934745 10.1093/bioinformatics/18.3.452
Morgenstern B DIALIGN 2: improvement of the segment-to-segment approach to multiple sequence alignment Bioinformatics 1999 15 211 218 10222408 10.1093/bioinformatics/15.3.211
Morgenstern B A simple and space-efficient fragment-chaining algorithm for alignment of DNA and protein sequences Applied Mathematics Letters 2002 15 11 16 10.1016/S0893-9659(01)00085-4
Morgenstern B DIALIGN: Multiple DNA and protein sequence alignment at BiBiServ Nucleic Acids Research 2004 32 W33 W36 15215344 10.1093/nar/gnh029
Morgenstern B Dress A Werner T Multiple DNA and protein sequence alignment based on segment-to-segment comparison Proc Natl Acad Sci USA 1996 93 12098 12103 8901539 10.1073/pnas.93.22.12098
Needleman SB Wunsch CD A general method applicable to the search for similarities in the amino acid sequence of two proteins J Mol Biol 1970 48 443 453 5420325 10.1016/0022-2836(70)90057-4
Notredame C Higgins D Heringa J T-Coffee: a novel algorithm for multiple sequence alignment J Mol Biol 2000 302 205 217 10964570 10.1006/jmbi.2000.4042
Pollard DA Bergman CM Stoye J Celniker SE Eisen MB Benchmarking tools for the alignment of functional noncoding DNA BMC Bioinformatics 2004 5 6 14736341 10.1186/1471-2105-5-6
Raghava G Searle SM Audley PC Barber JD Barton GJ OXBench: A benchmark for evaluation of protein multiple sequence alignment accuracy BMC Bioinformatics 2003 4 47 14552658 10.1186/1471-2105-4-47
Raphael B Zhi D Tang H Pevzner P A novel method for multiple alignment of sequences with repeated and shuffled elements Genome Research 2004 14 2336 2346 15520295 10.1101/gr.2657504
Stoye J Evers D Meyer F Rose: Generating sequence families Bioinformatics 1998 14 157 163 9545448 10.1093/bioinformatics/14.2.157
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 Research 1994 22 4673 4680 7984417
Thompson JD Plewniak F Poch O BAliBASE: A benchmark alignment database for the evaluation of multiple sequence alignment programs Bioinformatics 1999 15 87 88 10068696 10.1093/bioinformatics/15.1.87
Thompson JD Plewniak F Poch O A comprehensive comparison of protein sequence alignment programs Nucleic Acids Research 1999 27 2682 2690 10373585 10.1093/nar/27.13.2682
Walle IV Lasters I Wyns L Align-m – a new algorithm for multiple alignment of highly divergent sequences Bioinformatics 2004 20 1428 1435 14962914 10.1093/bioinformatics/bth116
Walle IV Lasters I Wyns L SABmark – a benchmark for sequence alignment that covers the entire known fold space Bioinformatics doi: 10.1093/bioinformatics/bth493.
Waterman MS Multiple sequence alignment by consensus Nucleic Acids Res 1986 14 9095 9102 3786145
| 15784139 | PMC1087830 | CC BY | 2021-01-04 16:02:48 | no | BMC Bioinformatics. 2005 Mar 22; 6:66 | utf-8 | BMC Bioinformatics | 2,005 | 10.1186/1471-2105-6-66 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-761579038810.1186/1471-2105-6-76Research ArticleAn Entropy-based gene selection method for cancer classification using microarray data Liu Xiaoxing [email protected] Arun [email protected] Adrian [email protected] Bioinformatics Institute, 30, Biopolis Street, #07-01, (S) 138671, Singapore2005 24 3 2005 6 76 76 13 7 2004 24 3 2005 Copyright © 2005 Liu 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
Accurate diagnosis of cancer subtypes remains a challenging problem. Building classifiers based on gene expression data is a promising approach; yet the selection of non-redundant but relevant genes is difficult.
The selected gene set should be small enough to allow diagnosis even in regular clinical laboratories and ideally identify genes involved in cancer-specific regulatory pathways. Here an entropy-based method is proposed that selects genes related to the different cancer classes while at the same time reducing the redundancy among the genes.
Results
The present study identifies a subset of features by maximizing the relevance and minimizing the redundancy of the selected genes. A merit called normalized mutual information is employed to measure the relevance and the redundancy of the genes. In order to find a more representative subset of features, an iterative procedure is adopted that incorporates an initial clustering followed by data partitioning and the application of the algorithm to each of the partitions. A leave-one-out approach then selects the most commonly selected genes across all the different runs and the gene selection algorithm is applied again to pare down the list of selected genes until a minimal subset is obtained that gives a satisfactory accuracy of classification.
The algorithm was applied to three different data sets and the results obtained were compared to work done by others using the same data sets
Conclusion
This study presents an entropy-based iterative algorithm for selecting genes from microarray data that are able to classify various cancer sub-types with high accuracy. In addition, the feature set obtained is very compact, that is, the redundancy between genes is reduced to a large extent. This implies that classifiers can be built with a smaller subset of genes.
==== Body
Background
DNA microarrays have become ubiquitous in analyzing the expression profiles of genes in the hope to distinguish between various disease types, such as discriminating between various cancer sub-types. Differential expression of genes is analyzed statistically and genes are assigned to various classes which may (or not) enhance the understanding of underlying biological processes. Alternatively, a reduced set of genes may be singled out and used as biomarkers for diagnosis and prognosis.
Microarray data is typically used both to discover new classes as well as in class prediction. Discovery of new classes [1-4] is usually achieved with the help of clustering techniques such as hierarchical clustering [5], k-means clustering [6] and self organizing maps (SOM) [7]. Class prediction, involving the assignment of labels to samples based on their expression patterns, is typically based on statistical or supervised machine learning methods. These range from the application of simple techniques such as nearest neighbor algorithms [8] to classical methods such as linear discriminant analysis [9] to more advanced techniques such as neural networks [10], support vector machines [11-13], fuzzy logic [14] and decision trees [15]. The challenge in dealing with microarray data lies in the fact that there are orders of magnitude differences between the number of samples (typically less than a hundred) and the number of genes (typically tens of thousands) that are studied. The measurements also typically contain both measurement noise as well as systemic noise. This could have a significant impact on classification accuracy. Classification must therefore be preceded by a step known as feature selection where a subset of relevant features is identified.
There are a number of advantages to feature set selection. The first lies in reducing the cost of clinical diagnosis. It is much cheaper to focus only on the expression of a few genes rather than on thousands of genes for diagnosis [16]. Feature set selection can also lead to a reduction in computational cost as a result of a reduction in problem dimensionality. Furthermore, feature set selection often gives rise to a much smaller and a more compact gene set. This could make it easier to identify genes of particular importance to the problem under study. Moreover, given the disparity in the magnitudes of the numbers of genes and samples, it is difficult to justify the development of a classifier based on a gene set where the number of genes is greater than the number of samples.
One way to categorize feature set selection approaches is to classify them as either filter (such as those based on statistical tests such as t-test, F-test etc.) or wrapper [17] methods. These methods have the advantage of having very low computational complexity as well as better generalization potential since they are uncorrelated to the learning method.
Wrapper type approaches are those in which the feature selection method is bundled together with the learning method. This implies that the usefulness of a feature is validated by the estimated accuracy of the learning method. In consequence, often, a small subset of the feature set with very high prediction accuracy can be obtained because the characteristics of the features match well with the characteristics of the learning method.
Another way of categorizing feature set selection approaches is as univariate or multivariate [18]. Univariate methods [1,19] consider the contributions of individual genes to the classification independently. In contrast multivariate methods such as recursive feature elimination (RFE) [12], leave one out (LOO) method [13], mutual information based approaches [20] etc., measure the relative contribution of a gene to the classification by taking the effect of other genes into consideration at the same time.
A serious deficiency of currently used multivariate approaches for feature set selection is that they are based on selecting genes which are maximally relevant with respect to the classes. The problem with this approach is that there might still be genes among the selected set that are heavily correlated with each other and thus leading to a redundancy in the selected feature set. Ding et. al. [20] have used mutual information for gene selection that has maximum relevance with minimal redundancy by solving a simple two-objective optimization.
In the study presented here, a similar approach has been followed for feature set selection by trying to maximize the relevance and minimize the redundancy of the selected genes. However, normalized mutual information has been used instead of mutual information. In addition, both Battiti's greedy selection algorithm [21] as well as a simulated annealing based approach [22] have been used. In order to find a more representative subset of features, an iterative procedure was adopted that incorporates an initial clustering followed by data partitioning and the application of the algorithm to each of the partitions. A leave-one-out approach then selects the most commonly selected genes across all the different runs and the gene selection algorithm is applied again to pare down the list of selected genes until a minimal subset that gives a satisfactory accuracy of classification is obtained. The algorithm was applied to three different data sets and the results obtained were compared to work done by others using the same data sets. Additionally the algorithm was also compared to work done by Ding and Peng [20] for three different datasets.
Results
Datasets
Three public microarray data sets were used to assess the performance of the algorithm.
SRBCT data
This data set includes 88 cDNA arrays for 63 training samples and 25 test samples from [10]. All samples were combined together and the 5 non-SRBCT samples were removed. The data set consists of four types of tumors in childhood, including Ewing's sarcoma (EWS), rhabdomyosarcoma (RMS), neuroblastoma (NB), and Burkitt lymphoma (BL). After filtering by [10], 2308 genes remained in the data set. The data was transformed to natural logarithmic values. Finally, each sample was also standardized to zero mean and unit variance.
Breast cancer data
This data set contains expression levels of 7129 genes in 49 breast tumor samples from [23]. The samples were classified according to their estrogen receptor (ER) status. 25 samples were ER positive while the other 24 samples were ER negative. In the pre-processing procedure, the data was thresholded with a floor of 100 and a ceiling of 16000 Affymetrix intensity units. Then those genes with ≤ 5 or max - min ≤ 500 were excluded. The filtered data was transformed to base 10 logarithmic values. Finally, each sample was standardized to zero mean and unit variance.
Colon cancer data
This data set contains expression levels of 40 tumor and 22 normal colon tissues. Only the 2000 genes with the highest minimal intensity were selected by [24]. The data waspre-processed by transforming the raw intensities to base 10 logarithmic values and standardizing each sample to zero mean and unit variance.
Results
The results of the application of the full algorithm using both the greedy selection algorithm as well as the simulated annealing algorithm for solving Problem 2 are shown in Table 1. The associated clustering dendrograms are shown in Figures 1, 3 and 5, respectively. For all the dendrograms, the samples are presented along the x-axis with the gene-set along the y-axis. Orange reflects up-expression while yellow represents no or little expression.
The results for SRBCT were the best with a 100% accuracy obtained. The number of genes selected in this case was 58 as opposed to the 96 genes selected by Khan et al. [10]. It is interesting to note that when the binary optimization algorithm was used to select genes for the SRBCT data, 50 of the genes selected were the same as those selected with the greedy algorithm. The accuracy rate for breast cancer data was similar for both cases with about 5 samples being misclassified. The final gene set for this data set contained 31 genes. For colon cancer data, there were 6 mis-classifications, with an overall accuracy rate of 90.3%. There were 29 genes in the final selected gene set.
There seems to be no quantitative or qualitative difference when using the greedy selection or the binary optimization algorithm. Moreover, since the simulated annealing procedure requires an inordinate amount of computation time (of the order of days) as compared to the greedy selection algorithm (of the order of a couple of hours), the iterative procedure was implemented with the greedy algorithm. The iterative approach shown in Figure 8 was used for all three data sets and the clustering dendrograms with the reduced feature sets are shown in Figures 2, 4 and 6 respectively. It is interesting to note that the classification accuracy is not affected by using a much reduced feature set. In fact, for colon cancer data, the accuracy improved to 91.9%.
One of the main concerns while carrying out a multi-objective optimization is the presence of the weight factor β. The selection of β is usually heuristic. Battiti suggested in [21] that the value of β between 0.5 and 1.0 is appropriate for most cases. The effect of changing β was studied by changing its value from 0 to 1 in steps of 0.2. using the colon cancer data set and the classification accuracy calculated (Table 2). A value of (0.5 – 1.0) for β seems appropriate. Also, the order of selection of the first 10 genes was examined (Table 3). It appears that varying β does affect the gene selection order to a certain extent. For example, comparing the gene selection orders for β = 0.6, 0.8 reveals that genes 267 and 513 swap places while genes 1256 (for β = 0.6) and 1727 (for β = 0.8) are not common to both cases. However, it must be kept in mind that the selection order in this case is not indicative of the relative importance of the genes since a greedy algorithm is being used.
We also compared our methodology to that of Ding and Peng [20] for three different datasets. The first dataset is the colon cancer dataset [24]. The second dataset is the leukemia dataset [1]. The third and final dataset used was the NCI dataset [25]. The results are tabulated in Table 4. As can be observed, the Uncertainity-based (UB) method (our method) seemed to do better than the DP (Ding and Peng) method for the colon dataset. On the other hand, for the leukemia dataset, DP proved superior to our method. For the NCI dataset, both methods performed poorly with the DP method having a slight edge. It must however be noted that the NCI dataset consists of 9 classes and only 60 samples. As a result, classifying the dataset with a very small sample size into 9 different classes and using only 15 genes is very difficult.
A further difficulty in comparing different methodologies lies in the fact that the initial pre-processing step could also play a role in classification accuracies. In the absence of a uniformity of preprocessing of the datasets, it is difficult to draw general conclusions regarding the relative performances of two different methodologies.
As commonly observed when analytical algorithms are compared, the performance shows mixed results. While Ding and Peng algorithm outperformed the one presented here (see table 4), it should be noted that the description of methods in their article did not allow us to compare both algorithms on equal terms as no gene ranking was provided, and thus the biological significance of their findings could not be assessed.
A comparison between the accuracies obtained by the original papers (from which the datasets were obtained) and our method is given in Table 5. The list of genes selected for each dataset and their ranks in the original papers are given in the supplementary file.
Discussion
The details of the selected genes and the comparison with the original data are listed in the supplementary material for all three data sets. This section presents a discussion of the comparison of genes selected by the algorithm presented in this work with those presented in earlier work (or as in the case of Breast cancer and SRBCT data, in the original work).
SRBCT data set
There were a total of 41 genes that overlapped between the selection methods presented in this work and those by [10]. The common genes were from all rank levels of the original method. Left out genes coded often, but not always for proteins from a functional system similar to those still selected here, as in the case of no. 233721 insulin-like growth factor binding protein (not selected here) and no. 296448 (insulin- like growth factor 2) and 207274 (insulin-like growth factor 2, exon 7 and additional ORF), which were selected by both methods. Interestingly, two viral oncogene sequences were not selected (nos. 417226 and 812965, v-myc avianmyelocytomatosis viral oncogene homologs), nor were some extra- cellular matrix associated genes (nos: 122159 and 809901, collagens type III and XV) both without replacement from similar genes. The seventeen newly selected genes that were not part of the original selection come from various functional systems. Of interest here is that while the original gene no. 245330 (Human krueppel-related zinc finger protein H-plk) was left out, gene no. 767495 (GLI-Krueppel family member GLI3) was newly selected. Such "nuclear localization signals" have been shown to be involved in processes determining proper nuclear localization [26], but may also be determinants of progression towards cancer [27].
Breast cancer data set
Out of the 31 genes selected here, 16 were not selected in the original publication [23], which selected 60 genes. The 45 genes not selected by the present method covered a large variety of physiological functions, without a specific pattern becoming obvious. Two genes linked to the ILGF were left out (no: s37730 and m62403), with no replacement. ILGF is linked to the development of a number of cancers (review in [28]). The fact that ILGF-linked genes are left out here may be discussed in two diametrically opposite ways. For once, leaving these genes out of the classification set may cause an oversight of the tissue's potential to induce further cancerous growth. More likely, though, it seems like whatever physiological role these genes play in the tissue, they do not contribute to distinguishing between various types of cancer.
Colon cancer data set
Contrary to the other two test sets, in the case of colon cancer, the original publication did not rank the gene set retrieved, so that a direct comparison of results was not possible. The same dataset, however, has been re- analyzed previously by Silvio Bicciato [29], using an auto associative neural network model, which yielded a ranked gene list. With the exception of Tetraspan-1, which heads the rank list with a weight of 0.9391, the top genes found by Bicciato for the reconstruction of the normal class concur with the rank list presented here, while only one gene (Heat shock 60 kD protein 1) is selected by both methods when compared to the gene list in [29] for the reconstruction of the tumor class. This tetraspan family of proteins is involved in cell adhesion processes at the gap junctions and one related protein was enhanced in highly metastatic gastric cancer [30].
Conclusion
Compared to the classification methods described in the original articles or previous third party analysis, the algorithm described here compares favorably in its capacity to select small sets of genes that distinguish between various cancer types. The observation that it leaves out several genes known to be involved in cancer development may indicate that this method's advantage lies more in good classification, but not in the detection of new dysfunctional regulatory mechanisms.
Although preliminary results using a greedy selection algorithm are encouraging, additional work needs to be done in order to develop alternative methodologies for multi-objective optimization that can select a more optimal and representative set of genes for discriminating between various cancer sub-types.
Methods
Algorithms for microarray data analysis typically focus on obtaining a set of genes that can distinguish between the different classes in a given sample set. Thus, the primary concern is to ensure the relevance of the genes to the classes under consideration.
Given a microarray data set with m samples belonging to k known classes and n genes, we want to select out those genes which are able to predict the differences in the gene expression patterns in different sample classes. Define ; |c| = k, as the vector labeling the classes of samples and ; i ∈ n as the gene expression profile of gene i. Let be the feature set of all genes and let S be the set of selected genes. Then, the feature set selection problem can be defined as follows:
Problem 1
Select a set S of genes, S ⊂ such that ∀ gene s ∈ S the relevance of s with is maximized.
However, the feature set of genes selected will contain a number of redundant genes with sometimes little relevance to the classes. This is due to the fact that the presence of genes that are closely related to each other imply that there is a possibility of genes orthoganal to those in the selected set being left out of the final feature set. Moreover, the presence of genes with little relevance to the classes leads to a reduction in the "useful information".
Ideally, selected genes should have high relevance with the classes while the redundancy among the selected genes is low. Most previous studies emphasized the selection of highly relevant genes. Ding et. al. [20] addressed the issue of the redundancies among the selected genes. The genes with high relevance are expected to be able to predict the classes of the samples. However, the prediction power is reduced if many redundant genes are selected. In contrast, a feature set that contains genes not only with high relevance with respect to the classes but with low mutual redundancy is more effective in its prediction capability.
Problem formulation
To assess the effectiveness of the genes, both the relevance and the redundancy need to be measured quantitatively. An entropy based correlation measure is chosen here. According to Shannon's information theory [31], the entropy of a random variable X can be defined as:
Entropy measures the uncertainty of a random variable. For the measurement of the interdependency of two random variables X and Y, some researchers [20,21] used mutual information, which is defined as:
I(X, Y) = H(X) + H(Y) - H(X, Y) (2)
In order to ensure that different values are comparable and have similar effects, normalized mutual information is used as a measure and is defined as:
U(X, Y) is symmetrical and ranges from 0 to 1, with the value 1 indicating that the knowledge of one variable completely predicts the other (high mutual relevance) while the value 0 indicates that X and Y are independent (low mutual relevance).
The mutual relevance between and can then be modeled by U () while the dependency between two genes is U ().
The total relevance of all selected genes is given by
The total redundancy among the selected genes is given by
Therefore, the problem of selecting genes can be reformulated as follows:
Problem 2
Select a set S of genes, S ⊂ such that ∀ gi ∈ S, the total relevance of all the selected genes with , J1, is maximized while the total relevance among all the selected genes gi ∈ S, J2, is minimized.
This is a two-objective optimization problem. To solve it, a simple way is to combine these two objectives into one:
where β is a weight parameter.
subsection* Algorithm
To solve the above problem, Battiti [21] proposed a greedy algorithm. The procedure can be described as follows (see Figure 7):
1. Initialization: F ← allgenes, S ← ∅.
2. First gene: select gene i that has highest relevance U (). gi ∈ S, F \ i.
3. Remaining genes: From F, select gene j that maximizes .
4. Repeat the above step until the desired number of genes are obtained.
The maximization problem (6) can also be re-formulated into a binary optimization problem. Let xi be a binary variable with value 1 for selecting gene i while value 0 for not. Thus, Equation (6) can be rewritten into:
It can be further rewritten into matrix form:
max UcTx - βxTUpx (8)
where Uc is the relevance vector, Up is matrix of pairwise redundancy.
Beasley et al. [32] discussed several heuristic algorithms to solve such binary quadratic programming problems. A heuristic simulated annealing method was employed to solve the problem. The pseudo codes of simulated annealing can be obtained from [32].
There are however limitations to both approaches. There is a possibility that the solution obtained for Problem 2 can lead to a local optimum. This could result in a sub-optimal feature set thereby affecting the prediction accuracy. In order to expand the search space, an iterative procedure was adopted. The data was initially clustered and partitioned into K groups, C1, C2,..., CK by using k-means clustering. The idea was to group genes with similar expression patterns together. The greedy or heuristic simulated annealing procedure was then applied to select a subset of genes, Sk, from each partition, k, such that the selected genes had low mutual relevance with respect to each other while at the same having maximal relevance with the different classes. The genes selected from each subset are then combined to obtain a single gene set, that is, S = S1 ⋃ S2 ⋃ S3,..., ⋃SK.
The final set of genes is selected by carrying out a leave-one-out cross validation (LOOCV). For each run, one sample is held out for testing whilei the remaining N - 1 samples are used to train the classifier. The genes are selected by the algorithm using the training samples and then are used to classify the testing sample. The overall accuracy rate is calculated based on the correctness of the classifications of each testing sample. In order to get a deeper understanding of the selected genes, those genes found in common for all the N different runs of the LOOCV experiment are finally listed out for further investigation. The process of gene selection is repeated by selecting a subset of genes from this feature set, that gives a classification error that is below a user defined threshold ε. Nearest neighborhood (k-NN) classification method is used to assess the discriminant power of the selected genes by the method. The process is stopped when the error becomes greater than ε. The full algorithm is presented in Figure 8.
Authors' contributions
LXX was responsible for the development and implementation of the algorithm as well as for writing parts of the paper. AK was involved in algorithm development as well as in writing the manuscript. AM was responsible for the analysis of the results as well as manuscript preparation. All authors read and approved the manuscript.
Supplementary Material
Additional File 1
Selected Genes for All Datasets. The file contains the list of selected genes for each of the three datasets used in this study as well as the corresponding ranks of those selected genes in the original papers.
Click here for file
Acknowledgements
The authors would like to thank the anonymous reviewers for their suggestions and critical reviews of the paper.
Figures and Tables
Figure 1 Clustering dendrogram of SRBCT data – First Iteration.
Figure 2 Clustering dendrogram of SRBCT data – Reduced Feature Set.
Figure 3 Clustering dendrogram of breast cancer data – First Iteration.
Figure 4 Clustering dendrogram of breast cancer data – Reduced Feature Set.
Figure 5 Clustering dendrogram of colon cancer data – First Iteration.
Figure 6 Clustering dendrogram of colon cancer data – Reduced Feature Set.
Figure 7 Greedy Algorithm.
Figure 8 Optimal Feature Set Selection Algorithm. The function CLUSTER uses the k-means clustering approach to partition the initial gene set into the desired number of partitions K, with G genes in each partition. K and G are user-specified. The function SELECT_GENES uses either the greedy approach (Figure 7) or the heuristic simulated annealing approach to solve Problem 2. The function CLASSIFICATION_ERROR uses kNN classification method to assess the discriminant power of the selected genes and returns the classification error.
Table 1 Classification accuracies and the number of selected genes for the two different optimizationmethods (Greedy and Simulated Annealing (SA)). For Greedy selection, the accuracies as well as the number of genes selected in iterations 1 and 2 are shown. The reported accuracies are LOOCV accuracies while the number of genes is the smaller subset common to all LOOCV experiments.
Algorithm Colon Breast SRBCT
% Acc # Genes % Acc # Genes % Acc # Genes
Greedy 90.3/91.9 29/9 89.8/89.8 31/12 100/100 58/14
SA 87.1/- 26/- 89.8/- 44/- 100/- 58/-
Table 2 Effect of varying β on classification accuracy. The effect of varying β was studied for the colon cancer data set. A value of between 0.5 – 1 as suggested by Battiti [21] seems appropriate.
β 0.0 0.2 0.4 0.6 0.8 1.0
accurate 87.1% 88.7% 90.3% 90.3% 90.3% 90.3%
Table 3 Effect of varying β on the selection order of genes. The first ten genes selected for each value of β are shown here. The numbers correspond to the gene numbers for the colon cancer data set. Varying β does seem to affect the order in which the genes are selected. However, selection order is not indicative of the relative importance of genes since a greedy-selection method is being used.
β g1 g2 g3 g4 g5 g6 g7 g8 g9 g10
0.0 377 267 765 493 1582 513 1635 1671 245 780
0.2 377 267 1582 513 765 493 1635 1671 780 1423
0.4 377 1582 267 513 493 765 1635 1671 780 1491
0.6 377 1582 267 513 1491 493 1635 765 1671 1256
0.8 377 1582 513 267 1491 1727 493 1635 1671 765
1.0 377 1582 1491 513 267 1727 1244 1256 1671 1873
Table 4 Classification accuracies and the number of selected genes for the two different mutual information based methodologies (Uncertainity based (UB) and Ding and Peng's (DP)). The accuracies as well as the number of genes selected in iterations 1 and 2 respectively are shown for the UB method while the accuracies and genes selected for two different runs are shown for the DP case. For both methodologies, the accuracies reported are LOOCV accuracies.
Algorithm Colon Leukemia NCI
% Acc # Genes % Acc # Genes % Acc # Genes
UB (ours) 90.3/91.9 29/9 80.6/76.4 21/5 57.6/52.5 59/15
DP 75.8/91.9 50/20 98.6/100 50/10 73.3/61.7 50/20
Table 5 Classification Results of Original Papers
Dataset Colon Breast SRBCT
Accuracy (original) only clustering 89.47 100
Accuracy (UB) 91.9 89.8 100
==== Refs
Golub TR Slonim DK Tamayo P Huard C Gaasenbeek M Mesirov JP Coller H Loh ML Downing JR Caligiuri MA Molecular classification of Cancer:class discovery and class prediction by gene expression monitoring Nature 1999 286 531 537 10.1126/science.286.5439.531
Alizadeh AA Eisen MB Davis RE Ma C Lossos IS Rosenwald A Boldrick JC Sabet H Tran T Yu X Distinct types of diffuse large B-cell lymphoma identified by gene expression profiling Nature 2000 403 503 511 10676951 10.1038/35000501
Bittner M Meltzer P Chen Y Jiang Y Seftor E Hendrix M Radmacher M Simon R Yakhini Z Ben-Dor A Molecular classification of cutaneous malignant melanoma by gene expression profiling Nature 2000 406 536 540 10952317 10.1038/35020115
Perou CM Sorlie T Eisen MB van de Rijn M Jeffrey SS Rees CA Pollack JR Ross DT Johnsen H Akslen LA Molecular portraits of human breast tumours Nature 2000 406 747 752 10963602 10.1038/35021093
Weinstein JN Myers TG O'Connor PM Friend SH Fornace AJ Kohn KW Fojo T Bates SE Rubinstein LV Anderson NLea An informatio-intensive approach to the molecular pharmacology of cancer Science 1997 275 343 349 8994024 10.1126/science.275.5298.343
Tavazoi S Hughes JD Campbell MJ Cho RJ Church GM Systematic determination of genetic network architecture Nature Genetics 1999 22 281 285 10391217 10.1038/10343
Tamayo P Slonim D Mesirov J Zhu Q Kitareewan S Dmitrovsky S Lander ES Golub TR Interpreting patterns of gene expression with self-organizing maps: methods and application to hematopoietic differentiation National Academy of Sciences 1999 96 2907 2912 10.1073/pnas.96.6.2907
Dudoit S Fridlyand J Speed T Comparison of discriminantion methods for the classification of tumors using gene expression data Tech rep 2000 University of Carlifornia, Berkeley
Xiong M Li W Zhao J Jin L Boerwinkle E Feature (Gene) Selection in Gene Expression-Based Tumor Classification Molecular Genetics and Metabolism 2001 73 239 247 11461191 10.1006/mgme.2001.3193
Khan J Classification and diagnostic prediction of cancers using gene expression profiling andartificial neural networks Nature Medicine 2001 7 673 679 11385503 10.1038/89044
Furey TS Cristiani N Duffy N Bednarski DW Schummer M Haussler D Support vector machine classification of cancer tissue samples using microarray expression data Bioinformatics 2000 16 906 914 11120680 10.1093/bioinformatics/16.10.906
Guyon I Weston J Barnhill S Vapnik V Gene selection for cancer classification using support vector machines Machine Learn 2002 46 389 422 10.1023/A:1012487302797
Mukherjee S Tamayo P Slonim D Verri A Golub T Messirov JP Support vector machine classification of microarray data AI memo CBCL paper 182 2000 MIT Press, Cambridge, MA
Ohno-Machado L Vinterbo S Weber G Classification of gene expression data using fuzzy logic Journal of Intelligent and Fuzzy Systems 2002 12 19 24
Cai J Dayanik A Yu H Hasan N Terauchi T Grundy W Classification of gene cancer types by support vector machines using microarray gene expression data International Conference on Intelligent Systems for Molecular Biology 2000
Xu M Setiono R Gene selection for cancer classification using a hybrid of univariate and multivariate feature selection methods Applied Genomics and Proteomics 2003 2 79 91
Kohavi R John G Wrapper for feature subset selection Artificial Intelligence 1997 97 273 324 10.1016/S0004-3702(97)00043-X
Liu H Motoda H Feature selection for knowledge discovery and data mining 1998 Boston, Kluwer Acad. Publishers
Slonim D Tamayo P Mesirov J Golub T Lander E Class prediction and discovery using gene expression data 4th Annual International Conference on Computational Molecular Biology (RECOMB), 2000 Apr 8–11; Tokyo, Japan 2000 Tokyo: Universal Academy Press 263 272
Ding C Peng H Minimum Redundancy feature selection from microarray gene expression data Computational Systems Bioinformatics 2003
Battiti R Using Mutual Information for Selecting Features in Supervised Neural Networks IEEE transactions on neural networks 1994 5
Kirkpatrick S Gelatt C Vecchi M Optimization by simulated annealing Science 1983 220 671 680
West M Predicting the clinical status of human breast cancer by using gene expression profiles Proc Natl Acad Sci USA 2001 98 11462 11467 11562467 10.1073/pnas.201162998
Alon U Broad patterns of gene expression revealed by clustering analysis of tumor and normal colon tissuesprobed by oligonucleotide arrays Proc Natl Acad Sci USA 1999 96 6745 6750 10359783 10.1073/pnas.96.12.6745
Ross DT Scherf Ut Systematic variation in gene expression patterns in human cancer cell lines Nature Genetics 2000 24 227 234 10700174 10.1038/73432
Quadrini K Bieker J Kruppel-like zinc fingers bind to nuclear import proteins and are required for efficient nuclear localization of erythroid Kruppel-like factor J Biol Chem 2002 277 32243 32252 12072445 10.1074/jbc.M205677200
Zhang F White R Neufeld K Phosphorylation near nuclear localization signal regulates nuclear import of adenomatous polyposis coli protei Proc Natl Acad Sci 2000 97 12577 12582 11050185 10.1073/pnas.230435597
Renehan A Zwahlen M Minder C O'Dwyer S Shalet S Egger M Insulin-like growth factor(IGF)-I, IGF binding protein-3, and cancer risk Systematic Review and Meta-Regression Analysis, Lancet 2004 363 1346 1353
Bicciato S Pandin M Didone G Bello CD Pattern Identification and Classification in Gene Expression Data Using an Autoassociative Neural Network Model Biotechnology and bioengineering 2003 81 594 606 12514809 10.1002/bit.10505
Lee S Baek M Yang H Bang Y Kim W Ha J DK K DI J Identification of genes differentially expressed between gastric cancers and normal gastric mucosa with cDNA microarrays Cancer Letters 2002 184 197 206 12127692 10.1016/S0304-3835(02)00197-0
Shannon CE Weaver W The Mathematical Theory of Communication 1949 University of Illinois Press
Beasley JE Heuristic algorithms for the unconstrained binary quadratic programming problem London, England 1998
| 15790388 | PMC1087831 | CC BY | 2021-01-04 16:35:47 | no | BMC Bioinformatics. 2005 Mar 24; 6:76 | utf-8 | BMC Bioinformatics | 2,005 | 10.1186/1471-2105-6-76 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-771579042710.1186/1471-2105-6-77Research ArticleComparative mapping of sequence-based and structure-based protein domains Zhang Ya [email protected] John-Marc [email protected] Chris [email protected] Stephen R [email protected] Physical Biosciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA2 Computational Research Division, Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA3 School of Information Sciences and Technology, The Pennsylvania State University, University Park, PA 16802, USA2005 25 3 2005 6 77 77 3 11 2004 25 3 2005 Copyright © 2005 Zhang 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 domains have long been an ill-defined concept in biology. They are generally described as autonomous folding units with evolutionary and functional independence. Both structure-based and sequence-based domain definitions have been widely used. But whether these types of models alone can capture all essential features of domains is still an open question.
Methods
Here we provide insight on domain definitions through comparative mapping of two domain classification databases, one sequence-based (Pfam) and the other structure-based (SCOP). A mapping score is defined to indicate the significance of the mapping, and the properties of the mapping matrices are studied.
Results
The mapping results show a general agreement between the two databases, as well as many interesting areas of disagreement. In the cases of disagreement, the functional and evolutionary characteristics of the domains are examined to determine which domain definition is biologically more informative.
==== Body
Background
The concept of protein domains has gained increasing interest from the biology research community because of its importance in protein classification [1], protein function assignment [2], and protein engineering [3]. Protein domains are generally considered as protein fragments of common structures which may independently fold [4] or have their own functions [5]. They have also been treated as evolutionary units [6]. Protein domains function as the building blocks of proteins and are often recombined to form different proteins [5], leading to high redundancy in protein structures. Currently, a few thousand protein domains have been identified, a total much smaller than the number of proteins. Classifying proteins based on their constituent domains is therefore one of the most effective and efficient approaches to organize protein data both by structures and by evolutionary relationships. However, such a classification requires the identification of domain composition for proteins, which is by no means an easy task. The challenge lies in the ambiguity of domain definitions, as well as the lack of useful structural information about most proteins.
Two types of approaches have been widely used to assign domains: one based on the three-dimensional (3D) structures of proteins and the other based on protein sequences. Structure-based approaches define domains primarily according to the compactness and conservation of protein structural regions, generally described as globular modules. The domain annotation is best achieved through an expert's visual inspection of protein three-dimensional structures. Currently, the Protein Data Bank (PDB) [7], the primary protein structural database, contains 26,610 protein structures. A number of structure-based domain classification databases such as SCOP (Structural Classification of Proteins) [1], FSSP (Families of Structurally Similar Proteins) [8], and CATH (Class Architecture Topology Homology) [9] are constructed using the available protein structures so that proteins can be easily analyzed for the presence of domains. Among them, the SCOP database is manually curated and considered the most reliable domain classification. However, this classification covers only about 2–3% of sequenced proteins. At this time, the Swiss-Prot+TrEMBL [10] sequence databases together contain over 1.5 millon entries. The gap between the number of sequenced proteins and that of proteins with experimentally determined 3D structures is still increasing, which has greatly constrained the development of structure-based protein classification databases. Although 58% of sequences can be modeled using comparative modeling [11], the accuracy of such comparative models decreases sharply below the 30% sequence identity cutoff. An alternative classification schema assigns domains to proteins by only sequence information. Sequence-based domain databases constructed with this classification schema include Pfam [12], ProDom [13] and InterPro [14]. These databases define domains based on sequence similarity and implied evolutionary relationships. In this manuscript we focus on the Pfam database in which domain boundaries are manually assigned by experts.
Since domains are structurally and evolutionarily independent units, we may ask whether either a structure-based or sequence-based classification alone is sufficient and how well they agree. A previous study compared three structure-based classifications: SCOP, CATH and FSSP [15], and concluded that the majority of their classifications agreed. Two sequence-based domain databases were also compared [16] and discrepancies between the two databases were attributed to their different philosophies. In this paper, we strive to improve domain definitions through examining the correspondence between sequence-based domains and structure-based domains, using the domain definitions in SCOP as the representative for structure domains and those of Pfam as the representative for sequence domains. Elofsson and Sonnhammer [17] compared the Pfam and SCOP databases in 1999. According to their comparison, 70% of the SCOP domain families and 57% of the Pfam families have counterparts in the other databases. However, since then, both databases have greatly increased in size and various revisions and updates have been made. For example, the domain representation in Pfam was revised to model discontinuous domains [12]. Therefore, it is now timely and important to revisit this topic and compare the two types of domains under the new setting. Furthermore, the aim of this comparison is to some extent different from what Elofsson and Sonnhammer had. Other than examining the extent that the two databases overlap, we focus more on their differences. When inconsistencies in domain definitions occurs, we propose to determine which domain definition is biologically more meaningful by inspecting the evolution of those domains.
We directly map SCOP domains to Pfam domains based on their corresponding locations in their member sequences. The approach assigns a mapping score to the pair of domains under comparison to quantitatively represent the quality of the match.
The mapping reveals a moderate agreement among Pfam families and SCOP domain families. Five types of relationships between the two classifications are clearly indicated in the mapping results and we therefore put them into five categories. Statistical analysis and individual instances are provided for each category of mapping. In the case of disagreement in domain classification, information from past literature, such as known domain functions, is used as external validation. We also propose to examine the evolutionary history of each individual domain when disagreement occurs.
An overview of SCOP and Pfam
The SCOP [1] database is manually curated by experts. It orders all proteins with known structures, according to their evolutionary and structural relationships. The database adopts a hierarchical organization: domains are grouped into families, then superfamilies, folds and classes in the highest level of the hierarchy.
Pfam [12] contains hidden Markov model based profiles (HMM-profiles) of many common protein domains based on multiple sequence alignments. While the construction of the HMM-profiles is semi-automatic, expert knowledge contributes in the grouping of proteins, the aligning of protein sequences, and the quality control of the HMM-profiles. Although Pfam is subclassified by 'type' in 2002 as 'family', 'domain', 'repeat' and 'motif', its organization is generally considered to be flat. We hence do not differentiate the subtypes in this comparison.
The Pfam database contains two parts: one is the curated section called Pfam-A and the other is an automatically generated supplement called Pfam-B which represents small families taken from the PRODOM database that do not overlap with Pfam-A. In this study, only Pfam-A families are mapped to SCOP domain families.
Methods
Materials
All PDB protein sequences, based on PDB SEQRES records, with less than 95% identity to each other were downloaded from the ASTRAL Compendium [18,19]. This data set contains 8259 protein chains. Pfam 14.0 was downloaded from . Only Pfam-A families were used for the comparison. This version contains 7459 Pfam-A families and corresponding HMM-profiles. The HMMER package, version 2.3.2, was used to compare PDB protein sequences to Pfam-A HMM-profiles. The Pfam 'trusted cutoff' was used to determine whether a Pfam domain matches a PDB chain. The SCOP domain definitions were from the SCOP parsable files version 1.65. Because the SCOP parsable files are based on the PDB ATOM records, the ATOM records were mapped to PDB SEQRES records using the RAF mapping provided by ASTRAL before the comparison.
We propose to map the Pfam-A families to SCOP domain families based on their locations in member sequences. Each Pfam-A family or SCOP domain family is treated as a set of member protein sequences. A mapping between a Pfam family and a SCOP domain family is defined as follows: (1) they have at least one member protein sequence in common; (2) their locations in the common protein sequences overlap; and (3) their mapping score is larger than the pre-set threshold m. For each PDB protein sequence, a comparison was then made for the overlaps and differences in the SCOP domain families and the Pfam families. The process of mapping is illustrated with Figure 1.
Mapping matrix
Ideally, if a SCOP domain family and a Pfam family are defined at the same location over the same set of protein chains, then they map exactly to each other. However, in most cases, the mapping is not exact, i.e. they only partially overlap at individual member protein sequences or their member sequences are not all the same. In order to measure the extent of overlap, a mapping score is assigned to each pair of SCOP domain families and Pfam families. Intuitively, if the SCOP domain family and the Pfam family have more members in common and their corresponding protein sequence segments overlap more, then they are more likely to be mapped to each other. However, this mapping criteria favors those domains whose frequencies are high. Since we use only PDB protein chains in the comparative mapping, this data set may be biased towards those proteins of interests to biologists or whose structures are easier to resolve. For both domain models, we observe a power law distribution of domain frequency, where a few domains occurs in a large number of protein sequences and many domains occur in very few protein sequences. To account for the frequencies of domains, the mapping score is normalized by the average frequency of the two domains under comparison. Let si denotes the i-th protein domain in SCOP and fj the j-th protein domain in Pfam. The mapping score M(si, fj) is defined as
where P represents the set of PDB protein chains with both domain si and domain fj; pk is the kth protein chain in the set; overlap(, ) is the length of the overlapped segment on pk; and length() is the length of si on pk. freq(si) and freq(fj) represent the frequencies of the ith SCOP domain and jth Pfam family, respectively. The factor is to counteract the influence of frequency differences between protein domains. Here min(length(), length()) is used as the denominator because we want to distinguish the cases where two domains overlap in a small part of their coverage and where one domain is completely covered by the other domain, as shown in Figure 2.
Properties of the mapping matrix
The mapping scores for all SCOP and Pfam domain pairs form a matrix M. The matrix representation of the mapping has some nice properties. First consider mapping the SCOP domain si to all possible Pfam domains. We look at the i-th row of M. The number of nonzeros, , in the row indicates how many Pfam domains that the SCOP domain si could possibly map to. Among the possible mapping, the most likely Pfam domain that the SCOP domain si will map to is
Note that the number of nonzeros, , could be large, which implies that si maps to many Pfam domains. However, sometimes, two domains overlap very insignificantly, say only a few amino acid residues. To eliminate the insignificant mapping, we set a threshold, m, and require mapping to satisfy Mij ≥ m.
Next consider mapping the Pfam domain fj to all possible SCOP domains. We look at the j-th column of M. The number of nonzeros, , in the column indicates how many SCOP domains could be mapped to. The most likely SCOP domain that fj will map to is
The threshold m is again used to reduce insignificant mapping.
Results
Domain mapping
A total of 2081 Pfam families and 2512 SCOP domain families are defined in the set of 8259 PDB protein chains. The average lengthes of Pfam families and SCOP domains are 96 and 174 residues, respectively. The threshold m for mapping scores is empirically set to be 0.01 to include as much mapping as possible here, because even a small portion of the overlapping may be informative.
From the mapping results, 2008 (80%) SCOP domain families overlap with at least one Pfam family, and these SCOP domain families correspond to 2075 (99.7%) of the Pfam families. On average, each SCOP domain maps to 1.3 Pfam families, and each Pfam domain maps to 1.0 SCOP families. This result is expected because Pfam domains are overall 16% shorter than SCOP domains. The lengths of protein domains in SCOP are plotted against those of the corresponding Pfam families in Figure 3. One-fifth (504) of SCOP domain families have no Pfam counterpart, while only six (0.03%) Pfam families are not mapped to SCOP domain families (Table 1). Further analysis reveals that all the sequence segments corresponding to the unmapped Pfam families represent regions of residues that were absent in the PDB structures. That is, all Pfam families with known PDB structures are mapped to at least one SCOP domain family. It is unclear why 20% of SCOP domain families do not correspond to any Pfam family. One possible explanation is that the there are too few examples of those SCOP domain families to build HMM-profiles for Pfam families.
Exploring the mapping results
Several types of sequence-structure domain relationships emerge during this study, including:
• One SCOP domain family maps to exactly one Pfam family, where the SCOP domain family and the Pfam family overlap with and only with each other. However, their member sequences and their coverages at each individual sequence may slightly differ.
• One SCOP domain family maps to many Pfam families, where for each member sequence, the coverage of the SCOP domain family corresponds to the summation of those corresponding Pfam families.
• Many SCOP domain families map to one Pfam family, where for each member sequence, the coverage of the Pfam family corresponds to the summation of those corresponding SCOP domain families.
• One SCOP domain family maps to sets of Pfam families, where the SCOP domain family corresponds to one Pfam family at each member sequence, but to different Pfam families at different member sequences.
• Sets of SCOP domain families map to one Pfam family, where the Pfam family corresponds to one SCOP domain family at each member sequence, but to different SCOP domain families at different member sequences.
Examples of each type are provided in Table 2. We present below a detailed analysis of our findings.
One-to-one exact mapping
996 SCOP domains each maps to exactly one Pfam family. That is, 39.65% of SCOP domain families and 47.86% of Pfam families have exactly one counterpart in the other type of domain classification. Among these Pfam families, 431 (43.3%) are labelled as 'Family' type, 558 (56.0%) are associated with 'Domain' type, 4 (0.4%) with 'Repeat' type and 3 (0.3%) with 'Motif' type. Thus, the SCOP domain families largely (99.3%) correspond to 'Family' or 'Domain' types in Pfam.
In the case of one-to-one mapping, these Pfam domains have an average length of 164.0, and the SCOP domains have an average length of 182.7, 11% longer on average than the corresponding Pfam domains. Even where two domains are mapped one-to-one, their definitions may slightly disagree. For instance, their member protein sequences may not be exactly the same, or their corresponding sequence segments may not completely overlap. A few examples of Pfam domains and SCOP domains are graphed onto the corresponding member protein structures using Pymol [20] as shown in Figure 4 to illustrate the latter case.
Figure 5 shows the histogram of the differences in domains' endpoints. For two domains fi and sj, their difference in the endpoints is calculated as the total length of the regions covered by fi or sj minus the length of the shared regions covered by fi and sj. More than 50% (511) of the mappings between Pfam families and SCOP domain families differ by less than 10 residues, while only 3.4% (34) of domain mappings differ by more than 100 residues. To quantify the extent of the one-to-one mapping, we define a mapping ratio as
where P is the common member protein sequences of the two types of domain families, intersect(, ) is the length of the overlapped portion of the ith Pfam family with the jth SCOP domain family at the kth member protein sequence, and union(, ) is the length of the regions covered by either of them. Figure 6 shows the distribution of the mapping ratios. Among these cases of one-to-one mapping, 61.24% have a mapping ratio larger than 0.9. That is, the two types of domain definitions vary in less than 10% of the domain sequences. 81.62% vary in less than 20% of the domain sequences, and 90.26% vary in less than 30% of the domain sequences.
One SCOP domain family to many Pfam families
A total of 76 SCOP domain families map to multiple Pfam families. About half (33) of these SCOP domain families correspond to several copies (repeats) of the same Pfam family. The corresponding Pfam families may be of Pfam type 'Family', 'Domain', or 'Repeat'. One example is provided for each case in Figure 7. SCOP domain a.118.1.8 (Pumilio repeat)corresponds to 8 copies of Pfam family PUF (Pumilio-family RNA binding repeat) of type 'Family' (Figure 7(A)), SCOP domain c.10.2.8 (Polygalacturonase inhibiting protein PGIP) corresponds to 8 copies of Pfam family LRR (Leucine Rich Repeat) of type 'Repeat' (Figure 7(B)), and SCOP domain a.39.1.10 (Polcalcin phl p 7) corresponds to 2 copies of Pfam family efhand (EF hand) of type 'Domain' (Figure 7(C)). It seems that these Pfam families all serve as building blocks for SCOP domains and more careful investigation is required to determine the validity of these domains.
Several Pfam families, such as LRR (Leucine Rich Repeat) and efhand (EF hand) have a high frequency of mapping to SCOP domain families. For instance, the SCOP domain c.10.1.2(Rna1p (RanGAP1), N-terminal domain) maps to two copies of the Pfam family LRR, the SCOP domain c.11.1.1 (Outer arm dynein light chain 1) maps to four copies of LRR, and the SCOP domain c.10.2.8 (Polygalacturonase inhibiting protein PGIP) maps to eight copies of LRR (Figure 7(B)). Most of the SCOP counterparts of LRR belong to the SCOP L domain-like superfamily. Pfam annotates LRR as Repeat type, and describes them as 'short sequence motifs present in a number of proteins with diverse functions'. These types of Pfam families actually represent structural components that form structural domains. They differ from domains in that they are functionally and evolutionarily dependent on other structure components. Therefore, we would suggest these Pfam families being removed from the Pfam-A family.
Many SCOP domain families to one Pfam family
There are 106 Pfam families mapped to multiple SCOP domains. Of them, 25 map to repeats of the same SCOP domain. Several examples for this type of mapping are shown in Figure 8. According to the mapping results for the bacterial multidrug efflux transporter AcrB (PDB ID 1iwgA), the Pfam ACR_tran (AcrB/AcrD/AcrF) family corresponds to eight SCOP domain families in the order of f.35.1.1 (Multidrug efflux transporter AcrB transmembrane domain), d.58.44.1 (Multidrug efflux transporter AcrB pore domain; PN1, PN2, PCI and PC2 subdomains), d.58.44.1, d.225.1.1 (Multidrug efflux transporter AcrB TolC docking domain; DN and DC subdomains), f.35.1.1, d.58.44.1, d.58.44.1, and d.225.1.1 (Figure 8(A)). Among these SCOP domains, only three are unique, and the second four SCOP domains are exact repeats of the first four SCOP domains. These SCOP domains are found to co-exist in PDB protein chains 1iwG, 1oy8, 1oyE, 1oy6, 1oy9, and 1oyD based on SCOP records. Further inspection reveals that these domains are always present together in the multidrug efflux transporter proteins in the same order, and they act collaboratively in the process of exporting toxic compounds out of the cell [21].
However, each functions independently: d.225.1.1 docks TolC into AcrB, f.35.1.1 translocates substrates from the cell interior, and d.58.44.1 translocates substrates into the TolC tunnel. In this sense, the SCOP domain classification is more accurate and the Pfam ACR_tran family may be chopped into eight small domains. Similarly, the Pfam family Glyco_hydro_42 (Beta-galactosidase), mapped to a series of the SCOP domain families c.1.8.1 (Amylase, catalytic domain), c.23.16.5 (A4 beta-galactosidase middle domain), and b.71.1.1 (alpha-Amylases, C-terminal beta-sheet domain), may be partitioned into three small domains.
One SCOP domain to sets of Pfam families
289 SCOP domains are mapped to sets of Pfam domains, one set at a time. For example, the SCOP domain d.81.1.2 (Homoserine dehydrogenase-like) maps to the Pfam family Homoserine_dh (Homoserine dehydrogenase) on the PDB protein chain 1ebf A (Figure 9(A)) and to the Pfam family Saccharop_dh (Saccharopine dehydrogenase) on the PDB protein chain 1e5qA (Figure 9(B)). Another example is the SCOP domain family e.8.1.1 (DNA polymerase I) which maps to the Pfam DNA_pol_A (DNA polymerase family A) and DNA_pol_B (DNA polymerase family B) on different PDB protein chains. Relationships are suggested between these Pfam families that are individually mapped to a same SCOP domain family. If several sets of Pfam families are mapped to the same SCOP domain, based on the fact that the SCOP domain families are functionally independent, these Pfam families are very likely to share both functions and structures. Therefore, close scrutiny may be required to determine whether these Pfam families should be merged or not.
Sets of SCOP domain families to one Pfam family
We find 314 Pfam families that map to multiple sets of SCOP domain families. Under this category a subtype of special interest is Pfam families corresponding to SCOP superfamilies. Some examples of this subtype are listed in Table 3. For instance, the SCOP domain families c.107.1.1 (Manganese-dependent inorganic pyrophosphatase (family II)) and c.107.1.2 (Exonuclease RecJ family) each individually map to the Pfam family DHH (DEE Family). Both of the SCOP domains belong to the SCOP superfamily c.107.1 (DHH phosphoesterases). Another example is the Pfam family Glyoxalase (Glyoxalase/Bleomycin resistance protein/Dioxygenase superfamily). The Pfam domain is independently mapped to the following four SCOP domain families: d.32.1.1 (Glyoxalase I (lactoylglutathione lyase)), d.32.1.2 (Antibiotic resistance proteins), d.32.1.3 (Extradiol dioxygenases), and d.32.1.4 (Methylmalonyl-CoA epimerase). These SCOP domains all belong to the SCOP superfamily d.32.1 (Glyoxalase/Bleomycin resistance protein/Dihydroxybiphenyl dioxygenase). From the Pfam annotation of the Pfam family Glyoxalase, we see that Pfam seems to be aware of it is a superfamily. But the flat organization of Pfam fails to reflect this property explicitly. In this sense, the comparative mapping between SCOP and Pfam could help Pfam to build a hierarchical organization. On the other hand, it is known that all SCOP classes higher than 7 are considered "not true SCOP classes" and their subtypes (folds, superfamilies, and families) are considered not "true", either. We can utilize this type of mapping to put those SCOP domains in meaningful classes. For example, the SCOP domain families c.96.1.1 (Fe-only hydrogenase) and i.4.1.1 (Electron transport chains) each individually map to the Pfam family Fe_hyd_lg_C (Iron only hydrogenase large subunit, C-terminal domain). It may be inferred that the SCOP domain family i.4.1.1 is related to SCOP superfamily c.96.1.
Combination of types
In many cases, a combination of several types is observed. For example, the Pfam TCP-1/cpn60 chaperonin (Cpn60.TCP1) family is mapped to two different sets of SCOP domains, each consisting of a series of three domains: {a.129.1.2 (Group II chaperonin (CCT, TRIC), ATPase domain), d.56.1.2 (Group II chaperonin (CCT, TRIC), intermediate domain), and c.8.5.2 (Group II chaperonin (CCT, TRIC), apical domain)} and {a.129.1.1 (GroEL chaperone, ATPase domain), d.56.1.1 (GroEL-like chaperone, intermediate domain), and c.8.5.1 (GroEL-like chaperone, apical domain)}. These two sets of SCOP domains usually occur together. However, the SCOP domain families c.8.5.1 and c.8.5.2 are also each present on their own in many PDB protein chains. This indicates that c.8.5.1 and c.8.5.2 are each an independent, single domain. According to Aroul-Selvam et. al [22], this three domain set is formed through two insertions as follows: a.129.1.1 and a.129.1.2 are the parent domains, followed by the insertion of d.56.1.1 into a.129.1.1 and d.56.1.2 into a.129.1.2. Finallyc.8.5.1 is inserted into d.56.1.1, and c.8.5.2 is inserted into d.56.1.2 (Figure 10). Members with the domain organization of {a.129.1.2, d.56.1.2, c.8.5.2} are the molecular chaperone GroEL and proteins with similar functions. These proteins are known to have three functional domains: equatorial (ATPase) domain, intermediate domain, and apical domain, each with its own distinct function. The whole protein functions as a molecular chaperone, which binds unfolded polypeptides in vitro, and has a weak ATPase activity. The apical domain is involved in substrate binding. The equatorial domain contains the nucleotide binding site and provides most of the intersubunit contacts. The linker domain serves to transmit allosteric effects between the other two domains.
Comparative mapping may help build Pfam clans
The Pfam database employs a flat organization, with a 'Type' annotation attached to each family. The annotation is to some extent similar to levels in SCOP hierarchical organization. Clans have been introduced in Pfam to reflect the evolutionary relationship between different families. Each clan contains two or more Pfam families that have arisen from a single evolutionary origin. However, Pfam release 14.0 contains only 15 clans covering less than 100 Pfam families. With our comparative mapping results, the SCOP hierarchy may be used to help Pfam generate the clans. For example, when one SCOP domain family is mapped to sets of Pfam families, a strong connection/relationship between those Pfam domains may be implied. A clan may be inferred from those Pfam families. Therefore, we compared our results with the existing Pfam clans. Table 4 lists the member families in existing Pfam clans and their corresponding SCOP domains. We only list 10 rather than 15 because the other five mostly contain Pfam families not used in the comparison. As can been seen from the Table, members of a clan usually correspond to a SCOP family or a SCOP superfamily. Therefore, we believe the results from comparative mapping could potentially be helpful in building Pfam clans.
Phylogenetic analysis
Domains are considered evolutionarily independent units, and the evolution history of each domain is expected to be characteristic. Similar domain evolutionary histories may indicate relations among domains. Therefore, we propose to use correlation in domain evolution to validate the domain definitions by Pfam and SCOP in the case of disagreement.
Tan et. al have designed a tool to compute the similarities between proteins' evolutionary histories [23]. This approach can be slightly modified to fit our needs for determining the similarities between domains' evolutionary histories. We define the evolutionary correlation between two domains as the average correlation between pairs of their member sequences. The correlation between two sequence segments is then defined as the Pearson correlation coefficient of the evolutionary distance matrices of the two sequences. It is computed using the following steps. First, Blastp is used to find the orthologous protein sequences in two sets of genomes; bacterial and eukaryotic. The bacterial data set contains proteins from the genomes of eighteen species: Acinetobacter sp ADP1, Fusobacterium nucleatum, Nitrosomonas europaea, Vibrio parahaemolyticus, Bacillus anthracis Ames, Geobacter sulfurreducens, Pyrococcus abyssi, Xylella fastidiosa, Campylobacter jejuni, Helicobacter hepaticus, Rickettsia conorii, Yersinia pestis KIM, Deinococcus radiodurans, Lactococcus lactis, Streptococcus pyogenes, Escherichia coli K12, Methanosarcina mazei, and Thermotoga maritima. The eucaryotic data set contains genome protein sequences from nine species, including Arabidopsis thaliana, Encephalitozoon cuniculi, Plasmodium falciparum, Caenorhabditis elegans, Homo sapiens, Rattus norvegicus, Drosophila melanogaster, Mus musculus, and Saccharomyces cerevisiae.
Second, for each species, the orthologous protein sequence with the highest E-value is selected (if a significant one exists). Third, ClustalW is then used to align these sequences. Fourth, the Pearson correlation coefficient of those mapping matrices is computed with Equation 3, which represents the correlation between the corresponding sequence pair.
where N is the number of species where orthologous sequences were retrieved, S and P are N × N distance matrices from ClustalW alignment of sequence segments in SCOP domain families and Pfam families, respectively. The correlation between two domains is then expressed as:
where abs (x) gives the absolute value of x, and Ni and Nj are the number of member sequences for domains i and j, respectively.
This correlation measures the relatedness of the two domains. Its value ranges from 0 to 1, where 1 means 100% similarity in the two domains' evolutionary histories and 0 means no similarity. Now we need to determine the lower threshold of the correlation which indicates co-evolution. We randomly select two Pfam families and compute their correlation. Similarly, the random correlation between two SCOP domains is calculated. The distributions of the correlations are shown in Figure 11.
When multiple Pfam families are mapped to a SCOP domain, we compute the evolutionary correlation of these Pfam families. The correlation may suggest whether those Pfam families should be merged or not. If two domains reside on the same set of sequences in close vicinity and share the same set of evolutionary characteristics, then we propose those domains should be considered as co-evolved and treated as a single, larger domain. Thus, domain definitions may depend on the relative evolutionary histories.
Conclusion
In this paper, we discuss the comparative mapping of structure-based domains to sequence-based domains in order to address the question of how each of these models individually captures the evolutionary, structural and functional features of protein domains. The ultimate purpose of our comparative mapping is to provide insight into protein domain definitions.
Using domain definitions from SCOP and Pfam, we mapped the two types of domain definitions to each other using their location information for each domain instance. Mapping results reveal a general agreement between the two types of domain definitions. To further analyze the problem, we introduce several subcategories (one/many SCOP domain to one/many Pfam domain, and vice versa), and provide detailed studies of the mapping using examples from each category.
In the subcategory of one SCOP to/from one Pfam mapping, often the mapping is not perfect: the two domains only partially overlap. Analysis shows that around 62% of the cases of one-to-one mapping agree on 90% or more of their coverage. The differences are usually in the domain boundaries. This result suggests that evolutionary history of the mapped region versus the unmapped region may be examined to see how those unmapped portions are evolutionarily related to the mapped region.
In many cases, a SCOP domain family is mapped to a series of repeats of a Pfam family. These Pfam families, such as LRR, are more likely domain components without the properties of structural domains. Therefore, we would suggest Pfam remove those families. The mapping results could also be used to infer classification for SCOP domain families that do not belong to the true classes (classes larger than 7). For example, in the cases that a set of SCOP domains are mapped to one Pfam family, structural and functional relationships are suggested among the set of SCOP domains. This information may be useful for the assignment of SCOP domains to true SCOP classes. On the other hand, the Pfam database employs a flat organization and fails to indicate the relationship between Pfam families. Although Pfam introduced clans to reflect the relationship between different families, the building of clans needs input from experts and as a result, there only 15 clans in Pfam release 14.0. Our comparison of the mapping results with the Pfam clans showed that members of a clan usually correspond to a SCOP family or a SCOP superfamily.
Therefore, the comparative mapping results may be used to help Pfam generate the clans. Perhaps most interesting, several sharp disagreements between SCOP domain families and Pfam families have been discovered, and studied in some detail. Further examination of those domain families using phylogenetic analysis would be beneficial. We have proposed using evolutionary correlation between domains to measure the fitness of the domain classification. Clearly, further studies on these sharp differences are necessary and future research may be targeted in this area.
Authors' contributions
SRH conceived and coordinated the study. YZ participated in the design of the study, implemented the comparative mapping methods, performed the data and statistical analyses, as well as drafted the manuscript. SRH, JMC, and CD participated in experimental design, and edited the final draft of the manuscript. All authors read and approved the manuscript.
Acknowledgements
This work was funded in part by grants from the US Department of Energy, Office of Biological Energy Research and Office of Office of Laboratory Policy and Infrastructure, through an LBNL LDRD, under contract No. DE-AC03-76SF00098. JMC was supported by NIH grant 1-P50-GM62412. We thank Hui Xiong and Xiaofeng He for helpful discussions.
Figures and Tables
Figure 1 Mapping between Pfam families and SCOP domain families. An instance of a SCOP domain (si, i = 1, ..., 5) on its member sequence is represented by a white rectangle while that of a Pfam domain (fj, j = 2, ..., 6) is represented by a black rectangle. Striped rectangles represent their overlap. Location information is used to map a Pfam family and a SCOP domain family. Each Pfam-A family and each SCOP domain family is treated as a set of member protein sequences. The mapping process finds overlapped regions of the two types of domains on their shared member protein sequences. The overlapped regions represent where the two types of domain definitions agree.
Figure 2 Two cases of domain mapping. An instance of a SCOP domain (s*, * = l, m, n) on its member sequence is represented by a white rectangle while that of a Pfam domain (f*, * = i, j) is represented by a black rectangle. (A) A Pfam domain and a SCOP domain overlap at a very small portion of their shared member sequence. This case is considered a partial agreement between the two types of domain definitions, and the mapping score is assigned as 0.5. (B) A Pfam domain overlaps with two SCOP domains over the full lengths of the two SCOP domains, respectively. In this case, we consider the Pfam domain maps to both SCOP domains. Therefore, a score of 1 is assigned to each mapping.
Figure 3 The lengths of SCOP domains are plotted against the lengths of their corresponding Pfam families based on the mapping. Each mapping is represented by a '+', whose x-axis and y-axis values represent the lengthes of the corresponding SCOP domains and Pfam domains, respectively.
Figure 4 Examples of one-to-one exact mapping between Pfam families and SCOP domain families. The domains are graphed onto the PDB structures of their corresponding member proteins using Pymol. The first row shows Pfam domains and the second row shows their corresponding SCOP domains. The structure regions of Pfam domains are marked in green and those of SCOP domains are marked in blue. Red regions lie outside the SCOP or Pfam domains. The differences in the domain coverage on the structures indicate disagreement between the domain definitions. The differences are usually in domain boundaries. The PDB proteins 1e17A, 1o7fA, and 3htsB are used for the illustration.
Figure 5 Histogram of differences in the endpoints of the domains. The differences in the endpoints show a power law distribution: more than 50% of the mappings between Pfam families and SCOP domain families differ by less than 10 residues and only 3.4% mapped domains differ by more than 100 residues.
Figure 6 Distribution of the mapping ratio for one-to-one exact mapping. The mapping ratios are calculated with Eq. 2. Among the cases of one-to-one exact mapping, 61.24% have a mapping ratio larger than 0.9, 81.62% have a mapping ratio larger than 0.8, and 90.26% have a mapping ratio larger than 0.7.
Figure 7 Structures of SCOP domains each mapped to several copies (repeats) of a Pfam family. The corresponding Pfam families may be of type 'Family', 'Domain', or 'Repeat'. PDB proteins lib2A, 1ogqA, and lk9uA are used for the illustration. (A) SCOP domain a.118.1.8 (Pumilio repeat) corresponds to 8 copies of Pfam family PUF (Pumilio-family RNA binding repeat) of type 'Family'. The regions marked by red, pink, blue, purple, green, cyan, orange, and yellow each represent a copy of PUF. (B) SCOP domain c.10.2.8 (Polygalacturonase inhibiting protein PGIP) corresponds to 8 copies of Pfam family LRR (Leucine Rich Repeat) of type 'Repeat'. The eight copies of LRR are each marked with a unique color: red, pink, blue, purple, green, cyan, orange, and yellow. (C) SCOP domain a.39.1.10 (Polcalcin phl p 7) corresponds to 2 copies of Pfam family efhand (EF hand) of type 'Domain'. The two copies of efhand are marked in red and green, respectively.
Figure 8 A series of SCOP domains are mapped to a Pfam family. (A) The Pfam family ACR_tran (AcrB/AcrD/AcrF family) corresponds to eight SCOP domain families for PDBID 1iwgA, three of which are unique. The regions marked with red and pink are two copies of the SCOP domain family d.225.1.1 (Multidrug efflux transporter AcrB TolC docking domain; DN and DC subdomains), marked with yellow and orange are two copies of the SCOP domain family f.35.1.1 (Multidrug efflux transporter AcrB transmembrane domain), and the rest are four copies of the SCOP domain family d.58.44.1 (Multidrug efflux transporter AcrB pore domain; PN1, PN2, PC1 and PC2 subdomains). (B) The Pfam family Glyco_hydro_42 (Beta-galactosidase) mapped to a series of the SCOP domain families {c.1.8.1 (Amylase, catalytic domain), c.23.16.5 (A4 beta-galactosidase middle domain), b.71.1.1 (alpha-Amylases, C-terminal beta-sheet domain)} in PDB protein 1kwgA. They are marked in cyan, green and blue, respectively.
Figure 9 One SCOP domain mapped to different sets of Pfam families. (A) The SCOP domain d.81.1.2 is mapped to the Pfam family Homoserine-dh (marked in blue) in PDB protein lebfA. (B) The SCOP domain d.81.1.2 is mapped to the Pfam family Saccharop-dh (marked in red) in PDB protein le5qA.
Figure 10 A Pfam family corresponds to two different sets of SCOP domains, each consisting of a series of three domains. The PDB proteins 1a6d and 1iokG are used for the illustration. The SCOP domains a.129.1.1 and a.129.1.2 are marked in purple. The SCOP domains d.56.1.1 and d.56.1.2 are marked in red. The SCOP domains c.8.5.1 and c.8.5.2 are marked in blue. The Pfam domain Cpn60_TCP1 is marked in green. (A)The Pfam family Cpn60_TCP1 is mapped to the set of SCOP domain families: {a.129.1.2 + d.56.1.2 + c.8.5.2}. (B)The Pfam family Cpn60_TCP1 is mapped to the set of SCOP domain families: {a.129.1.1 + d.56.1.1 + c.8.5.1} (C)Illustration of the insertion process which supports the SCOP domain definitions for this particular case. The SCOP domain families a.129.1.1 and a.129.1.2 are the parent domains. Later the SCOP domain families d.56.1.1 and d.56.1.2 are inserted into a.129.1.1 and a.129.1.2, respectively. Finally the SCOP domain c.8.5.1 is inserted into d.56.1.1, and the SCOP domain c.8.5.2 is inserted into d.56.1.2.
Figure 11 Distribution of correlations between two Pfam domains. The Pfam families are randomly selected and their correlation is calculated as described in Section Phylogenetic Analysis. The correlation represents the relatedness of two domains. Its value ranges from 0 to 1, with 1 indicating 100% similarity in the two domains' evolutionary histories and 0 no similarity. Genome protein sequences from bacteria are used in the computation. About 76% of the domain pairs have a correlation less than 0.5.
Table 1 Pfam families with no corresponding SCOP domain families. The annotations for Pfam families were retrieved from the Pfam database.
Pfam family Type Annotation
Cytochrom_B559a Family The lumenal portion of cytochrome b559 alpha chain.
MHC_I_C Family The C-terminal region of the MHC class I antigen.
STN Family Found at the N-terminus of the Secretins of the bacterial type II/III secretory system as well as the TonB-dependent recep- tor proteins, which are involved in TonB-dependent active up- take of selective substrates.
Phe_tRNA- synt_N Domain Aminoacyl tRNA synthetase class II, N-terminal domain.
RNA_pol_Rpb1_R Repeat The repetitive C-terminal domain (CTD) of Rpb1 (RNA poly-merase Pol II).
Prion_octapep Repeat Found at the amino terminus of prion proteins and shown to bind to copper.
Table 2 Types of mapping between SCOP and Pfam families.
Type of map Example
SCOP Pfam
One SCOP domain family to exactly one Pfam family b.81.2.1 CfAFP
One SCOP domain family to a series of Pfam families e.38.1.1 {PCRF, RF-1}
A series of SCOP domain families to one Pfam family {d.179.1.1, d.58.20.1} HMG-CoA_red
A SCOP domain family to several sets of Pfam families b.41.1.1 {PRCH, PRC}; PRC
Sets of SCOP domain families to one Pfam family {f.10.1.1, b.1.18.4}; i.6.1.1 Alpha_E1_glycop
Table 3 Examples for cases where a Pfam family corresponds to a SCOP superfamily.
Pfam Type SCOP
DHH Family c.107.1.1; c.107.1.2
OsmC Family d.227.1.2; d.227.1.1
Pec_lyase_C Domain b.80.1.2; b.80.1.1
Glyoxalase Domain d.32.1.3; d.32.1.1; d.32.1.4; d.32.1.2
TOBE Domain b.40.6.1; b.40.6.3; b.40.6.2
HhH-GPD Domain a.96.1.2; a.96.1.3; a.96.1.1
NAD_binding_1 Domain c.25.1.4; c.25.1.1; c.25.1.5; c.25.1.2
Glyco_hydro_15 Family a.102.1.1; a.102.1.5
Ricin_B_lectin Repeat b.42.2.1; b.42.2.2
Prenyltrans Repeat a.102.4.3; a.102.4.2
HHH Motif a.60.2.1; a.60.4.1; a.60.2.3; a.60.2.2
Table 4 Members of Pfam clans and their corresponding SCOP domains.
Clan ID Member families Corresponding SCOP domains
1 Laminin_EGF g.3.11.2
EGF_CA g.3.11.1
EGF g.3.11.1
2 Laminin_G_2 b.29.1.4
Laminin_G_1 b.29.1.4
3 Kazal_2 g.15.1.1
Kazal_1 g.15.1.1
4 KH_1 d.52.3.1
KH_2 d.52.3.1
5 SNF2_N -
ResIII c.37.1.19
Flavi_DEAD -
DEAD_2 -
DEAD c.37.1.19
6 ENTH a.118.9.1
ANTH a.118.10.1
7 SH3_2 b.34.2.1
SH3_1 b.34.2.1
8 V-set b.1.1.1
Ig b.1.1.1
I-set b.1.1.1
C2-set b.1.1.2
C1-set b.1.1.3
9 TAFII28 a.22.1.3
TAF a.22.1.3
Histone a.22.1.3
CBFD_NFYB_HMF a.22.1.3
10 Transpeptidase e.3.1.1
Peptidase_S11 e.3.1.1
Lactamase_B -
Betalactamase e.3.1.1
==== Refs
Murzin AG Brenner SE Hubbard T Chothia C Scop: a structural classification of protein database for the investigation of sequences and structures J Mol Biol 1995 247 536 540 7723011 10.1006/jmbi.1995.0159
Abascal F Valencia A Automatic annotation of protein function based on family identication Proteins:structure, fuction, and genetics 2003 53 683 692 10.1002/prot.10449
Gulich S Uhlen M Hober S Protein engineering of an igg-binding domain allows milder elution con-ditions during affinity chromatography J Biotechnol 2000 76 233 244 10656338 10.1016/S0168-1656(99)00197-2
Jaennicke R Folding and association of proteins Prog Biophys Mol Biol 1987 49 117 237 3327098 10.1016/0079-6107(87)90011-3
Teichmann SA Rison SC Thornton JM Riley M Gough J Chothia C Small-molecule metabolism: an enzyme mosaic Trends Biotechriol 2001 19 482 486 10.1016/S0167-7799(01)01813-3
Holm L Sander C Parser for protein folding units Proteins 1994 19 256 268 7937738
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
Holm L Sander C The fssp database of structurally aligned protein fold families Nucleic Acids Res 1994 22 3600 3609 7937067
Pearl FMG Lee D Bray JE Sillitoe I Todd AE Harrison AP Thornton JM Orengo CA Assigning genomic sequences to cath Nucleic Acids Res 2000 28 277 282 10592246 10.1093/nar/28.1.277
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
Baker D Sali A Protein structure prediction and structural genomics Science 2001 294 93 96 11588250 10.1126/science.1065659
Bateman A Coin L Durbin R Finn RD Hollich V Griffiths-Jones S Khanna A Marshall M Moxon S Sonnhammer ELL Stud holme DJ Yeats C Eddy SR The pfam protein families database Nucleic Acids Res 2004 32 D138 D141 14681378 10.1093/nar/gkh121
Servant F Bru C Carrére S Courcelle E Gouzy J Peyruc D Kahn D Prodom: Automated clustering of homologous domains Briefings in Bioinformatics 2002 3 246 251 12230033
Mulder NJ Apweiler R Attwood TK Bairoch A Barrell D The interpro database, 2003 brings increased coverage and new features Nucleic Acids Research 2003 31 315 318 12520011 10.1093/nar/gkg046
Hadley C Jones DT A systematic comparison of protein structure classifications: Scop, cath, and fssp Structure Fold Des 1999 7 1099 1112 10508779 10.1016/S0969-2126(99)80177-4
Studholme DJ Rawlings ND Barrett AJ Bateman A A comparison of pfam and merops: two databases, one comprehensive, and one specialised BMC Bioinformatics 2003 4 17 12740029 10.1186/1471-2105-4-17
Elofsson A Sonnhammer ELL A comparison of sequence and structure protein domain families as a basis for structure genomics Bioinformatics 1999 15 480 500 10383473 10.1093/bioinformatics/15.6.480
Brenner SE Koehl P Levitt M The astral compendium for protein structure and sequence analysis Nucleic Acids Res 2000 28 254 256 10592239 10.1093/nar/28.1.254
Chandonia JM Hon G Walker NS Lo Conte L Koehl P Levitt M Brenner SE The astral compendium in 2004 Nucleic Acids Res 2004 32 D189 D192 14681391 10.1093/nar/gkh034
DeLano WL The PyMOL Molecular Graphics System 2002 DeLano Scientific, San Carlos, CA, USA
Murakami S Nakashima R Yamashita E Yamaguchi A Crystal structure of bacterial multidrug efflux transporter acrb Nature 2002 20 587 593 10.1038/nature01050
Aroul-Selvam R Hubbard T Sasidharan R Domain insertion in protein structures J Mol Biol 2004 338 633 641 15099733 10.1016/j.jmb.2004.03.039
Tan S Zhang Z Ng S Advice: automated detection and validation of interaction by co-evolution Nucleic Acids Res 2004 32 W69 W72 15215353 10.1093/nar/gnh072
| 15790427 | PMC1087832 | CC BY | 2021-01-04 16:02:48 | no | BMC Bioinformatics. 2005 Mar 25; 6:77 | utf-8 | BMC Bioinformatics | 2,005 | 10.1186/1471-2105-6-77 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-831580435410.1186/1471-2105-6-83Methodology ArticleBayesian coestimation of phylogeny and sequence alignment Lunter Gerton [email protected]ós István [email protected] Alexei [email protected] Jens Ledet [email protected] Jotun [email protected] Department of Statistics, University of Oxford, 1 South Parks Road, Oxford OX1 3TG, UK2 MTA-ELTE Theoretical Biology and Ecology Group, Pázmány Péter sétány 1/c 1117 Budapest, Hungary3 Department of Zoology, University of Oxford, South Parks Road, Oxford OX1 3PS, UK4 Department of Mathematical Sciences, University of Aarhus, Ny Munkegade, Building 530, DK-8000 Aarhus C, Denmark2005 1 4 2005 6 83 83 20 1 2005 1 4 2005 Copyright © 2005 Lunter 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
Two central problems in computational biology are the determination of the alignment and phylogeny of a set of biological sequences. The traditional approach to this problem is to first build a multiple alignment of these sequences, followed by a phylogenetic reconstruction step based on this multiple alignment. However, alignment and phylogenetic inference are fundamentally interdependent, and ignoring this fact leads to biased and overconfident estimations. Whether the main interest be in sequence alignment or phylogeny, a major goal of computational biology is the co-estimation of both.
Results
We developed a fully Bayesian Markov chain Monte Carlo method for coestimating phylogeny and sequence alignment, under the Thorne-Kishino-Felsenstein model of substitution and single nucleotide insertion-deletion (indel) events. In our earlier work, we introduced a novel and efficient algorithm, termed the "indel peeling algorithm", which includes indels as phylogenetically informative evolutionary events, and resembles Felsenstein's peeling algorithm for substitutions on a phylogenetic tree. For a fixed alignment, our extension analytically integrates out both substitution and indel events within a proper statistical model, without the need for data augmentation at internal tree nodes, allowing for efficient sampling of tree topologies and edge lengths. To additionally sample multiple alignments, we here introduce an efficient partial Metropolized independence sampler for alignments, and combine these two algorithms into a fully Bayesian co-estimation procedure for the alignment and phylogeny problem.
Our approach results in estimates for the posterior distribution of evolutionary rate parameters, for the maximum a-posteriori (MAP) phylogenetic tree, and for the posterior decoding alignment. Estimates for the evolutionary tree and multiple alignment are augmented with confidence estimates for each node height and alignment column. Our results indicate that the patterns in reliability broadly correspond to structural features of the proteins, and thus provides biologically meaningful information which is not existent in the usual point-estimate of the alignment. Our methods can handle input data of moderate size (10–20 protein sequences, each 100–200 bp), which we analyzed overnight on a standard 2 GHz personal computer.
Conclusion
Joint analysis of multiple sequence alignment, evolutionary trees and additional evolutionary parameters can be now done within a single coherent statistical framework.
==== Body
Background
Two central problems in computational biology are the determination of the alignment and phylogeny of a set of biological sequences. Current methods first align the sequences, and then infer the phylogeny given this fixed alignment. Several software packages are available that deal with one or both of these sub-problems. For example, ClustalW [1] and T-Coffee [2] are popular sequence alignment packages, while MrBayes [3], PAUP* [4] and Phylip [5] all provide phylogenetic reconstruction and inference. Despite working very well in practice, these methods share some problems. First, the separation into a multiple-alignment step and a phylogenetic inference step, is fundamentally flawed. The two inference problems are mutually dependent, and alignments and phylogeny should ideally be co-estimated, a point first made by Sankoff, Morel and Cedergren [6]. Indeed, a proper weighting of mutation events in multiple sequences requires a tree, which in turn can only be determined if a multiple alignment is available. For instance, ClustalW and T-Coffee compute their alignments based on a neighbour-joining guide tree, biasing subsequent phylogenetic estimates based on the resulting alignment. Moreover, fixing the alignment after the first step ignores the residual uncertainty in the alignment, resulting in an overconfident phylogenetic estimate.
This leads on to the second issue, which is that heuristic methods are used to deal with insertions and deletions (indels), and sometimes also substitutions. This lack of a proper statistical framework makes it very difficult to accurately assess the reliability of the alignment estimate, and the phylogeny depending on it.
The relevance of statistical approaches to evolutionary inference has long been recognised. Time-continuous Markov models for substitution processes were introduced more than three decades ago [7]. Inference methods based on these have been considerably improved since then [8], and now have all but replaced older parsimony methods for phylogeny reconstruction. With alignments, progress towards statistically grounded methods has been slower. The idea to investigate insertions and deletions in a statistical framework was first considered by Bishop and Thompson [9]. The first evolutionary model, termed the TKF91 model, and corresponding statistical tools for pairwise sequence alignment were published by Thorne, Kishino and Felsenstein [10]. Its extension to multiple sequences related by a tree has been intensively investigated in the last few years [11-17], and has recently also been extended to RNA gene evolution [18]. Current methods for statistical multiple alignment often computationally demanding, and full maximum likelihood approaches are limited to small trees. Markov chain Monte Carlo techniques can extend these methods to practical problem sizes.
Statistical modelling and MCMC approaches have a long history in population genetic analysis. In particular, coalescent approaches to genealogical inference have been very successful, both in maximum likelihood [19,20] and Bayesian MCMC frameworks [21,22]. The MCMC approach is especially promising, as it allows the analysis of large data sets, as well as nontrivial model extensions, see e.g. [23]. Since divergence times in population genetics are small, alignment is generally straightforward, and genealogical inference from a fixed alignment is well-understood [20,24-26]. However, these approaches have difficulty dealing with indels when sequences are hard to align. Indel events are generally treated as missing data [27], which renders them phylogenetically uninformative. This is unfortunate as indel events can be highly informative of the phylogeny, because of their relative rarity compared to substitution events. Statistical models of alignment and phylogeny often refer to missing data. Not all of these can be integrated out analytically (e.g. tree topology), and these are dealt with using Monte Carlo methods. The efficiency of such approaches depend to a great extent on the choice of missing data. In previous approaches to statistical alignment, the sampled missing data were either unobserved sequences at internal nodes [28], or both internal sequences and alignments between nodes [13], or dealt exclusively with pairwise alignments [29,30]. In all cases the underlying tree was fixed. In [31] we published an efficient algorithm for computing the likelihood of a multiple sequence alignment under the TKF91 model, given a fixed underlying tree. The method analytically sums out all missing data (pertaining to the evolutionary history that generated the alignment), eliminating the need for any data augmentation of the tree. This methodology is referred to in the MCMC literature as Rao-Blackwellization [32]. As a result, we can treat indels in a statistically consistent manner with no more than a constant multiplicative cost over existing methods that ignore indels.
The only missing ingredient for a full co-estimation procedure is an alignment sampler. Unfortunately, there exists no Gibbs alignment sampler that corresponds to the analytic algorithm referred to above. In this paper we introduce a partial importance sampler to resample alignments, based on a proposal mechanism built on a partial score-based alignment procedure. This type of sampler supports the data format we need for efficient likelihood calculations, while still achieving good mixing in reasonable running time (see Results).
We implemented the likelihood calculator and the alignment sampler in Java, and interfaced them with an existing MCMC kernel for phylogenetics and population genetics [22]. We demonstrate the practicality of our approach on an analysis of 10 globin sequences.
Results
Definition of the TKF model
The TKF91 model is a continuous-time reversible Markov model describing the evolution of nucleotide (or amino acid) sequences. It models three of the main processes in sequence evolution, namely substitutions, insertions and deletions of characters, approximating these as single-character processes. A sequence is represented as a string alternatingly consisting of links and characters connected by these links. This string both starts and terminates with a link. Insertions and deletions are modeled through a time-continuous birth-death process of links. When a new link is born, its associated character (by convention, its right neighbour) is chosen from the equilibrium distribution of the substitution process. (The original TKF91 model used a simple substitution process, the Felsenstein-81 model [27]. It is straightforward to replace this by more general nucleotide or amino acid substitution models [33].) When a link dies, its associated character dies too. The leftmost link of the sequence has no corresponding character to its left, and is never deleted. For this reason it is called the immortal link.
Since subsequences evolve independently, it is sufficient to describe the evolution of a single character-link pair. In a given finite time span, this pair evolves into a finite subsequence of characters and links. Since insertions originate from links, only the first character of this descendant subsequence may be homologous to the original character, while subsequent ones will have been inserted and therefore not be homologous to ancestral characters. The model as applied to pairwise alignments was solved analytically in [10], see also [34]. Conceptually, the model can be trivially extended to trees, but the corresponding algorithms for likelihood calculations have been developed only recently [11,12,14-16].
Because the TKF91 model is time reversible, the root placement does not influence the likelihood, an observation known as Felsenstein's "Pulley Principle" [27]). Although the algorithms we developed are not manifestly invariant under changes in root placement, in fact they are. We have used time reversibility to check correctness of our implementations.
Computing the likelihood of a homology structure
The concept of homology structure [31], also known as effective alignment [35], refers to an alignment of sequences at leaves without reference to the internal tree structure, and without specifying the ordering of exchangable columns (see below for more details). We derived a linear-time algorithm that computes the likelihood of observing a set of sequences and their homology structure, given a phylogeny and evolutionary parameters, under the TKF91 model [31]. By definition, this likelihood is the sum of the probabilities of all evolutionary scenarios resulting in the observed data. It was previously shown that such evolutionary scenarios can be described as a path in a multiple-HMM ([13,28]), and the likelihood can thus be calculated as the sum of path probabilities over all such paths, in time polynomial in the number of states. However, this straightforward calculation is infeasible for practical-sized biological problems, since the number of states in the HMM grows exponentially with the number of sequences [16]. Since our algorithm does not feature this exponential blow-up of Markov states, we termed it the one-state recursion. In contrast to previous approaches [13,28], the one-state recursion relieves us from the need to store missing data at internal tree nodes, allowing us to change the tree topology without having to resample this missing data. This enables us to consider the tree as a parameter, and efficiently sample from tree space. The concept of homology structure referred to above is key to our algorithm, and we will presently define this concept more precisely. Let A1, A2, ...Am be sequences, related by a tree T with vertex set V. Let denote the jth character of sequence Ai, and let denote its k long prefix. A homology structure on A1, ..., Am is an equivalence relation ~ on the set of all the characters of the sequences, C = {}, specifying which characters are homologous to which. The evolutionary indel process generating the homology structure on the sequences imposes constraints on the equivalence relations that may occur. More precisely, the equivalence relation ~ has the property that a total ordering, <h, exists on C such that
(Here, a = h b is equivalent to: a ≮h b and b ≮h a.) In particular, these conditions imply that the characters constituting a single sequence are mutually nonhomologous. The ordering <h corresponds to the ordering of columns of homologous characters in an alignment. Note that for a given homology structure, this ordering may not be unique (see Fig. 1). This many-to-one relationship of alignment to homology structure is the reason for introducing the concept of homology structure, instead of using the more common concept of alignment.
The one-state recursion, which calculates the likelihood of a homology structure, is a convolution of two dynamic programming algorithms. The top-level algorithm traverses the prefix set of the multiple alignments representing the homology structure (see Figure 2). This repeatedly calls on a reverse traversal algorithm on the phylogenetic tree, which sums out the likelihood contributions of substitutions and indels under the TKF91 model. See [31] for full details.
A partial Metropolized independence sampler
Because our algorithm does not require the phylogenetic tree to be augmented with missing data, proposing changes to the evolutionary tree is easy, and mixing in tree space is very good. The drawback however is that without data augmentation, it is unclear how to perform Gibbs sampling of alignments, and we have to resort to other sampling schemes. One straightforward choice would be a standard Metropolis-Hastings procedure with random changes to the alignment, but we expect slow mixing from such an approach. Another general approach is Metropolized independence sampling. Its performance depends on the difference between the proposal distribution and the target distribution, and this will inevitably become appreciable with growing dimension of the problem, as measured by the number and length of the sequences to be aligned. We therefore opted for a partial Metropolized independence sampler [36], where we partly defy the "curse of dimensionality" by resampling only a segment of the current alignment. Above increasing the acceptance ratio, this method has the added advantage of being a more efficient proposal scheme, since the time complexity of the algorithm is proportional to the square of the window size, and so leads to an effective increase in mixing per processor cycle. Metzler et al. [29] followed a parallel approach, using a partial Gibbs sampler, and showed that this resulted in faster mixing compared to a full Gibbs sampling step. Since the realignment step may change the window length (measured in alignment columns), to have a reversible Markov chain we need all window sizes to have positive proposal probability. We chose a geometric length distribution, but other distributions can be considered equally well.
The proposal algorithm
The proposal algorithm is as follows. A window size and location is proposed, the alignment of subsequences within this window is removed, and a new alignment is proposed by a stochastic version of the standard score-based progressive pairwise alignment method. First, dynamic programming (DP) tables are filled as for a deterministic score-based multiple alignment, starting at the tree tips and working towards the root, aligning sequences and profiles. We used linear gap penalties, and a similarity scoring matrix that was obtained by taking the log-odds of a probabilistic substitution matrix. The underlying phylogeny was used to define divergence times, and served as alignment guide tree. After filling the DP tables, we applied stochastic traceback. The probabilities for the three possible directions at each position was taken to be proportional to exp(αs), where s is the deterministic score and α is a scale parameter (see Fig. 3). The set of paths that emerged in this way then determined the multiple alignment. All possible alignments can be proposed in this manner, and the proposal as well as the back-proposal probabilities can be calculated straightforwardly.
Correctness of the sampler
There are two problems with the proposal sampler introduced above. First, we propose alignments instead of homology structures. We need the latter, since the algorithm derived in this paper calculates the likelihood of the homology structure, not the particular alignment. Although it would be conceptually and (for the sampler) computationally simpler to use alignments, we are not aware of any efficient algorithm that can calculate such alignment likelihoods. The second problem is that calculating the proposal probability of a particular alignment is not straightforward. Any choice of window size and location may result in the same proposal alignment. To calculate the true proposal probability of particular alignments, we need to sum over all possible windows, which is prohibitively expensive.
Fortunately, we can solve both problems efficiently. We can sample alignments uniformly inside a homology structure, and at the same time sample homology structures according to their posterior probabilities. As biologically meaningful questions refer to homologies and not particular alignments, it seems reasonable to impose a simple uniform distribution over alignments within homology structures. The second problem is solved by not calculating an alignment proposal probability, but the proposal probability of the combination of an alignment and a resampling window. For a proposal of alignment X2 and window w from a current alignment X1, we use the following Metropolis-Hastings ratio:
where H1 and H2 are homology structures corresponding to the alignments X1 and X2 respectively, |H1| and |H2| are their cardinalities (i.e. the number of alignments representing these homology structures), and T is the proposal probability. Using this ratio, the Markov chain will converge to the desired distribution π(X) = π(H)/|H|, since the detailed balance condition is satisfied. Indeed,
where the final equality holds because of the symmetry of the left-hand side. The cardinality of a homology structure, |H1|, is the number of possible directed paths in the graph spanned by the one-state recursion; in other words, the number of permutations of alignment columns that result in alignments compatible with the given homology structure (see Fig. 2). This number can be calculated straightforwardly using a dynamic programming algorithm that traverses the one-state recursion graph [31,37].
Discussion
The one-state recursion provides a method for calculating the likelihood L = Pr{A, |T, Q, λ, μ} of observing the sequences with their homology structure (loosely, "alignment") given the tree and model parameters. Here A are the amino acid sequences, is their homology structure, T is the tree including branch lengths, Q is the substitution rate matrix, and λ, μ are the amino acid insertion and deletion rates. To demonstrate the practicality of the new algorithm for likelihood calculation we undertook a Bayesian MCMC analysis of ten globin protein sequences (see Additional file: 1). We chose to use the standard Dayhoff rate matrix to describe the substitution of amino acids. As initial homology structure we used the alignment computed by T-Coffee. We co-estimated homology structures, the parameters of the TKF91 model, and the tree topology and branch lengths. To do this we sampled from the posterior,
where Z is the unknown normalising constant. We chose the prior distribution on our parameters, f (T, λ, μ), so that T was constrained to a molecular clock, and λ = μL/(L + 1) to make the expected sequence length under the TKF91 model agree with the observed lengths; here L is the geometric average sequence length. All other parameters were sampled under uniform priors. We assume a molecular clock to gain insight into the relative divergence times of the alpha-, beta- and myoglobin families. In doing so we incorporate insertion-deletion events as informative events in the evolutionary analysis of the globin family. The posterior density h is a complicated function defined on a space of high dimension. We summarise the information it contains by computing the expectations, over h, of various statistics of interest. We estimate these expectations by using MCMC to sample from h. Marginalizations for continuous variables can be done in a straightforward manner; see for example Figure 4, which depicts the marginal posterior density of the μ parameter for two independent MCMC runs, showing excellent convergence.
For alignments, the maximum a-posteriori alignment is very hard to estimate from an MCMC sample run, as there are typically far too many plausible alignments contributing to the likelihood. Indeed, we found that almost all alignments in a moderately long MCMC run (50000 samples) were unique. However, it is possible to reconstruct a reliable maximum posterior decoding [38] alignment from such a moderate long sampling run. This alignment uses the posterior single-column probabilities, which can be estimated much more reliably since many alignments share particular columns, to obtain an alignment that maximizes the product of individual column posteriors. This alignment can be obtained by a simple dynamic programming algorithm [39], see Fig. 5. It is hard to visualise alternative suboptimal alignments, but the individual posterior column probabilities clearly reveal the more and less reliable regions of the alignment. We found that the reliable alignment regions broadly correspond to the alpha helical structure of the globin sequences.
Figure 6 depicts the maximum a posteriori (MAP) estimate of the phylogenetic relationships of the sequences. This example exhibits only limited uncertainty in the tree topology, however we observed an increased uncertainty for trees that included divergent sequences, such as bacterial and insect globins (results not shown).
The estimated time of the most recent common ancestor of each of the alpha, beta and myoglobin families are all mutually compatible (result not shown), suggesting that the molecular clock hypothesis is at least approximately valid. Analysis of a four sequence dataset demonstrate consistency in μ estimates between MCMC and previous ML analyses [16] (data not shown). Interestingly, the current larger dataset supports a lower value of μ. This is probably due to the fact that no indels are apparent within any of the subfamilies despite a considerable sequence divergence. The indel rate estimated by the current cosampling procedure is greater than the estimate on a fixed multiple alignment [31] (0.0207 vs. 0.0187), but this discrepancy is not significant for the current dataset. It should be stressed that the two MCMC analyses of the globin data set presented here are purely illustrative of the practicality of the algorithm described, and no novel biological results were obtained. The two MCMC runs of 5 million states each required less than 12 hours of CPU time each on a 2.0 GHz G5 Apple Macintosh running OS X, using an unoptimised implementation of the algorithm. From these runs we sampled 50000 states each. The estimated number of independent samples (estimated sample size, ESS) for the posterior probabilities was 250 and 240, respectively (see [22] for methods), while for the indel rate μ the ESSs were calculated at 5400 and 4000. We expect analyses of data sets of around 50 sequences to be readily attainable with only a few days computation.
Conclusion
In this paper we present a new cosampling procedure for phylogenetic trees and sequence alignments. The underlying likelihood engine uses recently introduced and highly efficient algorithms based on an evolutionary model (the Thorne-Kishino-Felsenstein model) that combines both the substitution and insertion-deletion processes in a principled way [31]. We show that the proposed method is applicable to medium-sized practical multiple alignment and phylogenetic inference problems.
One motivation for using a fully probabilistic model, and for using a co-estimation procedure for alignments and phylogeny, is that this makes it possible to assess the uncertainties in the inferences. Fixing either the alignment or the phylogeny leads to an underestimate of the uncertainty in the other, and score-based methods give no assessment of uncertainty whatsoever.
We show that the confidence estimates so obtained can contain biologically meaningful information. In the case of the multiple alignment of globin sequences, peaks in the posterior column reliabilities correspond broadly to the various conserved alpha helices that constitute the sequences (see Fig. 5). In the case of the tree estimate, the non-traditional phylogeny supported by the myoglobin subtree coincided with a significant polyphyly, as indicated by the posterior tree topology probabilities, and graphically represented by significantly overlapping 95% node height confidence boxes (see Fig. 6). It is clear that such confidence information significantly contributes to the usefulness of the inference.
At the heart of the method lies a recently introduced algorithm, termed the "indel peeling algorithm", that extends Felsenstein's peeling algorithm to incorporate insertion and deletion events under the TKF91 model [31]. This renders indel events informative for phylogenetic inference. Although incurring considerable algorithmic complications, the resulting algorithm is still linear-time for biological alignments (see also Figure 1). Moreover, our approach allows efficient sampling of tree topologies, as no data is presented at internal nodes.
We also developed a method for sampling multiple alignments, which is applicable for the data augmentation scheme we used for the efficient likelihood calculations. By combining the two samplers, we can co-sample alignments, evolutionary trees and other evolutionary parameters such as indel and substitution rates. The resulting samples from the posterior distribution can be summarized in traditional ways. We obtained maximum a-posteriori estimates of alignment, tree and parameters, and augmented these with estimates of reliability.
As was already mentioned in [10], it would be desirable to have a statistical sequence evolution model that deals with 'long' insertions and deletions, instead of single nucleotides at a time. For score-based algorithms, this is analogous to the contrast between linear and affine gap penalties. It is clear that the extension of the model to include long indels would result in considerable improvements, but the algorithmic complexities are considerable. We have made progress on a full likelihood method for statistical sequence alignment under such an evolutionary model [17], but the generalization of this method seems nontrivial. We believe that here too, Markov chain Monte Carlo approaches, combined with data augmentation, will be essential for practical algorithms. However, we also believe that in certain restricted but biologically meaningful situations, such as highly conserved proteins, the TKF91 model is reasonably realistic for the co-estimation procedure presented here to be of practical interest.
Availability and requirements
The BEAST package (AJ Drummond and A Rambaut), which includes the algorithm described in this paper, is available from , with full installation and requirement details. The data set used in this paper is avaliable (see Additional file: 1)
Authors' contributions
IM conjectured and GL proved the one-state recursion. GL and IM independently implemented the algorithms, and wrote the paper. JLJ simplified the proof of the recursion, GL suggested to use it within an MCMC phylogeny cosampler, and IM suggested to use a Metropolised importance sampler and proved its correctness. GL and AD interfaced the Java algorithms to the BEAST phylogeny sampling package [40], and AD carried out the MCMC analysis. JH provided project management. All authors read and approved the final manuscript.
Supplementary Material
Additional File 1
This XML file specifies the MCMC run for the example phylogeny and alignment co-estimation given in this paper (see Figs. 4, 5, 6). To run, download the BEAST package (AJ Drummond and A Rambaut, .)
Click here for file
Acknowledgements
The authors thank Yun Song, Dirk Metzler, Anton Wakolbinger and Ian Holmes for several useful suggestions and discussions. This research is supported by EPSRC (code HAMJW) and MRC (code HAMKA). I.M. was further supported by a Békésy György postdoctoral fellowship.
Figures and Tables
Figure 1 Alignments and homology structure. (Left:) Two alignments representing the same homology structure. A "homology structure" is defined as the set of all homology relationships between residues from the given sequences; residues are homologous if they appear in the same alignment column. Our recursion includes contributions from all alignments compatible with a given homology structure (itself represented by a single alignment). (Right:) Due to the evolutionary process acting on the sequences, homology relationships (arrows) will never 'cross' as depicted. This restriction on the equivalence relation ~ is codified by <h (see text).
Figure 2 Dynamic programming table traversal. The multiple alignment prefixes (represented by o symbols) traversed by the one-state recursion, when the input is the homology structure of Fig. 1. (For clarity, the vectors are plotted in two dimensions instead of the actual three.) The homology structure is represented by the graph, and each directed path on this graph uniquely corresponds to an alignment that is compatible with the homology structure.
Figure 3 Generating the proposal alignment. This figure illustrates the stochastic sequence aligner. In the deterministic fill-in process, the three scores are s1, s2 and s3, hence the value in this cell is max{s1, s2, s3}. In the stochastic traceback phase, the three neighbor cells are choosen with probabilities proportional to exp(αsi), where α > 0 is a scaling parameter. The chosen traceback path corresponds to the proposed alignment in the usual way.
Figure 4 Posterior distribution of deletion rate μ. Estimated posterior densities of the deletion rate μ sampled according to h (see text), for two independent runs, suggesting excellent convergence. The sampled mean is 0.0207; the 95% highest posterior density (HPD) interval was estimated to be (0.0121, 0.0316).
Figure 5 Maximum posterior decoding alignment, and column reliabilities. The maximum posterior decoding alignment of ten globins (human, chicken and turtle alpha hemoglobin, beta hemoglobin, myoglobin and bean leghemoglobin). Posterior probabilities for aligned columns were estimated as their rate in the Markov chain. Common alpha helices are indicated with 'α' symbols under the alignment. A broad correspondence between peaks in the posterior alignment reliability and the position of conserved secondary structure is apparent.
Figure 6 Maximum a-posteriori phylogeny. The maximum a posteriori tree (black) relating the ten globins of Fig. 5, and 95% confidence intervals of the node heights (grey boxes). Most of the tree's topology is well determined, with the exception of the myoglobin sub-tree. Alpha and beta chain sub-families both support the traditional ordering of birds, turtles and mammals, while the three myoglobin sequences support an unconventional phylogeny, as previously observed by Hedges and Poling [41]. However, the posterior probability for the topology of the myoglobin subtree is smaller than that for the remaining topology. The marginal posterior probability (estimated from the MCMC chain) for the monophyly of human and chicken myoglobin is 83.1%, followed by the conventional grouping of turtle and chicken at 11.9%. The third topologlcal arrangement of myoglobin occurred the remaining 5% of the time, suggesting significant homoplasy in this sub-family.
==== Refs
Thompson J Higgins D Gibson T CLUSTAL-W: improving the sensitivity of multiple sequence alignment through sequence weighting, position specific gap penalties and weight matrix choise Nucleic Acids Res 1994 22 4673 4680 7984417
Notredame C Higgins D Heringa J T-Coffee: A novel method for multiple sequence alignments Journal of Molecular Biology 2000 302 205 217 10964570 10.1006/jmbi.2000.4042
Huelsenbeck JP Ronquist F MRBAYES: Bayesian inference of phylogenetic trees Bioinformatics 2001 17 754 755 11524383 10.1093/bioinformatics/17.8.754
Swofford D PAUP* 4.0 Sinauer Associates 2001
Felsenstein J PHYLIP version 3.63 Dept of Genetics, Univ of Washington, Seattle 2004
Sankoff D Morel C J CR Evolution of 5S RNA and the non-randomness of base replacement Nature New Biology 1973 245 232 234 4201431
Jukes TH Cantor CR Munro Evolution of Protein Molecules Mammalian Protein Metabolism 1969 Acad Press 21 132
Whelan S Lió P Goldman N Molecular phylogenetics: state-of-the-art methods for looking into the past Trends in Gen 2001 17 262 272 10.1016/S0168-9525(01)02272-7
Bishop M Thompson E Maximum likelihood alignment of DNA sequences J Mol Biol 1986 190 159 165 3641921 10.1016/0022-2836(86)90289-5
Thorne JL Kishino H Felsenstein J An Evolutionary Model for Maximum Likelihood Alignment of DNA Sequences J Mol Evol 1991 33 114 124 1920447
Steel M Hein J Applying the Thorne-Kishino-Felsenstein model to sequence evolution on a star-shaped tree Appl Math Let 2001 14 679 684 10.1016/S0893-9659(01)80026-4
Hein J An algorithm for statistical alignment of sequences related by a binary tree Pac Symp Biocomp, World Scientific 2001 179 190
Holmes I Bruno WJ Evolutionary HMMs: a Bayesian approach to multiple alignment Bioinformatics 2001 17 803 820 11590097 10.1093/bioinformatics/17.9.803
Hein J Jensen JL Pedersen CNS Recursions for statistical multiple alignment PNAS 2003 100 14960 14965 14657378 10.1073/pnas.2036252100
Miklós I An Improved Algorithm for Statistical Alignment of Sequences related by a Star Tree Bul Math Biol 2002 64 771 779 10.1006/bulm.2002.0300
Lunter G Miklós I Song Y Hein J An efficient algorithm for statistical multiple alignment on arbitrary phylogenetic trees J Comp Biol 2003 10 869 889 10.1089/106652703322756122
Miklós I Lunter GA Holmes I A "Long Indel" model for evolutionary sequence alignment Mol Biol Evol 2004 21 529 540 14694074 10.1093/molbev/msh043
Holmes I A probabilistic model for the evolution of RNA structure BMC Bioinf 2004 5
Kuhner MK Yamato J Felsenstein J Estimating effective population size and mutation rate from sequence data using Metropolis-Hastings sampling Genetics 1995 140 1421 1430 7498781
Griffiths RC Tavare S Ancestral inference in population genetics Stat Sci 1994 9 307 319
Wilson IJ Balding DJ Genealogical Inference From Microsatellite Data Genetics 1998 150 499 450 9725864
Drummond AJ Nicholls GK Rodrigo AG Solomon W Estimating mutation parameters, population history and genealogy simultaneously from temporally spaced sequence data Genetics 2002 161 1307 1320 12136032
Pybus OG Drummond AJ Nakano T Robertson BH Rambaut A The epidemiology and iatrogenic transmission of hepatitis C virus in Egypt: a Bayesian coalescent approach Mol Biol Evol 2003 20 381 387 12644558 10.1093/molbev/msg043
Felsenstein J Estimating effective population size from samples of sequences: Inefficiency of pairwise and segregating sites as compared to phylogenetic estimates Genetical Research Cambridge 1992 59 139 147
Stephens M Donnelly P Inference in Molecular Population Genetics J of the Royal Stat Soc B 2000 62 605 655 10.1111/1467-9868.00254
Pybus OG Rambaut A Harvey PH An integrated framework for the inference of viral population history from reconstructed genealogies Genetics 2000 155 1429 1437 10880500
Felsenstein J Evolutionary trees from DNA sequences: a maximum likelihood approach J Mol Evol 1981 17 368 376 7288891
Jensen J Hein J Gibbs sampler for statistical multiple alignment Tech Rep 429, Dept of Theor Stat, U Aarhus 2002
Metzler D Fleißner R Wakolbringer A von Haeseler A Assessing variability by joint sampling of alignments and mutation rates J Mol Evol 2001 53 660 669 11677626 10.1007/s002390010253
Metzler D Statistical alignment based on fragment insertion and deletion models Bioinformatics 2003 19 490 499 12611804 10.1093/bioinformatics/btg026
Lunter G Miklós I Drummond A Jensen J Hein J Bayesian phylogenetic inference under a statistical indel model Lecture Notes in Bioinformatics 2003 2812 228 244
Casella G Robert CP Rao-Blackwellisation of sampling schemes Biometrika 1996 83 81 94 10.1093/biomet/83.1.81
Hein J Wiuf C Knudsen B Møller MB Wibling G Statistical Alignment: Computational Properties, Homology Testing and Goodness-of-Fit J Mol Biol 2000 302 265 279 10964574 10.1006/jmbi.2000.4061
Miklós I Toroczkai Z An improved model for statistical alignment Lecture Notes on Computer Science 2001 2149 1 10
Dress A Morgenstern B Stoye J The number of standard and of effective multiple alignments App Math Lett 1998 11 43 49 10.1016/S0893-9659(98)00054-8
Liu JS Monte Carlo Strategies in Scientific Computing 2001 Springer
Giegerich R Meyer C Steffen P A Discipline of Dynamic Programming over Sequence Data Science of Computer Programming 2004 51 215 263 10.1016/j.scico.2003.12.005
Durbin R Eddy S Krogh A Mitchison G Biological sequence analysis 1998 Cambridge University Press
Holmes I Durbin R Dynamic programming alignment accuracy J Comp Biol 1998 5 493 504
Drummond AJ Rambaut A BEAST v1.2.2 2004
Hedges SB Poling LL A molecular phylogeny of reptiles Science 1999 283 945 946 10075558 10.1126/science.283.5404.998
| 15804354 | PMC1087833 | CC BY | 2021-01-04 16:02:49 | no | BMC Bioinformatics. 2005 Apr 1; 6:83 | utf-8 | BMC Bioinformatics | 2,005 | 10.1186/1471-2105-6-83 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-861581396810.1186/1471-2105-6-86SoftwareHDBStat!: A platform-independent software suite for statistical analysis of high dimensional biology data Trivedi Prinal [email protected] Jode W [email protected] Jelai [email protected] Gary L [email protected] Vinodh [email protected] Stanislav O [email protected] Kyoungmi [email protected] Tapan [email protected] Jacob PL [email protected] Amit [email protected] Grier P [email protected] David B [email protected] Section on Statistical Genetics, Department of Biostatistics, University of Alabama at Birmingham, Birmingham, AL 35294, USA2 Department of Mathematics and Statistics, University of Missouri-Rolla, Rolla, MO 65409, USA3 Department of Agronomy, Iowa State University, Ames, IA 50011, USA4 Pennington Biomedical Research Center, 6400 Perkins Rd., Baton Rouge, LA 70808, USA2005 6 4 2005 6 86 86 29 11 2004 6 4 2005 Copyright © 2005 Trivedi 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
Many efforts in microarray data analysis are focused on providing tools and methods for the qualitative analysis of microarray data. HDBStat! (High-Dimensional Biology-Statistics) is a software package designed for analysis of high dimensional biology data such as microarray data. It was initially developed for the analysis of microarray gene expression data, but it can also be used for some applications in proteomics and other aspects of genomics. HDBStat! provides statisticians and biologists a flexible and easy-to-use interface to analyze complex microarray data using a variety of methods for data preprocessing, quality control analysis and hypothesis testing.
Results
Results generated from data preprocessing methods, quality control analysis and hypothesis testing methods are output in the form of Excel CSV tables, graphs and an Html report summarizing data analysis.
Conclusion
HDBStat! is a platform-independent software that is freely available to academic institutions and non-profit organizations. It can be downloaded from our website .
==== Body
Background
One of the most critical tasks in the field of biology is identifying how and which genes interact with each other under different conditions. Until a few years ago, researchers were only able to accomplish this task for a limited number of genes because the traditional methods in molecular biology allowed them to assess only one gene at a time. The advent of microarray technology has provided investigators the opportunity to simultaneously assess the expression levels of thousands of genes. Microarrays also generate a large amount of data in short period of time. Extracting statistically valid and biologically relevant information from such massive data sets is a major challenge. HDBStat! is a user-friendly and platform-independent software designed for the statistical analysis of microarray data using well-validated methods for quality control of experiments and the identification of differentially expressed genes.
Implementation
Data analysis in HDBStat! is divided into four steps – data import, data processing, quality control and hypotheses testing (Figure 1).
Data import
Data is imported into HDBStat! using two files, a gene expression data file (Figure 2) and chip level information file (Figure 3), both of which must be Microsoft Excel '97 or more recent format (.xls), or Comma Separated Values (.csv) files [see Additional file 1 and Additional file 2]. The gene expression data file contains the output from the chip image processing software, such as MAS 5.0, Bioconductor, or GenePix. The chip level file contains experimental variables such as treatment, time, experiment, and if appropriate, pairing variables for the chips. Upon import some descriptive statistics are automatically generated about the raw data such as Pearson's correlations between chips, mean, standard deviation, minimum and maximum values of gene expression levels for each chip and displayed in graphical and tabular formats.
Data preprocessing
Optionally, a normalization and/or transformation method(s) can be applied prior to the primary statistical analyses. Normalization is a procedure intended to remove variability among chips that is unrelated to treatment conditions of interest. HDBStat! offers Chip Mean normalization, which divides each observation by the chip mean, and Quantile-Quantile normalization, which ranks each observation on the chip based on expression value and then converts to the value of a deviation that would be expected from the standard normal distribution based on the observation rank. Quantile-quantile normalization results in data from each chip with a mean of zero and standard deviation of 1.0. Transformation is a process of applying a mathematical function to every observation in a data set in order to better satisfy assumptions of certain statistical models used for analysis. HDBStat! offers three different scales of logarithmic transformation, base-2, base-e, or base-10. Combinations of normalizations and transformations may be selected.
Quality control
HDBStat! provides a unique quality control procedure based upon Deleted Residuals (DR). Deleted residuals have traditionally been used in the statistical analysis of data when the number of observations in a group are small or may be influenced by outliers, as in the case in microarrays. In HDBStat!, the deleted residuals for each gene on each chip is calculated by taking the observed value of a gene on a chip subtracting the mean for the gene across all other chips in that group divided by the standard deviation of the mean for the gene across all the other chips in that group. The Probability Density Function (PDF) for the deleted residuals for a gene will follow a Student's t-distribution with n-2 degrees of freedom where n is the number of chips in the treatment group. If we assume that the genes across a chip are independent identically distributed (IID) the distribution of the deleted residuals should approximate a Student's t-distribution with n-2 degrees of freedom. The difference of the observed data from the expected t-distribution is graphically illustrated (Figure 4) and the significance of the difference is tested using a Kolmogorov-Smirnov test. If a chip is significantly different from the t-distribution it may be an indication that the particular chip is an outlier compared with the other chips in the group. Further, the user has the opportunity to remove chip(s) from the analysis and re-analyze the data.
Hypotheses testing
Currently, HDBStat! performs a series of pair wise comparison tests. Based on the information provided by user in chip level information file, a combination of all possible hypotheses is displayed in the user interface. User must select at least one hypothesis in order to perform two group comparisons.
HDBStat! includes parametric and non-parametric methods for estimating the significance of changes in gene expression between groups. Student's t-test, for which the user can choose an equal-variance t-test, which uses a pooled variance across treatments, or Welch's t-test, which assumes unequal variances between the two treatment groups [11]. Another method based on Chebyshev's inequality, Chebby Checker is extremely robust against departures from normality and equality of variance between treatment groups, but it also has very low power [2]. The Chebby Checker is useful for identifying genes that are almost certainly differentially expressed without considering any statistical assumptions. In addition a bootstrap resampling method [6,8] is implemented. One can either conduct an exact bootstrap (all possible permutations) or a random (used specified number of permutations) bootstrap. The bootstrap procedures implement both pivots and smoothes in order to calculate the significance more accurately. As exact bootstrap is more accurate than random bootstrap, it is preferred for computationally feasible cases, but once the n per groups exceeds 6 it is difficult to implement.
Because of the large number of simultaneous hypotheses tested in high dimensional biology experiments, an adjustment for multiple testing is appropriate in order to avoid falsely calling too many genes significant. In HDBStat! several multiplicity control methods are available to be applied to any hypothesis testing method. The available multiplicity control adjustments are Bonferroni [4], Sidak [10], two False Discovery Rate (FDR) estimation methods [3,5], and a method based on a mixture modeling of observed p-values [1], referred to as the "Mix-o-matic" (Figure 5) in the HDBStat! software. The Bonferroni and Sidak methods provide experiment-wise (or Family-wise) type I control. The FDR methods are designed to control the proportion of false positives among all genes declared differentially expressed. The mixture modeling method allows for the Bayesian estimation of the probability that each gene is a false positive or negative and this approach is also conveniently for projecting power estimates for future studies [9].
For the planning of future studies HDBStat! implements the method of Gadbury et al to extrapolate power from pilot data [9]. HDBStat! allows for the calculation of the expected discovery rate (EDR), posterior true positive, and posterior true negative rates for large and smaller samples sizes than were entered as pilot data. (Figure 6)
If an investigator is interested in empirically comparing the size of the observed differences in gene expression, an Empirical Bayes method is provided to provide shrinkage estimators of the true differences in gene expression [7]. In addition group means and fold changes in expression are calculated and output to results directory specified by the user.
Programming details
HDBStat! is implemented using the Java programming language using various licensed and open source libraries such as Visual Numerics JMSL, Jakarta POI, Velocity, and JFreeChart. Extensive software testing is performed using JUnit library.
Results
At the completion of calculating the deleted residuals and analyzing a hypothesis, results are output to a date/time-stamped directory into the user specified directory. Chip level statistics, preprocessed data, deleted residuals, standard outliers, various pair wise comparison tests (Table 1), mix-o-matic and power analysis results are output in the form of Excel CSV files. Graphs generated from chip level statistics, deleted residuals, mix-o-matic and power analysis results are output in .png format image files. HDBStat! also generates a HTML file that provides a summary of the analysis including the hypotheses tested, chips in each group. This mechanism of outputting results provides the user an opportunity to view quality control results and modify hypotheses, preprocessing methods, and/or chip selections before proceeding to the next step.
Discussion
The goal of HDBStat! is to help researchers analyze microarray data to extract valid inferences, estimates and interpretations via a flexible and user-friendly graphical interface. It allows the user to skip preprocessing and quality control methods by simply not selecting those methods. After previewing the preliminary results of raw data, preprocessed data or deleted residuals, user has flexibility to drop a chip by simply un-checking checkbox in the user interface. This feature allows the user to design any number of possible comparisons while the analysis is in progress.
To assist novice users with using HDBStat!, video clips demonstrating how to analyze paired and unpaired data, examples of how to set up input files for paired and unpaired data analyses, screen shots, and FAQ are available on our website. A detailed description of methods as well as additional explanations of the output files in this software is also available in a PDF format on our website.
Additional statistical methods and features are added on an ongoing basis. Support for data import from a text file and results output to a text file will be available for large data sets. In the current version, only single channel or common reference design microarray data can be analyzed using two group comparisons. In the near future, we will add the capability to analyze two channel data and support for ANOVA, and GLM.
There are many software programs available to analyze microarray data, each offering various features and functions. In Table 2, we have compared the features and functions of HDBStat! to SAM, BRB Array Tools and TM4.
Availability and requirements
System requirements for an end-user are the Java Runtime Environment (JRE 1.4.2 or higher), at least 256 MB RAM and 25 MB hard disk space. Using Java Web Start technology, HDBStat! can be easily downloaded from our website at .
Authors' contributions
JWE, JPLB, KK, GPP wrote the implementation specification documents for various methods. GLG and DBA developed mix-o-matic method, developed prototype implementation in S-Plus and wrote the implementation specification document. GPP and JWE developed the deleted residuals approach. DBA and JWE developed prototype implementation of Empirical Bayes Estimates. JWE and PT developed a prototype implementation of statistical methods in SAS. JW, PT, VS and TM implemented and tested the java code. JWE, PT, JW, SOZ, GPP, VS, TM and AP tested the software. GPP and JWE directed the content of HDBStat! All authors read and approved the manuscript.
Acknowledgements
This work is supported by grants from the UAB HSF GEF, NSF 0217651 and NIH U54CA100949.
Figures and Tables
Figure 1 Data analysis in HDBStat! is divided into four steps – data import, data preprocessing, quality control and hypotheses testing. At each step, user input is required and in return, the results are displayed in the interface and/or output to a file.
Figure 2 Screenshot of gene expression data file in Excel format
Figure 3 Screenshot of chip level information file in Excel format
Figure 4 Deleted residuals graph
Figure 5 Mix-o-matic graph
Figure 6 Power analysis graph
Table 1 Hypothesis testing results
Probe Mean (group1, Raw data) Mean (group2, Raw data) Fold change t (Equal variance t-test) p-value (Equal variance t-test) PTP, unrestricted (Equal variance t-test)
AFFX-MurIL2_at 105.16 200.34 1.905097 -0.81193 0.440318 0.226179
AFFX-MurIL10_at 176.52 253.58 1.436551 -0.49777 0.632041 0.194216
AFFX-MurIL4_at 146.56 197.86 1.350027 -0.05477 0.957668 0.147303
AFFX-MurFAS_at 670.06 487.42 0.727427 2.264698 0.05333 0.32061
AFFX-BioB-5_at 4074.98 5165.36 1.267579 -1.10225 0.302405 0.251946
AFFX-BioB-M_at 10258.4 12376.04 1.20643 -0.85246 0.418745 0.230018
AFFX-BioB-3_at 5600.76 6784.9 1.211425 -0.82656 0.432446 0.227573
AFFX-BioC-5_at 15688.4 17406.88 1.109538 -0.23558 0.819676 0.16602
AFFX-BioC-3_at 11812.54 13288.74 1.124969 -0.2571 0.803593 0.168321
AFFX-BioDn-3_at 65025.2 68451.72 1.052695 0.467328 0.65273 0.190965
AFFX-CreX-5_at 124944.9 152682.8 1.222001 -0.39488 0.70325 0.183174
AFFX-CreX-3_at 166352 192790.6 1.158932 -0.26323 0.799024 0.168979
AFFX-BioB-5_st 611.6 677.8 1.108241 0.280987 0.785851 0.170886
AFFX-BioB-M_st 556.26 779.84 1.401934 -1.1822 0.27107 0.25832
AFFX-BioB-3_st 195.62 359.32 1.836827 -1.51174 0.169049 0.281499
AFFX-BioC-5_st 114.36 183.8 1.607205 -2.17372 0.061461 0.316596
AFFX-BioC-3_st 57.36 94.86 1.653766 -0.62822 0.547374 0.2079
AFFX-BioDn-5_st 1981.88 2327.5 1.17439 -0.99146 0.350498 0.242595
AFFX-BioDn-3_st 1487.74 2265.06 1.522484 -2.01951 0.078118 0.309434
AFFX-CreX-5_st 950.56 1221.04 1.284548 -1.42444 0.192137 0.275814
AFFX-CreX-3_st 2610.3 3429.5 1.313834 -1.21894 0.257585 0.261147
AFFX-DapX-5_at 204.42 267.12 1.306721 0.187522 0.85592 0.16092
AFFX-DapX-M_at 299.3 347.62 1.161443 0.119799 0.907597 0.153866
AFFX-DapX-3_at 192.64 95.82 0.497404 1.5611 0.157123 0.284581
AFFX-LysX-5_at 132.46 115.9 0.874981 0.445594 0.667702 0.188634
AFFX-LysX-M_at 174.16 204.14 1.172141 -0.09641 0.92557 0.151477
AFFX-LysX-3_at 47.8 167.94 3.513389 -1.2216 0.256632 0.261348
AFFX-PheX-5_at 33.04 55.5 1.679782 -2.04285 0.07534 0.310549
AFFX-PheX-M_at 40.64 29.66 0.729823 1.223746 0.255865 0.261511
AFFX-PheX-3_at 497.66 217.12 0.436282 2.826877 0.022257 0.342932
AFFX-ThrX-5_at 92.06 184.74 2.006735 -0.26462 0.797994 0.169127
AFFX-ThrX-M_at 225.32 272.46 1.209214 -0.33299 0.747704 0.17649
AFFX-ThrX-3_at 114.54 95.08 0.830103 -0.17193 0.86776 0.15928
AFFX-TrpnX-5_at 73 96.68 1.324384 -0.82529 0.433126 0.227452
AFFX-TrpnX-M_at 30.12 37.74 1.252988 -0.31889 0.757976 0.174969
Table 2 Comparison of HDBStat! with other software packages. All these software packages are still in active development and new functions will undoubtedly be added over time.
HDBStat! SAM BRB-Array Tools 3.2.2 TM4
Two color data handling Common reference and balanced block designs No Yes Yes
Database No No No Yes
Ratio Statistics Yes No Yes Yes
Normalization Yes Yes Yes Yes
Max number of arrays No limit 255 249 No limit
Discriminate Analysis No No Yes Yes
ANOVA Yes No Yes Yes
Bootstrapping Yes Yes Yes Yes
Non-normal and heteroskedastic data handling Yes Yes Via normalization Via normalization
Non-parametric statistics Yes No Yes No
Cluster analysis No No Yes Yes
FDR (number) 8 1 2 1
Family Wise Error rate corrections 2 0 1 1
Quality Control Yes No No Yes
Power Analysis Yes No No No
Automatic report generation Yes No Yes No
Gene Class testing No No Yes No
Automatic Annotation No Link out Yes No
Platform Single program implemented in Java & available via Java Web Start technology Microsoft Excel Add-in Microsoft Excel Add-in 4 separate programs, 3 of which implemented in java & 1 in C++
==== Refs
Allison DB Gadbury GL Moonseong H Fernandez JR Lee C Prolla TA Weindruch R A mixture model approach for the analysis of microarray gene expression data Comp Statist & Data Anal 2002 39 1 20 10.1016/S0167-9473(01)00046-9
Beasley TM Page GP Brand JPL Gadbury GL Mountz JD Allison DB Chebyshev's inequality for non-parametric testing with small N and a in microarray research J R Statist Soc C 2004 53 95 108 10.1111/j.1467-9876.2004.00428.x
Benjamini Y Hochberg Y Controlling the false discovery rate: a practical and powerful approach to multiple testing J R Statist Soc B 1995 57 289 300
Bland JM Altman DG Multiple significance tests: the Bonferroni method BMJ 1995 310 170 7833759
Benjamini Y Yekutieli D The control of the false discovery rate in multiple testing under dependency Ann Statist 2001 29 1165 1188 10.1214/aos/1013699998
Davison AC Hinkley DV Bootstrap methods and their application 1997 Cambridge University Press, United Kingdom
Edwards JW Page GP Gadbury G Heo M Kayo T Weindruch R Allison DB Empirical Bayes estimation of gene-specific effects in micro-array research Funct Integr Genomics 2005 5 32 9 15455262 10.1007/s10142-004-0123-0
Effron B Tibshirani RJ An Introduction to the Bootstrap 1993 Chapmann and Hall New York
Gadbury GL Page GP Edwards JW Kayo T Prolla TA Weindruch R Permana PA Mountz J Allison DB Power and Sample Size Estimation in High Dimensional Biology Stat Meth Med Res 2004 13 325 338
Sidak Z Rectangular confidence regions for the means of the multivariate normal distributions J Am Stat Assoc 1967 62 626 633
Welch BL The significance of the difference between two means when the population variances are unequal Biometrika 1938 29 350 362
| 15813968 | PMC1087834 | CC BY | 2021-01-04 16:02:47 | no | BMC Bioinformatics. 2005 Apr 6; 6:86 | utf-8 | BMC Bioinformatics | 2,005 | 10.1186/1471-2105-6-86 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-931581999210.1186/1471-2105-6-93SoftwareWindows .NET Network Distributed Basic Local Alignment Search Toolkit (W.ND-BLAST) Dowd Scot E [email protected] Joaquin [email protected] Javier R [email protected] Melvin J [email protected] Paxton R [email protected] Livestock Issues Research Unit, Agriculture Research Service, USDA, Lubbock, TX, USA2 Plant Stress and Germplasm Development Research Unit, Agriculture Research Service, USDA, Lubbock, TX, USA2005 8 4 2005 6 93 93 1 11 2004 8 4 2005 Copyright © 2005 Dowd 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 one of the most common and useful tools for Genetic Research. This paper describes a software application we have termed Windows .NET Distributed Basic Local Alignment Search Toolkit (W.ND-BLAST), which enhances the BLAST utility by improving usability, fault recovery, and scalability in a Windows desktop environment. Our goal was to develop an easy to use, fault tolerant, high-throughput BLAST solution that incorporates a comprehensive BLAST result viewer with curation and annotation functionality.
Results
W.ND-BLAST is a comprehensive Windows-based software toolkit that targets researchers, including those with minimal computer skills, and provides the ability increase the performance of BLAST by distributing BLAST queries to any number of Windows based machines across local area networks (LAN). W.ND-BLAST provides intuitive Graphic User Interfaces (GUI) for BLAST database creation, BLAST execution, BLAST output evaluation and BLAST result exportation. This software also provides several layers of fault tolerance and fault recovery to prevent loss of data if nodes or master machines fail. This paper lays out the functionality of W.ND-BLAST. W.ND-BLAST displays close to 100% performance efficiency when distributing tasks to 12 remote computers of the same performance class. A high throughput BLAST job which took 662.68 minutes (11 hours) on one average machine was completed in 44.97 minutes when distributed to 17 nodes, which included lower performance class machines. Finally, there is a comprehensive high-throughput BLAST Output Viewer (BOV) and Annotation Engine components, which provides comprehensive exportation of BLAST hits to text files, annotated fasta files, tables, or association files.
Conclusion
W.ND-BLAST provides an interactive tool that allows scientists to easily utilizing their available computing resources for high throughput and comprehensive sequence analyses. The install package for W.ND-BLAST is freely downloadable from . With registration the software is free, installation, networking, and usage instructions are provided as well as a support forum.
==== Body
Background
What is BLAST?
Basic Local Alignment Search Tool (BLAST) answers the question: "What known nucleotide or amino acid sequence in an existing sequence database is most similar to an unknown input (query) sequence?" BLAST was originally intended for comparison of unknown and newly sequenced genetic sequences against annotated database sequences to find those sequences with the closest biological similarity to the query [1,2]. Many papers describe the BLAST algorithm and concepts for similarity searching in detail [1-3].
With the advent of high-throughput sequencing technologies, there is an ever increasing input of new genetic data, which typically needs to be evaluated using BLAST. As this ever increasing volume of new data is analyzed using BLAST and then annotated the new sequences are subsequently added to the public databases, which in turn continues to increase their size. This is a vicious cycle, in which new sequence data is generated at increasing rates and databases are continually increasing in size. The BLAST search process is therefore becoming an increasingly time-consuming productivity bottleneck in genomics. In example, a typical laboratory workstation (2.4 Ghz Pentium 4 processor, 1 GB physical memory, 60 GB SATA hard drive) can search one query sequence (800 bp) against a local copy of NCBI nucleotide (nt) database (320 million bp database) in just under 3 minutes. Thus, a typical high-throughput BLAST search, which involves 10,000 sequences, can take over 20 days to complete on a single workstation. This is an inordinate amount of time to wait for results.
There is a growing need to find ways of reducing these BLAST productivity bottlenecks when evaluating new sequence data. One of the most logical approaches is to increase overall BLAST performance by enhancing scalability, user-friendliness, and reliability. There are an increasing number of commercial and several non-commercial software packages that have been developed (Table 1). The obvious drawback to commercial products is a very high price tag. On the other hand, non-commercial and freely available BLAST solutions (Table 1) have been shown to increase the throughput of large BLAST jobs, but typically require a high level of computer literacy and networking skill to enable them to be effectively utilized. As examples, WU-BLAST [3] provides a dramatic increase in the speed and efficiency of the BLAST algorithm itself, but it is currently unavailable for Windows environments, does not have distribution capabilities inherent, and requires someone with above average computer skill to setup, operate and maintain. Similarly, MPI-BLAST [4], which is arguably the most popular and powerful distributed BLAST implementation available today, distributes BLAST across Windows, UNIX-based, or heterogeneous networks, yet need for extensive networking skill, lack of user interfaces, and low fault tolerance can result in limited utility for the average end-user.
Ultimately the high price, lack of user-friendliness, or other similar factor reduces the utility of most available technologies for a majority of scientists that could truly benefit from such high-throughput BLAST tools. Thus, after evaluating the available free software, we identified the need within the research community for a comprehensive BLAST toolkit with a high degree of user-friendliness and functionality. We sought to create a full Windows toolkit with the following characteristics:
• Utilizes the most common operating system (Windows 2000 and Windows XP)
• Utilizes Local Area Network (LAN) connected Windows workstations as a distributed environment
• Freely available for non-commercial and academic use.
• Does not require extensive networking, hardware or software skills to setup and utilize.
• GUI based formatting of custom databases from any fasta file
• GUI based importing and distribution of existing BLAST databases (architecture dependent)
• Automatic scanning of LANs for Windows workstations that are available and ready to run WND-BLAST
• Ability for users at child nodes to cancel computationally intensive BLAST searches without disrupting the entire job if they happen to require more resources than are available during searches.
• Ability to BLAST a multi-FASTA file (single file containing more than one fasta sequence)
• Ability to BLAST a folder containing multiple single fasta files (single folder with multiple files)
• Automatic creation and management of individual projects on any workstation in the network.
• Generation and real-time viewing of BLAST log files to show network performance and progress
• Progress bars for most time consuming functions such as database formatting and large high throughput BLAST searches
• Logical formatting and visualization of BLAST search outputs using any of the fields available (e.g. bit score, hit def, E-value etc.)
• Exporting of custom annotated FASTA files containing any of the BLAST output data fields or manually input data.
• Exporting of fasta files containing any combination of hits, "no hits", or "false hits".
• Ability to manually and automatically curate BLAST output files
• Indexing of input and output files related to annotation allowing for rapid database curation and intuitive manipulation and mapping of data along with other experiments.
• GUI based installation and easy to follow instructions for setting up required Windows network shares.
• A useful help file.
The solution developed to address all of these issues has been named W.ND-BLAST, which stands for Windows .NET Distributed Basic Local Alignment Search Toolkit.
Implementation
Programming language
The package was written using Microsoft Visual Studio .NET 2003 and C# as the primary programming language. The software relies on the use of .NET 1.1 framework, which can be freely downloaded from the Microsoft website .
Auxiliary programs
W.ND-BLAST has encapsulated the NCBI BLAST software including blastall, formatdb and scoring matrices (i.e. Blosum62, PAM). W.ND-BLAST only requires a user supplied database and input sequences for proper execution. W.ND-BLAST has the ability to execute any of the sub-programs of blastall (i.e. blastx, tblastx, blastn etc).
W.ND-BLAST Toolkit overview
The W.ND-BLAST toolkit consists of several integrated software solutions
• Project Control Panel
• Database Engine
• BLAST Engine
• W.ND BLAST Distribution Algorithm
• BLAST Output Viewer
• Annotation Engine
Project Control Panel
Considerable attention was put into the design of the W.ND-BLAST protocol, which is based on creation and organization of separate BLAST projects. Once the software is initiated, the user is presented with a simple Project Control Panel (PCP) (Figure 1). This control panel has two options: A) create a new project or B) work on a previous project. The user creates a new project by entering the investigators last name, the title of the project, and a database identifier or date. After the researcher defines the new job the combined Database and BLAST Engine (DBE) interface opens and a unique project folder is automatically created for the new project (Figure 2). Alternatively the second option available at the (PCP) is to work on an existing or previous project. If an existing project is selected the BLAST Output Viewer (BOV) is opened along with a list of available subprojects (Figure 3). Subprojects can represent multiple BLAST outputs of the data against different or updated databases or sub-queries from the input sequences.
Database Engine
The Database Engine is a separate part of the previously mentioned DBE (Figure 2), which allows for creation and distribution of new databases using a fully functional database formating (formatdb) interface. The formatdb interface provides all available parameters with the most common parameters defaulted. Thus, when formatting a new database only the location of the fasta file being formatted must be identified (using a browse feature) and the type of fasta file (protein or nucleotide) must be entered. The database is formatted and entered into the software appearing as an available database. Any number or type of database can be created and made available. The Database Engine also allows for automatic location and importing of existing databases, which are copied into the database (DB) directory and added to the available database file. The software also has the ability to scan a computer or specific folder and will find any existing databases. Following scanning for available workstations in the next section, the databases can be automatically distributed to any or all of the worker nodes after formatting or importing.
BLAST Engine
The BLAST Engine, which is the primary part of the DBE (Figure 2), provides the option to scan the LAN for available network workstations (nodes) running the W.ND-BLAST server. It also provides the alternate option to BLAST only on the local machine. The software scans and finds only those LAN workstations that are available and configured to run WND-BLAST. The same WND-BLAST package provided for the master node software must be installed on each of the worker nodes as well. The BLAST Engine allows for the selection of child nodes, fasta file or folder, and BLAST parameters. The user is also presented with a list of BLAST databases to choose. Note: the user must format or import these databases. Progress bars keep track of the job and a real-time log file viewer provides feedback to the user on the overall progress of the BLAST as well as on which workstations are working.
W.ND BLAST distribution algorithm
The software utilizes simple client/server architecture with a single source client, referred to as 'master node' and multiple servers that are referred to as 'child nodes'. An included W.ND-BLAST server must be installed and running on all participating child nodes. Each child node is used solely for communication with the master node and executing the BLAST program. The BLAST algorithm used is unchanged from the most current NCBI BLAST, version 2.2.9. As newer versions are released by NCBI the algorithm can be updated independently with an integrated update feature or with new versions of W.ND-BLAST provided through the Livestock Issues Research Unit's website. The child nodes are queried by the master node upon execution of the BLAST algorithm. Communication between the master and child nodes is achieved and maintained through a well known port using an underlying Transmission Control Protocol (TCP) connection.
The child nodes are located by W.ND-BLAST, which performs a search of all workstations within range (e.g. domain or subnet mask) of the master or local machine. This is accomplished by creating a unique multicast group for the master and child nodes. Once communication has been established by the master node, the child nodes are added to the available workstation pool. The user is given a choice of using all or any combination of child nodes in a pool for their job. Thereafter, the available child node pool can be used to add child nodes, while the BLAST is in progress. This is useful when a user knows his colleagues no longer need their compute resources and wishes to add this newly freed resource to his currently running BLAST job. Before the master node executes a BLAST query, each participating child node must have the database the user wishes to query available locally. If not available locally, the master will distribute the database to the child nodes lacking the database. Once each child node has the required database, the master begins running queries on that child.
To add flexibility to input data processing, the user has the option to select a single multi-fasta file or an entire folder containing individual fasta files. If a multi-fasta file is selected, W.ND-BLAST automatically parses the file into single fasta files. If a whole folder is selected, W.ND-BLAST will attempt to BLAST each fasta file contained in that folder, but each file in the directory must only contain a single sequence. Once the user identifies either a file or folder every sequence in the pool is marked as not done in the master node job log at initialization of the "BLASTem" command. The sequences are then marshalled and sent one by one to child nodes. The master log reflects this distribution by updating the status of each sequence to working. Upon receipt, the child node un-marshals the query and begins BLAST operations. After the initial sequence distribution to each node, the master node is placed in a wait state tied to that sequence. The master node remains in this state until it receives a call-back from one of the child nodes or there are zero sequences remaining in its waiting sequence pool. Every child node, upon completion of a BLAST, sends a call-back message to the master informing of its finished state. The master node then retrieves the BLAST results and writes them to a file in the project folder. Upon receipt of the call-back, the master node also sends the child node the next query sequence from the pool to begin BLAST operations and update the sequence status to done. This process continues for every sequence until the sequence pool is empty.
BLAST Output Viewer
Once a BLAST is completed, the user has the option to view the BLAST output in the BLAST Output Viewer (BOV, Figure 3). The viewer scans through each BLAST output, one input sequence at a time, and displays as many hits for each sequence as the user desires. The viewer displays hits in a tabular format that can be rearranged based upon the user's desired preference. The viewer works with any type of XML formatted BLAST output such as BLASTx, BLASTn, BLASTp, tBLASTx, etc. There is also an alignment option that brings up a separate window that displays any of the output alignments. The alignment view of the sequence shows the query-seq, hit-seq, and mid-line fields of the hit aligned for easy viewing and analysis (Figure 4). The curation of BLAST hits is performed by clicking the checkbox corresponding to a single BLAST hit. This manual curation is based on which hits the user decides are significant based on his/her criteria. The researchers' criteria may consist of bit-score, e-value, and hit length for example. Curated hits can be exported as tables, association files and even annotated fasta sequences.
Annotation Engine
The BOV acts not only as a way to view and curate high-throughput BLAST data, but also has an effective way to perform automated annotations based upon top hits. The type of annotation can be defined dynamically based upon hit definition and can include any of the BLAST output fields that are available. The Annotation Engine (Figure 5) is used to export the BLAST output data according to the user's specifications. The Annotation Engine can automatically create annotated multi-fasta files, tab delimited text files containing all annotation, or common association files that are fully annotated based upon the user's criteria or selections.
In addition, as one of the most powerful aids to downstream analysis, the software automatically creates separate FASTA file folders within a project that contain the "no hits". This allows the "no-hits" from previous jobs to be queried against separate databases or future updated database versions. Similarly, during manual curation (as described above), those hits not "curated" or considered significant by the scientist are considered "false hits". As with the "no-hit" results the "false-hit" files can be exported separately as fasta files. As indicated, no-hits are those queries which did not get a significant match against the database.
An example of a typical annotated fasta output of an unknown sequence (contig 111) searched against a typical protein databases that hit against the database sequence with accession number Q143235 could be exported as follows (the following sequence is for demonstration only): >contig 111- similar to GB:AF166120.1 -E-value 1 × 10e-180- transcription activator AAGCTTCACGCCCAATTCTGCATCATTTTCATAAAGAGCAGGATTGGACAATACTTGAAAACTCAGTTTATTGGCTGCGTCGGAGGCGTTGGAGATCAACTCACGTAGG...
Workstation test network
The workstation test network used for testing was limited to the number of workstations available within our facility (25 windows workstations). All of the workstations have various versions of Microsoft Windows (e.g. XP, 2000, 2003 server, and 98). The workstations used in the testing process are detailed in Table 2.
Performance testing
Various combinations of workstations from the test network were utilized to test the functionality and reliability of W.ND-BLAST. For performance tests, the job start times were automatically recorded as well as the time when the job finished. Duration of the job was calculated by subtracting the software recorded initiation time from the software recorded end time. Performance of W.ND-BLAST was tested by varying the number of input sequences, database size, and number of workstations. Two sizes of databases were used for testing the functionality of W.ND-BLAST, a 332 MB (total formatted size) protein (able to reside in physical memory) and 1.5 GB (total formatted size) nucleotide database (too large to reside in physical memory). The primary rational for testing the larger database was that the larger database exceeds the caching ability of the physical memory and requires the computer to access disk. In addition, larger databases are automatically segmented by formatdb. For this reason it is necessary to test both the database engine and BLAST engine for their abilities to handle large and segmented databases. The current version of W.ND-BLAST improves BLAST scalability only by distributing jobs across networks. As noted in the future work section at the end of the manuscript future versions of W.ND-BLAST will be further optimized to enhance performance on increasingly large databases using database partitioning (segmentation) and dynamic task allocations (Quality of Service partitioning). The smaller database was populated by protein sequences so BLASTx was executed. The larger database was populated by a subset of nucleotides from NCBI's nt database, therefore, BLASTn was executed. Current versions of NCBI nt and NCBI nr databases were tested to ensure that the software was able to handle much larger databases. Finally, all of the common BLAST programs were tested (BLASTp, BLASTx, BLASTn, and tBLASTx) and found to perform correctly.
Fault tolerance/recovery implementation
As with any distributed system, fault tolerance is always a key issue. W.ND-BLAST displays several layers of fault tolerance. The primary level of fault tolerance is achieved by allowing the master node to assign individual and distinct jobs to each child node. Because jobs (each individual sequences) are submitted independently by the master node to each child node, if a node fails, only the single sequence the failed node was processing at the time of failure is temporarily lost. The lost sequence is ultimately reallocated to a separate node by the master so that there will be no missing data. A log is kept at the master node of all of the queries in the sequence pool (jobs waiting to be executed) and their respective status. When a result is received at the master node, the validity of the result is checked, the sequence status is changed to done in the log, and the sequence is removed from the pool. The master node will only allow the sequence to remain in the working state for a user specified amount of time before returning the status to not done. This prevents jobs from being stalled on slow nodes toward the end of large jobs. When a failed child node is restarted on the network, it will broadcast a message to the master informing it of its return to ready status. The master then prompts the user to add or omit this node to its working nodes. If the master node goes down (upon rebooting of the master node or restart of the software), the job can be resumed at the point of failure and all data up to that point is maintained intact. All remaining query sequences in the sequence pool will then resume execution on the available children. The final layer of fault tolerance occurs at the end of a large BLAST job when the software performs a final check of data integrity and ensures that all the input sequences have generated a quality output. Any sequence that did not generate an output file is automatically subjected to a second BLAST.
Efficiency
Efficiency (calculated as a percentage) was used in order to determine the scalability of W.ND-BLAST. Efficiency was calculated by means of the following equation:
E = [σ / (n * x)]*100%.
where:
E = efficiency expressed as a percentage.
σ = Time to process a given number of sequences (job) on a single average class E (Table 2) workstation.
n = Number of workstations being evaluated for a given data point (e.g. if 9 workstations are being tested n = 9). This does not include the master node which does not perform queries.
x = Time to process job on n workstations.
Results and discussion
Performance of W.ND BLAST
Figure 6 provides data to illustrate the time savings realized with use of W.ND-BLAST on increasing numbers of workstations. This data was obtained by performing BLASTx for various numbers of sequences against a 332 MB (formatted size) database. The figure indicates a decrease in the amount of time required for BLAST of large input files as the number of workstations is increased. W.ND-BLAST exhibits an almost inverse linear relationship between the time taken to complete a given job and the number of workstations utilized for that job. This trend is very evident when considering Figure 7, which displays the efficiency of W.ND-BLAST as additional machines are added. Note that the number of workstations was limited to available workstations within our laboratory and we cannot project how network performance would affect this linear relationship when increasingly large numbers of workstations are utilized. Table 3 provides the raw data for the completion of test jobs as time in minutes.
When performing the test using 1500 sequences on one workstation the BLASTx took on average 662.68 min. Theoretically, when increasing the number of workstations by three the time should decrease by no more than 3 times (220.89 minutes). The actual time, derived from Table 3, was measured at 222.67 minutes, which correspond to a 99% efficiency rating using Equation 1. With 6 workstations the time should theoretically be 110.33 minutes. The actual average time achieved was 109.12 minutes, which gives an efficiency rating of better than 100%. This trend can also be seen when doubling the number of workstations from 6 to 12. It is likely that those efficiencies above 100% are likely attributed to the effect of adding worker nodes with slightly more resources (e.g. Class D) as part of the child node pool. This is because efficiency is calculated in relation to the single machine times, which are determined on our Class E nodes (Table 2).
Figure 7 exhibits the sequence sample size efficiency expressed as a percentage (Equation 1) when plotted for all benchmarks. These data demonstrate that W.ND-BLAST is more efficient when using a larger number of query sequences. The efficiencies for 15 and 17 workstations were not calculated because experiments were forced to be performed on workstations of much lower (Class F – I) or of much higher compute performance capabilities (Class A). However, the results using these additional workstations illustrate scalability and are provided in Table 3. Table 4 displays the results of BLAST scalability using even larger databases (1.5 GB) indicating that W.ND-BLAST can accommodate and make more efficient such intensive database queries. As a final operability test we obtained and tested W.ND-BLAST on 12-15-04 full and segmented versions of both NCBI nr and nt databases and found it was easily able to utilize even very large databases (data not shown).
Fault tolerance/recovery
W.ND-BLAST allowed workstations to fail, be turned off, or to be turned on during jobs. It easily adapted by redirecting failed queries, adding new workstations, or ignoring failed workstations. Even when the master node was shut off during testing and although the job did not progress, W.ND-BLAST was able to continue the job at the point of failure once the master node was rebooted.
Shortcomings in current version
W.ND-BLAST was designed as a user-friendly distributed software implementation of BLAST. For large databases such as current versions of nr or nt from NCBI there is a inherent decrease in the efficiency of the BLAST algorithm as described in the introduction. By distributing tasks among computers W.ND-BLAST still reduces the time it takes to perform searches on any size databases (in proportion to the number of worker nodes utilized) but does not enhance the efficiency of the BLAST algorithm itself. With tools such as MPI-BLAST [4], which utilizes a database segmentation scheme, queries against large databases are more efficient. As noted in the following section on future work the next evolution of W.ND-BLAST will include a database segmentation scheme that will dramatically improve overall efficiency on increasingly large databases.
In the current version of W.ND-BLAST there is the inability of the software to handle multiple projects simultaneously originating from a single master node. However on large networks multiple instances of the software can be running simultaneously if separate master processes are activated from different workstations and worker nodes are partitioned between jobs. This potentially allows more than one scientist to be utilizing the software on the same network.
W.ND-BLAST software is able to do a BLAST locally on the master node alone (local BLAST) but it is unable to BLAST on the master node, while performing distributed BLAST. This means that the number of machines for distributed BLAST is always n-1 if the master node is counted. It should be noted that all calculations based on time in this paper do not count the master node. In W.ND-BLAST the master node distributes tasks, coordinates, and compiles results.
Finally, there is a lack of ability in the current version to utilize more than one processor on multi-processor machines. In future versions this feature will be reincorporated as a feature.
Future work
There are several areas where W.ND-BLAST would benefit from future work. Firstly, W.ND-BLAST provides for sufficient fault tolerance when the master node fails, but implementing a shadow master would be the most efficient addition to the algorithm. The shadow master would act as a clone to the master node, and in the event of failure the shadow would continue the execution of the application. If the master node fails, the current implementation would not resume until the user arrived back at the master workstation to restart it. With a shadow master, the software would continue. Secondly, one of the most dramatic planned performance improvements in the W.ND-BLAST implementation will be the development of a database segmentation scheme similar to MPI-BLAST [4]. Considering the efficiency of W.ND-BLAST on smaller databases a segmentation scheme should provide an even greater increase in efficiency on large databases such as current versions of nr or nt. This increase in performance would be primarily based on the ability of a given node to maintain a designated segment (small piece) of large databases in its memory so that it does not exceed the physical memory capacity and allowing it to cache the database files in memory after the first search. This prevents excessive writing to disk to load additional portions of the database. Instead by segmenting the database a node will typically search all sequences against its own small piece of the database and return results to the master to be compiled. Thirdly, several levels of Quality of Service (QoS) will be implemented within the W.ND-BLAST system. The current implementation only allows the remote users (users at nodes) to end the application if necessary. Future efforts will allow the user at the worker node or at the master node to decrease the amount of CPU usage BLAST is allotted on each node on a node by node basis. QoS will also be implemented related to the availability or lack of availability of disk space on child nodes. In example, if the child node is unable to hold the database it can automatically be pointed or manually assigned to access a file share database.
Conclusion
BLAST is one of the most widely used applications in modern biology, including genomics, microbiology, and molecular biology in general. Its significance can be shown in the thousands of publications that refer to its use every year. In science, when a tool is this significant, the next logical step is to improve its functionality (ease of use). WND-BLAST provides this with its no hassle installation and intuitive user interfaces. Also, when a tool is this widely used the need arises for enhancing performance. WND-BLAST accomplishes this by allowing the user to distribute BLAST jobs from a single workstation to all available computing resources without the need for a server class machine. WND-BLAST provides the research community with a more comprehensive, fault tolerant, user-friendly, reliable, and time efficient toolkit to perform BLAST queries distributed across Windows based networks. W.ND-BLAST's output viewer and annotator also provide the user with a high-throughput method to analyze, process, and export BLAST results in a well-organized fashion.
Availability and Requirements
WND-BLAST can be downloaded free-of-charge from the web page . WND-BLAST requires Windows 2 k or higher with the .NET 1.1 framework installed. As mentioned previously, WND-BLAST was written primarily using Microsoft Visual Studio .NET 2003 using C#. The WND-BLAST software is provided 'as is' with no guarantee or warranty of any kind. The WND-BLAST software is available for all non-commercial use. Any other use of the software requires special permission from the primary author.
USDA disclaimer
Mention of trade names or commercial products in this publication is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the U.S. Department of Agriculture.
Authors' contributions
SD conceived of the project, devised the W.ND-BLAST algorithm, designed the functionality of each aspect of the software, and edited very early versions and final drafts of the manuscript; JZ encoded much the software and wrote the primary version of the manuscript; MO assisted with testing, finding bugs in the software, and editing early version of the manuscript; JR assisted with all coding of the software; PP assisted in writing the manuscript and testing the software.
Figures and Tables
Figure 1 Screenshot: Project Control Panel (PCP). The Project Control Panel (PCP) displays the user's options for W.ND BLAST. This example screenshot shows where the user is able to "Create New Project" by simply entering in a new project name. The "Create New Project" can also be changed to "Load Saved Project" where the user has the option to choose a previous project from the left text box.
Figure 2 Screenshot: Database and BLAST Engine (DBE). This example screenshot displays the Database Engineer, which is used for scanning for child nodes, creating custom databases and distributing existing databases. Within the same interface is the BLAST engine, which is used for executing inputting query sequences, choosing BLAST parameters, and executing BLAST jobs. Also available is the ability to control child nodes during BLAST jobs. Lastly, the user has the option of resuming a failed BLAST.
Figure 3 Screenshot: BLAST Output Viewer (BOV). This example screenshot of the BLAST Output Viewer displays some of the available columns information from an individual BLAST hit. This particular window is displaying the first five hits for a sequence named, "Contig_318." From here the user can move forward or backward through all BLAST output sequences. The user can also click on the "View Alignment" column to bring up the "Sequence Alignment Viewer" window.
Figure 4 Screenshot: Sequence Alignment Viewer. This screenshot of the Sequence Alignment Viewer gives an example the BLAST output for one alignment. The query sequence, midline and hit sequence are displayed in parallel and pair-wise equal can be seen in black
Figure 5 Screenshot: Annotation Engine. The Annotation Engine window shown here gives the user the options to export their data in several different formats. The user can choose from any of the available BLAST output result fields and any available type of output file and associated delimiters.
Figure 6 WND-BLAST Performance. This graph presents the performance of WND-BLAST observed when varying the number of workstations and input sequences. To collect this data, we utilized W.ND-BLAST to perform BLASTx against a 332 MB (formatted size) database with the varying number of input (query) sequences shown.
Figure 7 Workstation Efficiency Percentage with Best Fit trend lines. This graph shows the overall efficiency of performance of the data presented in Figure 1, as the number of machines is increased. Equation 1 from the text is used to generate this data. The linear best fit regressions for the data points are also provided.
Table 1 BLAST tool comparison
Program Name Reference Increase BLAST Performance Freely Available Distributed BLAST GUI Based GUI BLAST Results Viewer GUI Database Formatter OS Available1 Fault Recovery2
NCBI BLAST 1 N/A yes No No No No W,U Low
BLAST++ 5 yes yes No No No No W,U Low
WU-BLAST 3 yes yes No No No No U,O Low
DeCypher BLAST™ 6 yes No No yes No No O High
S-BLAST 7 yes yes yes yes No No W,U High
BeoBLAST 8 yes yes yes No No No U Medium
StarBLAST 9 No No No yes yes yes W Low
mpiBLAST 4 yes yes yes No No No W,U Low
Condor BLAST 10 yes yes yes No No No W,U Medium
Soap-HT-BLAST 11 yes yes yes yes No No W,U Medium
Paracel 12 yes No yes yes yes No U,O Unknown
W.ND-BLAST 13 yes yes yes yes yes yes W High
1 – W = Microsoft Windows, U = UNIX/LINUX, O = Other
2 – Based on ability to recover from node failure, master failure, and complete failure
This table provides a list of many of the most popular BLAST programs as well as their major features. The primary goal of most BLAST packages is to increase the performance of the BLAST algorithm. There seems to be a dichotomy between increased performance and user friendliness that has been encapsulated within W.ND BLAST. W = Windows based, U = Unix/Linux based, O = Other
Table 2 Workstation specifications
Class Processor Operating System Number of workstations RAM Harddisk Space
A Dual Xeon 3.0 Ghz Microsoft Windows XP Pro 1 6 GB 80 GB Ultra320 SCSI RAID0
B Pentium 4 Extreme 3.2 Ghz Microsoft Windows Server 2003 1 1 GB 140 GB SATA RAID0
C Pentium 4 2.8 Ghz Microsoft Windows XP Pro 2 2 GB 40 GB SATA
D Pentium 4 2.4 Ghz Microsoft Windows XP Pro 2 2 GB 80 GB SATA
E Pentium 4 2.4 Ghz Microsoft Windows XP Pro 12 1 GB 80 GB SATA
F Pentium 4 2.0 Ghz Microsoft Windows XP Pro 1 1 GB 40 GB
G Pentium 4 1.8 Ghz Microsoft Windows 2000 Advanced Server 1 512 MB 40 GB
H Pentium 4 1.4 Ghz Microsoft Windows 2000 Pro 1 128 MB 20 GB
I Pentium 3 M 1.0 Ghz Microsoft Window 2000 Pro 1 512 MB 20 GB
Different classes of workstations used in testing W.ND-BLAST are given. They are classified according to processor type and speed, OS, RAM, and hard disk space. The number of each workstation class in our lab is also given.
Table 3 Performance evaluation Number of Sequences
Number of Sequences
100 500 1000 1500
Number of Workstations W.ND-BLAST W.ND-BLAST W.ND-BLAST W.ND-BLAST
1 38.15 194.68 461.81 662.68
3 19.20 70.36 154.28 222.67
6 13.28 32.79 77.88 109.12
8 10.09 29.50 58.58 81.70
12 6.74 17.45 40.10 54.66
15 5.04 13.92 33.79 46.67
17 4.05 13.33 32.58 44.97
This table provides the time in minutes to perform the specified BLAST jobs on the number of sequences shown. The database size used for this analysis was 332 MB (formatted size) protein database. BLASTx was utilized to search high-throughput jobs ranging from 100–1500 sequences on various heterogenous workstations ranging from 1 to 17 nodes per job. At 15 and 17 nodes it should be noted that older machines were utilized.
Table 4 BLAST performance for large databases gives the average times (n = 3) for a W.ND-BLAST project to complete when querying 50 sequences on Class E machines against a 1.5 GB nucleotide database (too large to fit in physical memory). The times, shown in minutes and seconds, show a relatively large decrease in execution time when the number of machines is increased from one to seven nodes.
Sequence Number Database Size Number of Machines Total Time(mm:ss)
50 1.59 GB 1 63:57
50 1.59 GB 7 12:14
==== Refs
Altschul S Gish W Miller W Myers E Lipman DJ Basic Local Alignment Search Tool Journal of Molecular Biology 1990 215 403 410 2231712 10.1006/jmbi.1990.9999
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 402 9254694 10.1093/nar/25.17.3389
Gish W WU-BLAST 1996
Darling A Carey L Feng W The Design, Implementation, and Evaluation of mpiBLAST ClusterWorld 2003 conference 2003
BLAST++
DeCypherBLAST™
Dowd SE Rodriguez JR Khurana V Sobolewski M Soorinarayanan S S-BLAST: Federated BLAST Using Sorcer Publication pending 2004
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
StarBLAST
Condor BLAST
Soap-HT-BLAST
Paracel
W.ND-BLAST (Current Paper)
NCBI BLAST
| 15819992 | PMC1087835 | CC BY | 2021-01-04 16:02:50 | no | BMC Bioinformatics. 2005 Apr 8; 6:93 | utf-8 | BMC Bioinformatics | 2,005 | 10.1186/1471-2105-6-93 | oa_comm |
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-951582630510.1186/1471-2105-6-95SoftwareStatistical Viewer: a tool to upload and integrate linkage and association data as plots displayed within the Ensembl genome browser Stenger Judith E [email protected] Hong [email protected] Carol [email protected] Elizabeth R [email protected] Margaret [email protected] Pascal J [email protected] Jeffery M [email protected] The Duke Center for Human Genetics, Duke University Medical Center, Durham, North Carolina 27710, USA2 Department of Medicine, Duke University Medical Center, Durham, NC 27710, USA2005 12 4 2005 6 95 95 15 12 2004 12 4 2005 Copyright © 2005 Stenger 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
To facilitate efficient selection and the prioritization of candidate complex disease susceptibility genes for association analysis, increasingly comprehensive annotation tools are essential to integrate, visualize and analyze vast quantities of disparate data generated by genomic screens, public human genome sequence annotation and ancillary biological databases. We have developed a plug-in package for Ensembl called "Statistical Viewer" that facilitates the analysis of genomic features and annotation in the regions of interest defined by linkage analysis.
Results
Statistical Viewer is an add-on package to the open-source Ensembl Genome Browser and Annotation System that displays disease study-specific linkage and/or association data as 2 dimensional plots in new panels in the context of Ensembl's Contig View and Cyto View pages. An enhanced upload server facilitates the upload of statistical data, as well as additional feature annotation to be displayed in DAS tracts, in the form of Excel Files. The Statistical View panel, drawn directly under the ideogram, illustrates lod score values for markers from a study of interest that are plotted against their position in base pairs. A module called "Get Map" easily converts the genetic locations of markers to genomic coordinates. The graph is placed under the corresponding ideogram features a synchronized vertical sliding selection box that is seamlessly integrated into Ensembl's Contig- and Cyto- View pages to choose the region to be displayed in Ensembl's "Overview" and "Detailed View" panels. To resolve Association and Fine mapping data plots, a "Detailed Statistic View" plot corresponding to the "Detailed View" may be displayed underneath.
Conclusion
Features mapping to regions of linkage are accentuated when Statistic View is used in conjunction with the Distributed Annotation System (DAS) to display supplemental laboratory information such as differentially expressed disease genes in private data tracks. Statistic View is a novel and powerful visual feature that enhances Ensembl's utility as valuable resource for integrative genomic-based approaches to the identification of candidate disease susceptibility genes. At present there are no other tools that provide for the visualization of 2-dimensional plots of quantitative data scores against genomic coordinates in the context of a primary public genome annotation browser.
==== Body
Background
The search for genes contributing to complex human diseases
The availability of the complete DNA sequence of the human genome, along with advances in gene expression, proteomics, metabolomics technology and bioinformatics databases, presents new opportunities for integrative approaches to identify candidate susceptibility genes for complex human diseases. Complex diseases, which include such diverse illnesses as Alzheimer disease, Parkinson disease, cardiovascular disease and asthma, account for the majority of chronic illnesses that plague our society today. These non-Mendelian diseases are attributable to inherited polymorphisms in perhaps several risk-associated or modifier genes that are triggered by exposure to environmental agent(s). Because of the multitude of factors ultimately contributing to the disease phenotype and the numerous confounding variables presented in studying human diseases, isolating the genetic components that confer an underlying predisposition to a complex disease is an inherently daunting undertaking.
The importance of data integration
Thus, to improve the odds of successfully identifying complex disease susceptibility genes, several diverse approaches, each of which must capitalize on cutting-edge technical, informatics and analysis, should be exploited. The integration of disparate biological, statistical and clinical databases, both public and private, into whole-genome annotation are of paramount importance to comprehend and efficiently interpret the vast quantities of DNA sequence data, gene expression data, proteomics and other "-omics" data. As genotyping is a costly endeavor, ever-more effective computational tools are needed to readily access, organize and comprehend the massive quantities of data generated to identify and prioritize candidate genes for genotyping. Only when a disease causing genetic mutation is confirmed can the underlying molecular mechanisms of complex diseases be unraveled so that tests, prevention, new knowledge-based therapeutic approaches can eventually be devised.
Great strides have been made towards the federation of bioinformatics databases as a result of concerted efforts over the last few years by leading bioinformaticians to develop controlled vocabularies [1,2], common platforms and tools for integration [3]. As a result, despite the exponential growth of bioinformatics data, the number of individual web-based resources that genetic researchers have to navigate has been substantially reduced from a multitude of disparate web sites on single chromosomes and individual physical and genetic maps to essentially three major on-line resources that facilitate access, analysis and retrieval of data from the recently completed human genome.
The visual presentation of the immense quantities of incongruent data types, however, presents challenges in itself. Effective integrated informatics tools must be capable of representing essential data of ever increasing complexity in a format that is both comprehensive and easily synthesized by the human brain [4]. Towards this end, three well-annotated web-based public genome browsers with improved interfaces have been developed and are continually evolving. These are: 1) the NCBI Entrez map viewer [5,6], 2) the EMBL-EBI / Sanger Institute collaborative Ensembl project [7,8], and 3) the University of California at Santa Cruz's Golden Path Genome Browser [9,10]. All are easily queried and capable of visually presenting overviews of the large regions of the genome while allowing the user to zoom in on an area of interest revealing detailed information on the numerous features mapped within. Each of these has become invaluable tools in the arsenal of genetic researchers.
Genomic convergence
We and others have embraced a multi-faceted integrated approach to identify and prioritize candidate genes for complex human diseases that we call "genomic convergence" [11]. This approach combines the list of genes obtained two or more distinctly different methods (e.g. gene expression and linkage analysis) to obtain a list of top candidate genes. Theoretically, including genes identified by other independent, yet biologically relevant, lines of evidence could increase the sensitivity of the approach as well as the specificity when the genes identified by several approaches are giving higher priority.
Rationale
To increase the efficiency and accuracy of the identification and prioritization of candidate genes for genotyping, we needed to facilitate the identification and extraction of genomic features of interest that are within linkage regions as well as fully exploit the public databases and the human genome assembly and annotation projects. A strategy relying on the assignment somewhat arbitrary "fitness" thresholds to reduce the number of candidate genes for follow-up analysis poses the risk of excluding the causative gene from the list of candidates if the thresholds are too stringent. For this reason a tool capable of displaying quantitative and positional data within a single integrated browser view that would facilitate the synthesis and interpretation of disparate data types is superior to the more simplistic approach using the intersection of sets of genes so that the thresholds are set empirically.
Customizing Ensembl to incorporate two-dimensional linkage and association data plots
For reasons discussed later in this paper, we opted to use a local implementation of Ensembl as the basis of our internal bioinformatics infrastructure. To fully meet our needs for integrating linkage and association data we developed software to customize Ensembl as an analytical tool for genomic convergence approaches to identify potential disease-susceptibility genes for follow-up analysis. For this purpose we have developed software modules that add functionality to the Ensembl genome annotation systems so that the browser will display quantitative data points plotted against chromosome position (e.g. statistical results, from genomic screens, fine mapping and association studies and expression levels), which are seamlessly integrated into the Contig View and Cyto View web pages. The new panels fully support the functionality of the Ensembl system so genome regions corresponding selected by the user within the Statistic View will be displayed in the Overview and Detailed View panels and additional information on the statistical data-points may be displayed in pop-up views with hyper-links to individual feature information pages. The Statistical Viewer package includes software to facilitate the upload, query, storage, integration, display, analysis, and retrieval of private quantitative data into a public open-source genome browser so that all public annotation, DAS sources and links can be fully exploited.
We have created a software package called "Statistical View" that includes an enhanced upload server that facilitates the upload, query, storage, integration, display, analysis, and retrieval of private quantitative data into a public genome browser to help geneticists make connection between the disease phenotype and the genetic features that are associated with risk on a genomic scale. From there an understanding of the molecular basis of disease can lead to testing, prevention, and perhaps ultimately, to pharmaceutical or alternative means of intervention so that the hopes of translational medicine promised by the completion of the human genome sequencing project will eventually be realized.
Implementation
Mapping genetic marker locations as genomic positions in the human genome assembly
Linkage data is traditionally visually displayed as a graph in which Lod scores are plotted along the ordinate against the genetic location of the markers that were used in the genetic screen and the subsequent analysis. The length of the abscissa represents the length of the chromosome in centimorgans (cM), a function of the recombination frequency between two loci. Recombination frequency is influenced by factors such as regional genetic content and gender so that one can only make a rough approximation between in the average number of base pairs that are in a single centimorgan map unit. Therefore, to incorporate a visual representation of statistical results into the Ensembl genome browser in a meaningful way, it is imperative that the abscissa be expressed in base pairs so that the position of markers along the abscissa correctly align with, and strictly correspond to, the horizontal illustration of the ideogram that is displayed immediately above the linkage study graph in the Ensembl ContigView. Towards this end we have developed a tool "GetMap" that uses linear interpolation relative to deCODE Genetics' published map [12] to approximate the physical chromosomal coordinates denoted in bps of markers that were a) not mapped by the deCODE group and b) we have insufficient information for successful ePCR. The algorithm first uses a binary search (the "divide-and-conquer" paradigm) to find and extract known marker coordinates from a database when available. Otherwise the linear interpolation algorithm is activated, using the closest markers flanking markers with known coordinates to approximate the position as accurately as possible. The algorithm is based on that described in by Kong et al. [12].
To convert the map units from centimorgans into their chromosome sequence coordinates in base pairs, it is necessary to map the markers used as probes to the physical sequence of the human genome. The most recent version of the human genome assembly (currently NCBI build 35) can be downloaded from the UCSC genome site [10] and the most recent version of the Ensembl annotation system [8].
The first step in converting the genetic location of a marker into its position in base pairs is the creation of a database of unique reliable and valid markers (typically microsatellites) based on the NCBI UniSTS database [13,14], an integrated non-redundant database of markers (sequence tagged sites, STSs [15]), as a starting point. UniSTS integrates mapping information gathered from various resources primary sources, and is the source of marker probe sequence information. The markers, their aliases, sequences and available locations on genetic maps are retrieved via the NCBI ftp server [16]. Once the most recent version of the human genome assembly and UniSTS has been uploaded onto our local servers we re-map the positions of STSs by using e-PCR [17-19], or BLAT [20] as a next resort. In a preprocessing step, markers that may have erroneous map positions and/or show inconsistent ordering when compared to other maps are flagged, removed from our verified database and saved elsewhere for reference. The confirmed genomic position and other information, including its genetic location(s), is then used to populate the statistical results table in the database illustrated Figure 3. Then, to facilitate the interconversion of map locations from pre-existing genetic maps and analysis into the human genome nucleotide position, we have developed a web-based tool called "Get Map" for uploading a file containing marker locations, as well as their position as provided by the user. This tool also exists in a stand-alone version for processing batch files and is described in the Get Map documentation file provided as supplementary material.
Software for viewing statistical data as 2-dimensional plots within the context of Ensembl
The drawing program consists of four basic Bioperl modules that build on the Ensembl open source software and genome annotation system to display a linkage plot of statistical results, that are easily uploaded by laboratory personnel using the upload server we devised, in the context of the annotated human genome sequence. These modules add an additional display panel, "Statistic View" that appears automatically when the Contig View (see Figure 1) and the Detailed View in CytoView (see Figure 2) pages are opened. However, if no statistical data for any studies has been uploaded into the mySQL db for a chromosome of interest no plots are drawn, but a compressed panel appears displaying text indicating that "there is no statistical data pertaining to any study available for this chromosome" (not shown). Like other Ensembl panels, to save space, this panel can be compressed when not needed. Once the data is uploaded into a local Distributed Annotation System (DAS) [21] server, researchers can use a pull-down menu to select the plot for a particular study of interest that is then drawn in a panel placed between ideogram of the chromosome and the Overview in both Cyto View and Contig View.
Description of BioPerl modules for embedding "Statistic View" panels in Ensembl
Using the Perl programming language we developed BioPerl modules (reviewed by Stajich et al. [22] that draw a linkage (or other) plot using the data uploaded in the DAS server for the particular study, which the researcher must provides the name of as a required field. The data flow for the Statistic View module is illustrated in figure 4. The source code is provided as supplemental material. The four basic BioPerl modules are: 1) Bio::EnsEMBL::Linkage.pm, 2) Bio::EnsEMBL::DBSQL::LinkageAdaptor.pm, 3) Bio::EnsEMBL::GlyphSet::lodplot.pm and 4) WebUserConfig::chrplot.pm and Bio::EnsEMBL::GlyphSet::FineLODplot provides for a detailed statistical plot, described below, that is drawn beneath the Detailed View panel in Ensembl's Cyto- and Contig- View pages. The source code for each of these is provided as rich text files (see Additional file 1 through 5, respectively). Also documentation and executables are provided within a compressed archive file (Additional file 6).
Briefly, the Linkage.pm module features a constructor to create the linkage object, essentially a table that encapsulates the record for a single linkage point, allowing for linkage data to be added populating the database. The database table, Linkage_Table, contains up to eleven fields shown in figure 3A. This information can be viewed in the Ensembl Contig- and Cyto- View browser pages within a pop-up box when moving the cursor over ("mousing-over") the point in the plot as shown in figure1 below the Detailed View)
The GlyphSet::lodplot.pm module is next implemented. This module contains all the information for drawing the graph of the statistical data. This module served as the basis for the FineLODplot.pm that displays a graph of statistical data below the Detail View Panel that corresponds to the same range (from the slice of the chromosome selected when the user slides the red box in either the ideogram or the Statistic View panel) of the Detailed View Panel. The WebUserConfig::chrplot.pm module is then added to configure the lod plot so that it is displayed in the Statistic View panel in CytoView and ContigView.
The LinkageAdaptor object module, LinkageAdaptor.pm, provides the functionality for accessing linkage data from the DAS database. Linkage adaptor creates a "slice" object, corresponding to the sliding red box that the view uses to select a region. The slice object defines the region from which linkage data will be retrieved and returns the records from the linkage objects that are mapped to the chromosome between the coordinates within the boundary of the red box that are displayed in the detailed view below. Finally the plot image was scored into the Ensembl Contig View and Cyto View by modifying the Contig- and Cyto- view code so that the linkage plot is displayed into the Ensembl Contig View.
Figure 1 is a screen shot of the Ensembl Contig View web browser running off of our local server that illustrates the Statistic View panel with a plot comparing lod and hetlod scores against their corresponding genomic location. Figure 2 similarly illustrates the incorporation of Statistic View into Ensembl's CytoView. Note how the graph in Figure 1 features data points interconnected by dotted lines while that in Figure 2 shows solid lines. The database table provides a field called link_type that allows a statistical analyst to specify the display to quickly allow one type of analysis, such as single point, from another, such as multi-point, when a convention is adhered to.
Although Statistic View is an invaluable tool for surveying chromosomes for interesting genomic features in linkage peaks derived from a genomic screen, it is of limited use in fine mapping and association studies due to insufficient resolution at the chromosomal level to see the desired features as the density of markers increases. Thus, recognizing the utility of having the ability to display detailed information in a region selected for detailed association studies, we have developed code (provided within Additional file 5) to plot and display association data in a separate panel, below the Detailed View in Contig- and Cyto- View. Essentially, this panel, that we refer to as "Detailed Statistic View" is an enlargement of a region selected from either "Statistic View" or the Overview panel that will directly corresponds with the Detailed View outlined in red (See figures 1 and 2).
Results
Here we describe the functionality of a program package called "Statistical Viewer" that was written for the purpose of integrating statistical genetic data with human genome sequence annotation. The name "Statistical Viewer" refers to the name of the software package while "Statistic View" refers to the display panel that is labeled in Ensembl's "Contig View and Cyto View" pages.
Usage
The first step in adding statistical data to Ensembl is formatting the data for upload. As the abscissa must correspond to genomic coordinates for integration into the human genome assembly, the position of the linkage point (for the microsatellite or other marker) needs to be defined by the name of chromosome, and chromosome start and end position in base pairs.
Simplified data entry via an improved web-based upload service
To facilitate the upload of statistical data as well as other types of private data, we developed a customized upload server that allows members of a group with permissions to upload their own data Users may add new data to be plotted into "Statistical View" or data for annotation as DAS tracts. Our improved DAS upload server allows users to append data into previously instantiated data tracks. The web-based interface is also user-friendly so that formatting difficulties and failures and errors that typically ensue when cutting and pasting tab-delimited text into the Ensembl web form, are avoided.
Upload file format requirements
The data must be in the form of a simple two-dimensional table containing attributes (columns) and tuples (rows). The upload serves accepts either tab-delimited text files or MS excel spreadsheet files as input using a browse feature to specify the directory. When using an Excel file, the top worksheet page must contain all the data to be uploaded and be devoid of any merged fields. We have designed the upload server to require a minimum of eight essential fields that serve as column headings in a table, although additional attributes such as the name of the researcher or technician entering the data, the method, or date are permitted and encouraged.
These eight required fields (attributes) are:
1. Study: The term defines the data set and is the title of the study. The field is a string of characters giving the name of a disease, project or a sub-study. One study can have data across multiple chromosomes or in several analysis groups. Data from multiple studies can be combined into a single spreadsheet if the records are listed sequentially in tuples (rows).
2. Analysis: This field enables representation of data sub-categories from stratification within a study. The character string can be different statistical analysis methods, or the different populations, etc. Again, several data categories can be included into the same spreadsheet. The fields will be used to provide the key or legend for the points plotted on the Statistic View graph. The units should be included in parenthesis.
3. Link_point: This field is typically the name of the marker for linkage study. The value will be represented as a data point on the Statistic View graph that will display the name of the marker in a pop-up window when the point is "moused-over".
4. Score: The statistical score for the Link_point. The units may be indicated here as well the score for inclusion in pop-up windows.
5. Chr_name: All human chromosome names (1 through 22 as well as X, Y) are acceptable character values.
6. Chr_start: The chromosome start location of the Link_point in base pairs.
7. Chr_end: The chromosome end location of the Link_point also in base pairs. This position may be identical to the chr_start value in the case of a SNP.
8. Link_type: This attribute specifies the type of line used for connecting the Link_points in the plot: for example, "dot" denotes a "dotted line (- - - - -)" that we use as a convention two-point analysis. "Line" is "solid line (____) " and typically indicates a multi-point analysis. By placing the term "point" in this field, Statistical View will draw a scatter plot without lines. If a value is not supplied for this field a solid line, the default, will connect the link points.
Viewing integrated statical data in Ensembl
The StatisticalView panel depicts a graph plotting the linkage, or association, statistics along the ordinate and the length of the chromosome, in megabases (Mbs), along the abscissa. The length of the abscissa diametrically corresponds to the length of the ideogram that is illustrated directly above. The program generates this panel, which is capable of displaying two-dimensional graphs plotting data for pertaining to a study of interest. Refreshing the page following selection of another study or analysis method can bring up different plots. This plot is seamlessly integrated into the Cyto View and the Contig View pages. Like other panels in these web pages, the Statistic View panel features a selection box, a movable red rectangle with adjustable width, to highlight the boundaries of a region of interest in the linkage plot. The boundaries of the region selected in Statistic View directly correspond to the selection box in the ideogram and the physical map position that is denoted in bps in the Detailed View. The selection box also correctly corresponds with the enlarged box that is dynamically drawn in the overview. As with the corresponding selection box in Ensembl's other display panels in Contig- and Cyto- View, the width of the selection box in Statistic View can be altered with a mouse, or a similar input devise, to select either a larger or a narrower region of interest to dynamically alter the other display panels. Likewise, changing bp coordinates in the Detailed View, or altering the size or location of any of the sliding selection boxes in the other panels, has the appropriate effect on the selection box in the Statistic View panel.
Discussion
Following the much-anticipated first release of the draft sequence of the human genome by the international consortium [23] and the Celera Corporation [24] in February 2001, human geneticists were eager to apply this resource data to use to map and identify disease susceptibility genes. Even in its incomplete and unverified state, the data represented a tremendously powerful resource to help resolve the inconsistencies that confounded the use of independently derived genetic and radiation hybrid maps. Because of the computational complexity of dealing with large and incomplete human pedigrees [25], the production of these maps was a significant accomplishment. The deCODE Corporation immediately incorporated the draft sequence data in constructing a new meiotic map [12] that represented a significant improvement over the Marshfield [26] and Genethon maps [27]. It was also immediately obvious that a tremendous amount of work remains to translate this data into knowledge that will eventually improve the overall health of the public and that the processes of analyzing and interpreting this data presents many challenges in itself.
In spite of the exponential growth of biomedical data, the task of mining data is less daunting than it was just a few years ago. Via the World Wide Web (WWW) geneticists have at their disposal three distinct, high-quality, well-annotated genome repositories that provide free access to the most recent genome assemblies for humans as well as an increasing, diverse assortment of model organisms. Each of these public genome browsers, NCBI, UCSC and EBI/Sanger's Ensembl, employ their own annotation pipeline but should contain the same nucleotide sequence from the latest, or at least the second most recent, release of the NCBI assembly. As is the case with the NCBI and UCSC genome browsers, the public Ensembl site contains genomic sequences and a plethora of useful features extending well beyond known and predicted genes, microsatellite markers, and SNPs that are linked to their corresponding records hosted at their respective primary sources (see [28] for a review of current molecular biology databases).
We have chosen Ensembl as the system to underlie Statistic View for several reasons: 1) the horizontal presentation of the genome annotation makes it amenable to displaying a linkage map, 2) Ensembl has stably incorporated DAS for displaying customized data from sources outside of the Ensembl annotation pipeline mapped to the genome as tracts for several years, 3) the developers have intended from the start that the project would be open source and thus have taken great care in documenting its source code, and 4) Ensembl's EnsMart genome data retrieval tool is a very sophisticated and flexible data mining tool containing extensive filters and several good output options. Thus for the purpose of integrating linkage data, as well as other types of internal data, we believe Ensembl currently provides the best architecture to serve as the basis of an data integration infrastructure for use by a genetics laboratory in an academic setting.
A local implementation of the Ensembl genome annotation databases and software system is ideal for integrating third party annotation features with public human genome sequence annotation and serves as the foundation of our bioinformatics infrastructure to assist in the identification and prioritization of candidate genes in disease fine mapping and association studies. The DAS system has enabled us to display the location of markers used in a particular study as features in tracks displayed in Contig- and Cyto-View. Additionally we have mapped other features such as the location of differentially expressed genes from experiments using both microarray [29] and SAGE technology [11]. These studies employed Statistic View and illustrate the utility of having such a tool for rapidly identifying and visualizing genomic features that are in regions of linkage.
We have found that the benefits of locally maintaining a mirror of the public human genome sequence along with the Ensembl genome browser software and a DAS server exceed the costs. It was far more costly to maintain individual databases with different formats across different projects. This integrated system allows users to manually curate genomic, computationally derived, statistical, genetic, and experimental (e.g. gene expression) results for many projects. Additionally, sensitive data can be password protected.
Although the ability to map the location of microsatellite or SNP markers used in genomic screens and genotyping into DAS tracts in the Ensembl overview has proved to be very useful in genomic convergence or integrative genomic strategies thus far, Statistic View will continue to enable us to efficiently screen regions of the genome for additional features that may be indicative of a gene warranting further investigation. Towards this end, we are conducting research to identifier better predictors of successful outcome. We have mapped to the genome sequence DNA motifs that we predict are "at-risk" for genome instability such as full length, highly identical, closely spaced inverted repeats such as Alu pairs [30,31] and long simple tandem repeats (Stenger, unpublished) in coding sequences (see detailed view in Figures 1 and 2). Displaying these features in DAS tracts in conjunction with linkage data has enabled us to hone in on a potential candidate gene even when the function of gene is unknown. Although the number of genes of unknown function is dwindling with better recognition of pseudogenes, they still represent a large enough portion of genome that it is prudent to not exclude from follow-up sequencing and association studies. However, genes of unknown function frequently are overlooked since such genes are excluded from strategies based on biological plausibility and may not be represented on microarrays.
In addition, we are in the process of mapping trans-acting transcription factors to the genome. Aberrantly expressed genes may not map to linkage regions, but it is likely that co-expressed genes are regulated by a common cis-acting regulatory element. The identification of proteins that may bind to these regulatory motifs may further inform our search for candidate genes in regions displayed in Statistic View. We have also been using in silico subtractive hybridization methods to identify genes expressed uniquely in tissues that exhibit pathology in the diseases that are under investigation (e.g. a DAS tract displaying the location of genes specific to the substantia nigra may help identify candidate genes for Parkinson Disease) and have mapped these to the genome as DAS tracks.
It is worthwhile to mention that the upload of data into the DAS database using Ensembl's upload server can be problematic. The upload server provided on the public server had a number of shortcomings: 1) it is often difficult to properly format the data as text so that the desired output is generated, 2) a full email address is required for the user ID at login, and 3) each login results in data put into a separate track so that individual records could not easily be added to an existing data track. To overcome this, the user who initialized the track was required to delete the original data file and append the new data to a file and reload the entire data set. These limitations detracted from the usefulness of our private Ensembl-DAS system because data entry became a bottleneck in the analysis project with bioinformaticians serving as gatekeepers. We extended the capabilities of the upload server so that researchers with appropriate permissions could easily add individual records to the DAS database without having to delete and recreate a new data set and its resulting tract displayed in the detailed view panel. By improving the data upload procedure, laboratory personnel who generate data can enter and maintain their data with minimal training.
To facilitate the upload of data, we developed a utility that thus far represents significant improvements over the upload server provided with the Ensembl package. Our tool works well for uploading data with base pair coordinates that is to be displayed in tracks in Contig- and Cyto- View's Detailed View panel. We have found that to have data entry proceed at the same rate as data generation, it is best to provide laboratory researchers with the tools to add their own annotation when high throughput genotyping methods are being used, although some labs may prefer to have a gatekeeper to manually curate data at the time of entry to ensure that data integrity is preserved. Recognizing the need to minimize errors, which are inevitably propagated, and faced with limited human resources to enter and curate data we in the process of developing greater functionality to our upload server by automating conversion to the proper physical location so that potential mapping errors are avoided. This software is provided in its current level of development as Additional files 7 and 8and their respective user manuals are provided in Additional files 9 and 10.
Conclusion
We have developed software to enhance the Ensembl open source software package as a private laboratory bioinformatics infrastructure to assist in the identification of candidate complex human disease susceptibility genes. We have improved the upload server thereby empowering laboratory personnel to add project specific data to the local DAS database for display in the context of the human genome using the Ensembl genome browser's contig view. By creating an additional panel in Ensembl's Contig View and Cyto View, called Statistic View, we are able to display statistical results from gene mapping experiments in the context of human genome sequence annotation. Statistic View displays a plot of linkage and association statistics directly above the Overview. Statistic View is fully and seamlessly integrated into the Ensembl genome browser. The user can navigate the chromosome by mousing-over a selection box outlined in red that directly corresponds to the one drawn by the Ensembl software in the ideogram of the chromosome. This capability facilitates the selection of regions of the chromosome in linkage disequilibrium for easily visualization of features mapped therein that are displayed in data tracts below in the Overview. This capability allows rapid screening of regions of interest to a particular study to identify genes that deserve further screening.
Availability and requirements
The source code is provided as supplementary material for this publication and will also be available from the Duke Center for Human Genetics public web pages [32] as well as through the public Ensembl site [8].
Setting up a local implementation of Ensembl
To use this software the Ensembl Genome Browser Software Annotation System must be installed and running locally. The Statistical View package runs on version 26.1, which can be accessed through the Download Ensembl Wiki web URL [33]. Genome files and other database files can be access through the Ensembl ftp server [34]. Ensembl requires a server with a UNIX or Linux type of operating system (e.g. OS X, SGI's IRIX and Sun's Solaris).
All software and the full complement of mySQL genome sequences and databases currently occupies 150 gigabytes of storage space and requires just as much swap space, but may not be needed depending on your requirements as specified in the installation pdf file [35].
We maintain a current local implementation of the Ensembl open source software system, including the human genome sequence assembly and related databases running on a SunFire 12 K reference server in conjunction with a separate server, a Sun Blade 2000 acting as an annotation server to store and overlay public genome data with private laboratory data for integration using the Distributed Annotation System.
List of abbreviations
BLAT – BLAST-Like Alignment Tool
cM – centimorgan, a single map unit
db – database
DAS – Distributed Annotation System
DNA – deoxyribonucleic acid
EMBLI – European Molecular Biology Laboratory
EBI – European Bioinformatics Institute
Mb – megabase, one million base pairs
NCBI – National Center for Biotechnology Information
SAGE – Serial Analysis of Gene Expression
SNP – single nucleotide polymorphism
SQL – structured query language
STS – sequence tagged sites (markers)
Authors' contributions
J E S – scientific director of CHG bioinformatics core, headed bioinformatics database integration project, contributed to concept, implementation ideas, helped formulate requirements and provided suggestions for improvements. Secured funding for hardware.
H X – senior programmer responsible for implementation.
C H – member of database integration and post-processing groups, contributed to concept, and implementation ideas. Helped design database tables, helped formulate requirements and provided suggestions for improvements.
E R H – Principal Investigator for CEGS bioinformatics Component, headed post-processing project, contributed to concept, implementation ideas and requirements
M P-V – Director of CHG, provided funding, human resources and the impetus for the project
P J G-C – Chairman of the Duke Department of Medicine, provided funding and supported project development.
J M V – senior author, provided cause for action, contributed ideas and requirements and funding
Supplementary Material
Additional File 1
The source code for Bio::EnsEMBL::DBSQL::LinkageAdaptor
Click here for file
Additional File 6
The executable BioPerl modules and documentation can be decompressed with either the GNU (Free Software Foundation) unzip command [36] in UNIX or the WinZip Windows archive utility (WinZip Computing, Inc.) [37] in an MS Windows OS. Please see the respective web sites that are referenced for additional details on acquiring and using these compression utilities, as well as the UNIX man pages for the tar command: i. Readme A text file the provides a description of modules and the path to the code in the Ensembl Server Root directory ii. Support.pm (Format: Perl Module) Modified source code for the ContigView::Support package that includes the Duke Center for Human Genetics extension to add the LOD score plot panel iii. Chrplot.pm (Format: Perl Module) The WebUserConfig::chrplot BioPerl module iv. Contigview (Format: text file) Modified source code for the configuration based version of the Ensembl Contig view package including the necessary additions to incorporate the LOD score or other statistical data plot v. Cytoview (Format: text file) Source code for the Ensembl Cyto view package including the modifications to incorporate the LOD score plot vi. DBAdaptor.pm (Format: Perl Module) The modified source code for the Bio::EnsEMBL::DBSQL::DBAdaptor module that includes the Duke Center for Human Genetics extension to add the LOD score plot panel into the genome browser vii. Linkage.sql (Format: Structured Query Language) This file specifies the table structure for the linkage object viii. Linkage.pm (Format: Perl Module) Bio::EnsEMBL::Linkage Ensembl BioPerl module ix. LinkageAdaptor.pm (Format: Perl Module) Bio::EnsEMBL::DBSQL::LinkageAdaptor BioPerl module This Module includes the POD documentation – the main documents preceding the code x. Linkageview (Format: BioPerl package) This file provides the package to display the chromosome linkage plot and the information for each linkage point xi. Linkageview.pm (Format: Perl Module) The EnsEMBL::Web::UserConfig::linkageview module xii. Lodplot.pm (Format: Perl Module) The Bio::EnsEMBL::GlyphSet::LodPlot module xiii. HTML.pm (Format: Perl Module) This file is the modified source code for the ContigView::HTML package that includes the Duke Center for Human Genetics extension to add the LOD score plot panel
Click here for file
Additional File 5
The source code for Bio::EnsEMBL::GlyphSet::FineLODplot BioPerl module
Click here for file
Additional File 7
The source code for the CHG enhanced upload service for integrating local data as features for display in the Ensembl genome browser as DAS tracks
Click here for file
Additional File 8
The source code for the CHG enhanced upload service for integrating linkage or other statistical data for display in plots (graphs) as a panel within the Ensembl genome browser Contig- and Cyto- View pages.
Click here for file
Additional File 9
User manual for uploading Linkage or other statistical data to be plotted in a display panel called "Statistic View" within the Contig- and Cyto- view genome Browser pages of a local implementation of Ensembl.
Click here for file
Additional File 10
The manual for using the Duke CHG enhanced upload server web interface to import data into MySQL database so that the features can be displayed as DAS-tracks showing annotation for experimental data within the context of the genome sequence assembly on a local implementation of Ensembl.
Click here for file
Additional File 2
The source code for Bio::EnsEMBL::Linkage
Click here for file
Additional File 3
The source code for Bio::EnsEMBL::GlyphSet::LodPlot
Click here for file
Additional File 4
The source code for WebUserConfig::chrplot BioPerl module
Click here for file
Additional File 11
Manual for using the Get Map tool for the interconversion of genomic and genetic positions for markers. The algorithm is also described.
Click here for file
Acknowledgements
We are grateful to Jason Stajich for pioneering this endeavor at the DCHG. We are indebted to the Ensembl team, especially Tony Cox at Sanger for providing guidance in setting up our local DAS server and Ewan Birney for his provocative discussions, encouragement, and willingness to make Statistic View available through the public site. We would like to thank Richard Cornwell for assistance in developing some of the code. We also acknowledge CHG fellows, students and technicians of the CHG for their helpful suggestions. Also, none of this would have been possible if not for dedicated researchers of the International Genome Consortium who collectively accomplished the remarkable feat of sequencing the human genome in its entirety, as well as those who tirelessly worked to continuously improve public access to biological databases. This work was supported in part by the Alzheimer's Association grant "Identification of Novel Age-at-Onset Genes on Chromosome 10 in Alzheimer Disease" as well as the following NIH grants: NHLBI grant P01 HL73042-01 (P.J.G.-C., E.R.H., H.X. M.P.-V., J.E.S., J.M.V), NINDS grants R01 AG021547 (M.P.-V), P01 NS26630-15 (M.P.-V), R01 NS36768-07 (M.P.-V), NIA grant R01AG19757-01, P50 NS39764-04 and its supplement P50 NS39764-04S1 (J.M.V.), NIEHS ES11375 (H.X.) J.E.S. is also supported in partby NIEHS K-22 grant ES00372.
Figures and Tables
Figure 1 Contig View featuring Statistic View. Shown is a screen shot depicting a local implementation of Ensembl's contig view page hosted by the Duke Center for Human Genetics, captured using SnagIt 7.0. The figure illustrates the appearance of the Statistic View panel and some of its features in the context of Contig View. It also demonstrates how the selection of a peak can enable the researcher to easily see where linkage data and other features such as potentially unstable tri-nucleotide repeats and gene expression data converge to suggest priority gene for association and sequence analysis.
Figure 2 Cyto View featuring Statistic View. A screen shot of private data hosted at primer.duhs.duke.edu illustrating the integration of Statistic View in the Ensembl CytoView page.
Figure 3 Table structures for the linkage object and the user access database. Part A of the figure shows that the study field serves as the key relating the User Data to the Linkage Data. Also shown are examples (part B) of three data records that define plot coordinates along with other attributes. User databases are created to allow restricted access for uploading data.
Figure 4 Flow chart illustrating the flow of data in Statistic View. A security feature is built in so that the process aborts if the user is not authenticated.
==== Refs
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 Geneontology: tool for the unification of biology. The Gene Ontology Consortium Nat Genet 2000 25 25 29 10802651 10.1038/75556
Stein LD Integrating biological databases Nat Rev Genet 2003 4 337 345 12728276 10.1038/nrg1065
Hubbard T Biological information: making it accessible and integrated (and trying to make sense of it) Bioinformatics 2002 18 S140 12385995
Heumann K Harris C Mewes HW A top-down approach to whole genome visualization Proc Int Conf Intell Syst Mol Biol 1996 4 98 108 8877509
Tatusova TA Karsch-Mizrachi I Ostell JA Complete genomes in WWW Entrez: data representation and analysis Bioinformatics 1999 15 536 543 10487861 10.1093/bioinformatics/15.7.536
NCBI Map Viewer
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
Project Ensembl Genome Browser
Kent WJ Sugnet CW Furey TS Roskin KM Pringle TH Zahler AM Haussler D The human genome browser at UCSC Genome Res 2002 12 996 1006 12045153 10.1101/gr.229102. Article published online before print in May 2002
UCSC Genome Bioinformatics
Hauser MA Li YJ Takeuchi S Walters R Noureddine M Maready M Darden T Hulette C Martin E Hauser E Xu H Schmechel D Stenger JE Dietrich F Vance J Genomic convergence: identifying candidate genes for Parkinson's disease by combining serial analysis of gene expression and genetic linkage Hum Mol Genet 2003 12 671 677 12620972 10.1093/hmg/12.6.671
Kong A Gudbjartsson DF Sainz J Jonsdottir GM Gudjonsson SA Richardsson B Sigurdardottir S Barnard J Hallbeck B Masson G Shlien A Palsson ST Frigge ML Thorgeirsson TE Gulcher JR Stefansson K A high-resolution recombination map of the human genome Nat Genet 2002 31 241 247 12053178
UniSTS integrated makers and maps
Schuler GD Pieces of the puzzle: expressed sequence tags and the catalog of human genes J Mol Med 1997 75 694 698 9382993 10.1007/s001090050155
Olson M Hood L Cantor C Botstein D A common language for physical mapping of the human genome Science 1989 245 1434 5 2781285
NCBI UniSTS ftp site
Schuler GD Sequence mapping by electronic PCR Genome Res 1997 7 541 50 9149949
Rotmistrovsky K Jang W Schuler GD A web server for performing electronic PCR Nucleic Acids Res 2004 W108 12 15215361
NCBI e-PCR
Kent WJ BLAT – the BLAST-like alignment tool Genome Res 2002 12 656 664 11932250 10.1101/gr.229202. Article published online before March 2002
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
Stajich JE Block D Boulez K Brenner SE Chervitz SA Dagdigian C Fuellen G Gilbert JG Korf I Lapp H 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
McPherson JD Marra M Hillier L Waterston RH Chinwalla A Wallis J Sekhon M Wylie K Mardis ER Wilson RK Fulton R Kucaba TA Wagner-McPherson C Barbazuk WB Gregory SG Humphray SJ French L Evans RS Bethel G Whittaker A Holden JL McCann OT Dunham A Soderlund C Scott CE Bentley DR Schuler G Chen HC Jang W Green ED Idol JR Maduro VV Montgomery KT Lee E Miller A Emerling S Kucherlapati Gibbs R Scherer S Gorrell JH Sodergren E Clerc-Blankenburg K Tabor P Naylor S Garcia D De Jong PJ Catanese JJ Nowak N Osoegawa K Qin S Rowen L Madan A Dors M Hood L Trask B Friedman C Massa H Cheung VG Kirsch IR Reid T Yonescu R Weissenbach J Bruls T Heilig R Branscomb E Olsen A Doggett N Cheng JF Hawkins T Myers RM Shang J Ramirez L Schmutz J Velasquez O Dixon K Stone NE Cox DR Haussler D Kent WJ Furey T Rogic S Kennedy S Jones S Rosenthal A Wen G Schilhabel M Gloeckner G Nyakatura G Siebert R Schlegelberger B Korenberg J Chen XN Fujiyama A Hattori M Toyoda A Yada T Park HS Sakaki Y Shimizu N Asakawa S Kawasaki K Sasaki T Shintani A Shimizu A Shibuya K Kudoh J Minoshima S Ramser J Seranski P Hoff C Poustka A Reinhardt R Lehrach H A physical map of the human genome Nature 2001 409 934 941 11237014 10.1038/35057157
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 FV 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 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 The sequence of the human genome Science 2001 291 1304 1351 11181995 10.1126/science.1058040
Lander ES Green P Construction of multi-locus genetic linkage maps in humans Proc Natl Acad Sci USA 1987 84 2363 2367 3470801
Marshfield Medical Center Genetics
Dib C Faure S Fizames C Samson D Drouot N Vignal A Millasseau P Marc S Hazan J Seboun E Lathrop M Gyapay G Morissette J Weissenbach J A comprehensive genetic map of the human genome based on 5,264 microsatellites Nature 1996 380 152 4 8600387 10.1038/380152a0
Galperin MY The molecular biology database collection: 2005 update Nucleic Acids Res 2005 D5 24 15608247
Li YJ Oliveira SA Xu P Martin ER Stenger JE Hulette C Scherzer CR Hauser MA Scott WK Small GW Nance MA Watts RL Hubble JP Koller WC Pahwa R Stern MB Hiner BC Jankovic J Goetz CG Mastaglia F Middleton LT Roses AD Saunders AM Welsh-Bohmer KA Schmechel DE Gullans SR Haines JL Gilbert JR Vance JM Pericak-Vance MA Glutathione S-transferase omega-1 modifies age-at-onset of Alzheimer disease and Parkinson disease Hum Mol Genet 2004 13 573 10.1093/hmg/ddh059
Lobachev KS Stenger JE Kozyreva OG Jurka J Gordenin DA Resnick MA Inverted Alu repeats unstable in yeast are excluded from the human genome EMBO J 2000 19 3822 3830 10899135 10.1093/emboj/19.14.3822
Stenger JE Lobachev KS Gordenin D Darden TA Jurka J Resnick MA Biased distribution of inverted and direct Alus in the human genome: implications for insertion, exclusion, and genome stability Genome Res 2001 11 12 27 11156612 10.1101/gr.158801
Duke Center for Human Genetics Software Repository
Download Ensembl Wiki web URL
Ensembl Anonymous FTP site
Ensembl installation instructions pdf file
GZIP
WinZip Windows Archive Utility
| 15826305 | PMC1087836 | CC BY | 2021-01-04 16:02:48 | no | BMC Bioinformatics. 2005 Apr 12; 6:95 | utf-8 | BMC Bioinformatics | 2,005 | 10.1186/1471-2105-6-95 | oa_comm |
==== Front
BMC DermatolBMC Dermatology1471-5945BioMed Central London 1471-5945-5-41585422410.1186/1471-5945-5-4DatabaseIntroducing the National Library for Health Skin Conditions Specialist Library Grindlay Douglas [email protected] Maged N Kamel [email protected] Hywel C [email protected] National Library for Health Skin Conditions Specialist Library, Centre of Evidence-Based Dermatology, Queen's Medical Centre, University of Nottingham, Nottingham NG7 2UH, UK2 School for Health, University of Bath, Claverton Down, Bath BA2 7AY, UK2005 26 4 2005 5 4 4 17 2 2005 26 4 2005 Copyright © 2005 Grindlay et al; licensee BioMed Central Ltd.2005Grindlay 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
This paper introduces the new National Library for Health Skin Conditions Specialist Library .
Description
The aims, scope and audience of the new NLH Skin Conditions Specialist Library, and the composition and functions of its core Project Team, Editorial Team and Stakeholders Group are described. The Library's collection building strategy, resource and information types, editorial policies, quality checklist, taxonomy for content indexing, organisation and navigation, and user interface are all presented in detail. The paper also explores the expected impact and utility of the new Library, as well as some possible future directions for further development.
Conclusion
The Skin Conditions Specialist Library is not just another new Web site that dermatologists might want to add to their Internet favourites then forget about it. It is intended to be a practical, "one-stop shop" dermatology information service for everyday practical use, offering high quality, up-to-date resources, and adopting robust evidence-based and knowledge management approaches.
==== Body
Background
Quality information all in one place
Dermatology is currently undergoing enormous changes in the way it is practised, and much of this is a result of a fundamental change in the manner in which information is exchanged through information technology and the Internet. The Web is increasingly becoming an indispensable tool for healthcare professionals caring for people with skin disease in their everyday clinical practice [1-8].
There are many quality online information resources on skin conditions, but locating these sources of information quickly and easily, and bringing together related information into one place, remains a big challenge for the average Internet user. Internet search engines such as Google or Google Scholar can be tried, but such an approach can be very time consuming and frustrating. Search engine results, including those from Google, contain too much "noise" in the form of irrelevant or low-quality material. Choosing suitable search terms, including skin disease or drug synonyms/variants, is an art in itself, and once a list of hits is obtained, the user has to wade through them to separate "the wheat from the chaff".
Although some quality dermatology Internet portals specialise in images, patient information leaflets, online tutorials, etc., nowhere does this information appear to be pulled together into one trustworthy source.
This is where the new UK National Library for Health Skin Conditions Specialist Library ( – Figure 1) should help. This new online digital library aims to catalogue, organise and harness the Web to the advantage of our users – primarily UK healthcare professionals caring for people with skin disease.
Figure 1 Homepage of the NLH Skin Conditions Specialist Library. Screenshot of the homepage of the NLH Skin Conditions Specialist Library as launched in February 2005.
How the Skin Conditions Specialist Library fits into the National Library for Health
In 2003, 19 Specialist Libraries were commissioned within the National electronic Library for Health (NeLH – ), to bring together resources for a range of clinical subject areas. Just as there are specialist sections in a real library branching off from the central general reading section, so too there are virtual specialist libraries in the NeLH, catering for more specialist groups and knowledge requirements. The Skin Conditions Specialist Library was one of these, and now even more Specialist Libraries are in the process of being commissioned, taking the number up to 26.
The aim of the NeLH was to organise clinical knowledge and to promote evidence-based decision making. The NeLH brought together over 70 electronic resources such as sources of guidelines, systematic reviews and bibliographic databases in one place, so that they could be accessed at any time wherever an Internet connection is available. Examples of key generic resources included in the NeLH are shown in Table 1.
Table 1 Examples of key generic resources included in the NLH. Examples of key generic resources (or core content) included in the National Library for Health . Access to all of these information sources in full is now completely free, although some, such as Drug and Therapeutics Bulletin and many of the full-text journals, require a valid Athens account.
Resource Internet address
National Institute for Clinical Excellence (NICE) Guidelines
NeLH Guidelines Finder
PRODIGY
Hitting the Headlines
Cochrane Database of Systematic Reviews
NHS Health Technology Assessment Programme
CRD Database of Abstracts of Reviews of Effects (DARE)
CRD Health Technology Assessment (HTA) Database
Clinical Evidence
Drug and Therapeutics Bulletin
Bandolier
Full-text journals
Recently, the NeLH was subsumed into a bigger whole, the National Library for Health (NLH), with a new URL and home page . Hence, the Skin Conditions Specialist Library is now to be known as the NLH Skin Conditions Specialist Library. The NLH is a major component of the National Knowledge Service , which in turn is part of England's National Programme for IT . The NLH is funded by the NHS (National health Service) to combine the national digital resource managed by the existing NeLH with the physical content of books and journals and the skills and resources of health libraries (i.e. libraries in the NHS, health libraries in Higher Education, and the libraries of national organisations such as Royal Colleges and major charities).
Construction and content
Aims and scope
The NLH Skin Conditions Specialist Library is an Internet information gateway for dermatology. It aims to provide an organised, easily accessible and up-to-date directory of key documents, reviewed evidence, and appraised information on skin conditions, including patient information resources. The emphasis is on quality and an evidence-based approach.
The subject remit of the Skin Conditions Specialist Library is the diagnosis, treatment, management and prevention of skin conditions, and the effect of skin conditions on quality of life. The scope is deliberately wide and includes disfigurement, skin cancer, skin surgery, wound care, sexually transmitted diseases that affect the skin, alternative and complementary treatments, consumer skin care, and cosmetic aspects such as skin ageing.
Identification of potential subject overlap is important, so that the various Specialist Libraries within the NLH can work together collaboratively to avoid unnecessary duplication of effort. The database used for the NLH Specialist Libraries allows cross-referencing where resources overlap and a synergistic, efficient use of resources.
The Skin Conditions Specialist Library has identified subject overlaps with cancer, child health, wound care, plastic surgery, sexually transmitted diseases, allergy, and alternative and complementary medicine. It has been proposed that plastic surgery and disfigurement should be housed within the Skin Conditions Specialist Library, at least for now.
Audience
Although anyone connected to the Internet in the UK and elsewhere in the world, including patients with skin diseases and their relatives, is able to access the Skin Conditions Specialist Library, the Library's purpose is primarily to support UK healthcare professionals. As the Library's content develops, it is hoped that the Skin Conditions Specialist Library will become an important and well-used resource for all National Health Service (NHS) professionals dealing with skin conditions, including dermatology consultants and specialist registrars, junior doctors, dermatology nurses, general practitioners, practice nurses and health visitors. The Library should also prove useful to medical students and student nurses studying dermatology at undergraduate and postgraduate levels.
Core Project Team, Editorial Team and Stakeholders Group
The National Library for Health Skin Conditions Specialist Library is based at the Centre of Evidence Based Dermatology at the University of Nottingham (which also hosts the Cochrane Skin Group – ). The core Project Team consists of the Clinical Lead, Professor Hywel Williams, Professor of Dermato-Epidemiology and Co-ordinating Editor of the Cochrane Skin Group, and the Information Specialist, Dr Douglas Grindlay.
A small Editorial Team, which includes members from the British Association of Dermatologists (BAD – ) and the British Dermatological Nursing Group (BDNG – ), along with other information scientists, provides advice on major issues of policy, content and quality (see ).
A wide-ranging Stakeholders Group has also been set up to ensure that the needs and views of all potential users are taken into account. The Stakeholders Group includes representatives from professional organisations such as the BAD, the BDNG, the Primary Care Dermatology Society , the British Association for Sexual Health and HIV , the British Association of Plastic Surgeons , and the Cochrane Skin Group . There are also representatives from an overarching body for patient support groups (the Skin Care Campaign – ) and five of the larger patient support groups. Another group of stakeholders included are health information providers, with representation from other NLH Specialist Libraries, NHS Direct Online , the OMNI database , the TRIP database , and the Chartered Institute of Library and Information Professionals (CILIP – ).
Resource and information types
Table 2 lists the resource and information types that are included in the National Library for Health Skin Conditions Specialist Library, and those that are excluded from it under present policy.
Table 2 Skin Conditions Specialist Library resource and information types. Resource and information types that are included in the National Library for Health Skin Conditions Specialist Library, and those that are excluded from it.
Resource and information types included in the Skin Conditions Specialist Library
- Health policies and strategies
- Health initiatives
- National standards
- National and selected international guidelines and guidance
- Commissioning statements and service specifications
- Health Care Needs Assessment (HCNA)
- Pathways (from )
- Audit
- Skin disease statistics
- National Service Frameworks (NSFs) – when available
- Systematic reviews
- Sources appraising primary research
- Critically Appraised Topics (CATs)
- Clinical answering services (e.g. has two dermatology questions as at 31 January 2005)
- Selected research studies on skin conditions (as determined by the Editorial Team)
- Bibliographic databases
- Internet resource guides and gateways
- Special resources and tools (e.g. special collections of electronic journals, diagnostic aids, and dermatology images – )
- Education, teaching and continuing professional development resources
- Evaluated patient information
- Professional bodies and health organisations
- Patient support groups and societies
- News
- Events
Resource and information types are excluded from the Skin Conditions Specialist Library
- Research protocols or research in progress
- Cochrane protocols – the full systematic review will be added once it is published
- Product information provided by product manufacturer or supplier that has not been independently evaluated
- Web sites of commercial organisations (e.g. pharmaceutical companies, private hospitals)
- Personal Web sites
- Sites under construction (unless significant information is available to Editorial Team at the time of assessment)
- Information in a language other than English
Collection building strategy
Individual Web resources on specific topics are discovered through systematic and regular trawling and searching of a core set of "sources of resources" that have been identified for this purpose:
- Sources for guidelines, pathways and policy documents, such as the NLH Guidelines Finder , PRODIGY and BAD
- Sources for systematic reviews and quality-assured synopses of evidence, such as the Cochrane Library
- Sources appraising primary research, such as Bandolier
- Sources for education and continuing professional development, such as BAD and BDNG
- Internet resource guides, such as Health on the Net Foundation HONselect and OMNI
- Web sites of professional bodies, skin research charities, support groups and other organisations concerned with skin conditions
- Sources of evaluated patient information/leaflets, such as BAD and PRODIGY
- Sources for news and events, e.g. NLH Hitting the Headlines and other news feeds and e-mail alerts
The relevant sections of each "source of resources" (in particular content lists and indexes) will be regularly visited and scanned for new documents, with the date of each search and the search strategy being recorded.
Internet search engines will also be regularly searched for relevant resources, helped by meta-search tools like Copernic . Again, the date of each search and query terms used will be recorded.
Quality standards and selection criteria
Quality benchmarking and filtering of candidate resources is an issue of prime importance in ensuring that only those resources providing the best current knowledge and evidence are indexed in the National Library for Health Skin Conditions Specialist Library [9]. Resources will be selected if they meet all of the following criteria:
1. The resource falls under the subject scope for the Skin Conditions Specialist Library and is relevant to its intended users.
2. The resource and information it contains are of a type specified for inclusion in the Skin Conditions Specialist Library (see 'Resource and information types' above).
3. The resource has a stable enough presence on the Internet to be useful.
4. The resource is freely available on the Internet, or is accessible to NHS health professionals.
5. The resource is not strictly local in context.
6. Authorship of the resource can be ascertained and contact details are available.
7. Qualified individuals or groups take responsibility for the resource, coming from reputable organisations or having recognised expertise and authority in the field.
8. There is a process of refereeing of the clinical content by qualified health professionals or appropriate specialists.
9. The source of the content is verifiable and there is proper attribution.
10. Sources of information are clearly identified.
11. The information is current (e.g. latest version of guidelines) and kept up-to-date. Resources published more than ten years ago will not be included, unless they are judged to have special significance and value by the Editorial Team. Inclusion of resources will be reconsidered every year, and resources will be removed if out-of-date or superseded by others.
12. In the view of the Editorial Team, the information does not have any inaccuracies or bias.
13. For sites sponsored by commercial companies, in particular drug companies, for a site to be accepted, any sponsorship should be explicit and transparent, the content should be editorially independent, and there should be explicit steps to minimise bias or commercial selectivity.
It should be noted that the quality criteria apply to the resources as catalogued, not to any further links they may contain.
An important policy of the Skin Conditions Specialist Library is that resources on national NHS Web sites and those which are identified via the NLH core content are automatically eligible for inclusion, unless specific reasons for exclusion are identified by the Editorial Team.
Patient information resources present particular problems with assessing provenance and quality. The resource must be written, edited or reviewed by a qualified medical professional and there should be an explicit process of medical refereeing and regular review. Priority will be given to patient information with acceptable accreditation criteria, such as NHS Information Partners Web sites or the Centre for Health Information Quality (CHIQ) TriangleMark scheme . Readability of patient information material [10] may perhaps be considered in a future revision of our quality benchmarking criteria.
A formal record will be kept of all resources that are rejected, with reasons. For example, we have recently come across a non-UK online dermatology image database that seemed a very good candidate for inclusion in our resource database, but was nevertheless rejected on quality grounds because it had no information as to provenance and the authority of the person producing it.
Taxonomy for content indexing, organisation and navigation
A major advantage of the Skin Conditions Specialist Library is the organisation of its content by skin topic and resource type (e.g. Guidance and Pathways, Evidence, etc.), to facilitate retrieval by users (Figure 2). The navigation menu for skin conditions that is used has been derived from the logical, diagnostic structure of the BAD Diagnostic Index ( and Figure 3), with the kind assistance of Dr Robert Chalmers from the BAD Health Informatics sub-Committee, who is also a member of our Editorial Team. This means that the skin condition categories are mappable to those used by the Cochrane Skin Group. The structure of the BAD Index is due to be incorporated into the SNOMED CT (Systematized Nomenclature of Medicine Clinical Terms – ) terminology project in the future, so that as NHS electronic patient records are developed using SNOMED CT, it should be relatively easy to refer across to the Skin Conditions Specialist Library [11].
Figure 2 Browsing the Library by skin topic and resource type. Screenshot of a typical page from the National Library for Health Skin Conditions Specialist Library, for the menu topic "Treatment and management > Sunscreens & sun protection". On the left is the navigation menu which contains different topics such as skin conditions, treatments and organisations. Towards the top of the page, below the total number of records matched, can be seen the six tabs that are used to display records for different resource types. In this case the tab for "Evidence" has been selected, which includes systematic reviews.
Figure 3 The Library's navigation menu for skin conditions. Screenshot of the Library's main topic headings for skin conditions, which are based on the BAD Diagnostic Index. The topic menu is always accessible on the left-hand side of the screen. The topic headings under 'Conditions' have been split into two in a graphics editor to reduce the height of this screenshot. Only higher-level categories are shown here, except for the 'Immunobullous (blistering) conditions' category (highlighted in light blue), which has been expanded.
Tagging of resources in the Skin Conditions Specialist Library has been as detailed as possible, with tagging to multiple topics when this will help retrieval by users (Figure 4). Resources are tagged to the different Treatment & Management headings as well as to Condition headings. Thus, for example, it is possible to click on the menu topic for 'Methotrexate' (under 'Systemic drugs') and find all resources with significant discussion of this drug, such as guidelines, Cochrane Reviews, patient safety alerts, etc. Resources are also tagged for 'Speciality' when they are likely to be of interest to particular users, such as nursing (for pressure ulcers, leg ulcers, wounds, diabetic foot, etc.), surgery, or management.
Figure 4 A typical resource record from the new Skin Conditions Specialist Library. Screenshot of a typical resource record page for a resource titled 'Probiotics for atopic diseases' from the new Skin Conditions Specialist Library database of resources. Note the 'Keywords' and 'Topics' fields near the bottom of the screenshot.
However, the authors appreciate that their navigational classification of skin conditions based on the BAD Diagnostic Index will more likely suit the needs of specialist dermatologists rather than those of non-specialists and novices. In fact, in a classification task reported by Norman et al. [12], novices and non-specialists categorised skin lesions presented to them on slides by their surface features (e.g. scaly lesions), intermediates grouped the slides according to diagnosis, and expert dermatologists organised the slides according to superordinate categories such as viral infections, which reflected the underlying pathophysiological or aetiological taxonomy.
Fortunately, the NLH also incorporates powerful search functions, allowing users to search just the Skin Conditions Specialist Library, other Specialist Libraries, or all NLH collections in one go. This complementary search facility should care for the needs of many users, especially those who are non-specialists and less familiar with the formal classifications, terminology and diagnostic features of skin diseases used in our navigation menu.
User interface and usability issues
The usability of the Skin Conditions Specialist Library is very much determined centrally by the parent organisation, the NLH. The user interface of the Library is consistent with the interfaces of other NLH Specialist Libraries in order to permit familiarity when moving from one Specialist Library to another, as a primary care practitioner will often need to do. In fact, nearly all of the Specialist Libraries are currently using the same content management system, and all should migrate to this system in the near future.
We know the fate of Web interfaces that fail to take into account all the stakeholders, and indeed the reader might think that in our case interface requirements are imposed by management rather than coming from actual users. However, this is not true for the NLH Specialist Libraries. Their shared content management system/interface, which is under continual improvement, is the result of ongoing analyses of users' requirements and consultations handled centrally by the NLH and covering a wide range of user types. One aim is to ensure interface consistency and a unified look and feel across the different parts of the NLH. Another aim is to free the time of the different Specialist Library Teams, so that they may concentrate on their primary roles in analysing the information needs of their specific users, and in finding and indexing quality online information.
Utility and future developments
A "one-stop-shop" service
The National Library for Health Skin Conditions Specialist Library is intended to provide a "one-stop shop", a single portal that can be used as a service to find quality, evidence-based information on dermatology that is relevant for UK health professionals. The Skin Conditions Specialist Library provides an organised, easily accessible and up-to-date collection of key documents, reviewed evidence, and appraised information on skin conditions, including selected patient information resources.
Current priorities
As with the rest of the NLH, the initial focus on content is currently on NHS-branded or NHS-funded information. The Skin Conditions Specialist Library is very much a work in progress, and with time we expect it to grow to include more and more external resources (as long as they meet our quality standards – see 'Quality standards and selection criteria' above). The core resources will always be those based on the highest level of clinical evidence, such as systematic reviews, critically appraised synopses of the evidence and guidelines.
With just one full-time information specialist working on the Skin Conditions Specialist Library, the priority has to be searching for, identifying, and signposting existing information that is relevant and of high quality. The resources and opportunities for specific new content generation in house will be limited, at least initially.
Patient information in the new Library
The initial source of information for patients is intended to be NHS Direct Online . However, there are so many chronic skin diseases where patients can become experts in their own conditions [13]. They will often want more information than can be obtained from typical consumer information sources such as NHS Direct Online, and they should be able to find this on the NLH.
For this reason, we hope the Library will also act as a single source for high quality patient information, such as that produced by the BAD, NHS Direct Online, DermNet , PRODIGY, and Informed Health Online . While this information will then be available directly to patients and relatives using the NLH, it will also be an important resource for dermatologists, GPs and nurses when they are looking for information to give out during consultations.
As skin conditions are often chronic, disfiguring and difficult to treat, we believe very strongly in this role for the Library and in the need for patients to be consulted as the Library develops (hence the inclusion of patient groups in our Stakeholders Group).
Identifying gaps in evidence and knowledge base
The Skin Conditions Specialist Library should have an important role in identifying gaps in the evidence and knowledge base, which will help to prioritise future research, reviews of the evidence and policy development. Our consultations so far have identified a need for the Library to include quality, evidence-based information on the "rarer" skin conditions. Finding such information is probably the main reason experienced dermatologists will want to use the service initially, but ironically it is just this information that is most often lacking, at least in electronic form. As stated above, the main function of the Library is to bring together existing quality resources rather than generating new content. A potential role in discovering the gaps and fostering the production of appropriate electronic resources to fill them has been identified, working in co-operation with our Stakeholders' organisations.
A current awareness service
Another important function of the Skin Conditions Specialist Library is to provide a current awareness service, to alert our users to important new research, systematic reviews, guidelines, policy developments, news and conferences that are relevant to the Library's remit. This function should develop and expand with time.
Future directions
To be used regularly by its target audience, the Skin Conditions Specialist Library needs to provide an added-value service over and above a mere directory of links. The aim is to build an active community of repeat users who will make our Library their starting point in answering all their dermatology information needs.
Depending on users' particular needs and current situation, not all quality resources will have the same practical relevance to them (some quality resources might even act as distracting "noise"). One of the Library's major functions is knowledge management, i.e., avoiding information overload by striving to present results relevant to our users (according to their roles), and by highlighting the most important resources (e.g. those that provide a comprehensive overview of a subject). Key resources should eventually be tagged in the Library using a new "Editor's pick" function.
Another major issue for the Library to resolve is the tension between organising resources on the basis of a condition/diagnosis, and accessing information when a diagnosis is unknown, for example by symptoms/lesion morphology and location on the body. Both approaches are needed, but the latter would be important for non-specialists such as GPs and NHS Direct nurses, particularly in view of the large number of skin conditions that exist. Probably some form of diagnostic and learning facility/decision support tool will be required in the future. An example of such a tool is VisualDX . As well as helping to access information in the absence of a pre-existing diagnosis, a system of this kind would be invaluable for differential diagnosis and continuing professional development ("learn as you practise"). It would also be of great use to students. There are two (complementary) possibilities for such systems, either as a separate diagnostic program or as pop-up tool embedded into electronic patient records (contextual decision-support system). In the meantime, the Library already includes a topic heading for Images, with links to several dermatology image databases.
Another long-term goal for the Skin Conditions Specialist Library team is to develop special features such as learning and continuing professional development resources, quizzes and commissioned briefings. Contributions from users could also be incorporated, for example suggestions on useful Web sites, hence the importance of including a comment return system on the Library's homepage . Another possibility is to have a "Web site of the month" feature.
There is also the possibility of using RSS (Really Simple Syndication) feeds to provide our users with regular updates on relevant new Library content, a feature already available elsewhere on the NLH (see ).
These possibilities will be investigated further once the core Library content has been fully established.
Conclusion
The new National Library for Health Skin Conditions Specialist Library is an exciting project that aspires to be of real, practical use to all health professionals who treat, manage and support people with skin conditions. The Library is not just another dermatology Web site that users might add to their Internet favourites, but rather a single source where users can view quality-assured information rapidly, easily and for free. It also has the potential to highlight existing gaps in the evidence and knowledge base for dermatology, and to foster the development of teaching and learning resources.
Although commissioned for, and targeted at, UK users, we hope that the free access to the Skin Conditions Specialist Library will mean that users from other countries around the world will visit and use our site.
We consider all users (and potential users) of the Skin Conditions Specialist Library as our stakeholders. Comments and suggestions for improvement are always welcome, and we hope with time to develop a community of interest in skin conditions in the UK supported by newsletters and regular updates via the Library pages.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
As member of the Library's Editorial Team, MNKB advised on and participated in formulating many of the Library's aspects, including its editorial policies and quality checklist. MNKB also conceived and drafted this manuscript. The Library's core Project Team comprises DG, the Library's Information Specialist, and HCW, the Library's Clinical Lead, who together are responsible for developing and managing the Library, and organising the activities of its Editorial Team and Stakeholders Group. DG and HCW also revised and edited the first drafts of this manuscript and provided valuable input that helped improving it. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The National Library for Skin Conditions is funded by the NHS National Library for Health (NLH).
==== Refs
Dill SW Digiovanna JJ Changing paradigms in dermatology: information technology Clin Dermatol 2003 21 375 82 14678717 10.1016/j.clindermatol.2003.08.006
Gjersvik PJ Nylenna M Aasland OG Use of the Internet among dermatologists in the United Kingdom, Sweden and Norway Dermatol Online J 2002 8 1 12546756
Pollock B Goulden V Sheehan-Dare R Goodfield M Dermatology and the World Wide Web Br J Dermatol 2001 144 914 5 11298569 10.1046/j.1365-2133.2001.04165.x
Wainwright BD Clinically relevant dermatology resources and the Internet: An introductory guide for practicing physicians Dermatol Online J 1999 5 8 10673461
Sitaru C Dermatology resources on the Internet: a practical guide for dermatologists Int J Dermatol 1998 37 641 7 9762811 10.1046/j.1365-4362.1998.00545.x
Diepgen TL Eysenbach G Digital images in dermatology and the Dermatology Online Atlas on the World Wide Web J Dermatol 1998 25 782 7 9990769
Macey WH Dermatology and the World Wide Web Dermatol Nurs 1997 9 159 62 9233152
Krejci-Papa NC Bittorf A Diepgen T Huntley A Dermatology on the Internet. A source of clinical and scientific information J Dermatol Sci 1996 13 1 4 8902646 10.1016/0923-1811(95)00510-2
Boulos MN Roudsari AV Gordon C Muir Gray JA The use of quality benchmarking in assessing Web resources for the Dermatology Virtual Branch Library of the National electronic Library for Health (NeLH) Journal of Medical Internet Research 2001 3 e5 11720947
Baker GC Newton DE Bergstresser PR Increased readability improves the comprehension of written information for patients with skin disease J Am Acad Dermatol 1988 19 1135 41 3204187
Boulos MN Roudsari AV Carson ER A Dynamic Problem to Knowledge Linking Semantic Web Service Based on Clinical Codes Medical Informatics & The Internet in Medicine 2002 27 127 137 12507259 10.1080/14639230210147359
Norman GR Rosenthal D Brooks LR Allen SW Muzzin LJ The development of expertise in dermatology Arch Dermatol 1989 125 1063 8 2757402 10.1001/archderm.125.8.1063
Muir Gray JA The Resourceful Patient 2002 Oxford, UK: eRosetta Press
| 15854224 | PMC1087837 | CC BY | 2021-01-04 16:29:46 | no | BMC Dermatol. 2005 Apr 26; 5:4 | utf-8 | BMC Dermatol | 2,005 | 10.1186/1471-5945-5-4 | oa_comm |
==== Front
BMC GastroenterolBMC Gastroenterology1471-230XBioMed Central London 1471-230X-5-141582900910.1186/1471-230X-5-14Research ArticleThe accuracy of the report of hepatic steatosis on ultrasonography in patients infected with hepatitis C in a clinical setting: A retrospective observational study Hepburn Matthew J [email protected] Jeffrey A [email protected] Eric P [email protected] Eric J [email protected] Departments of Medicine Brooke Army Medical Center, Fort Sam Houston Texas, USA2 Department of Pathology, Brooke Army Medical Center, Fort Sam Houston Texas, USA2005 13 4 2005 5 14 14 22 11 2004 13 4 2005 Copyright © 2005 Hepburn 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
Steatosis is occasionally reported during screening ultrasonography in patients with hepatitis C virus (HCV). We conducted a retrospective observational study to assess the factors associated with steatosis on ultrasonography and the relationship between steatosis on ultrasound versus biopsy in patients infected with HCV in a clinical setting. Our hypothesis was ultrasonography would perform poorly for the detection of steatosis outside of the context of a controlled study, primarily due to false-positive results caused by hepatic fibrosis and inflammation.
Methods
A retrospective review of ultrasound reports was conducted on patients infected with HCV in a tertiary care gastroenterology clinic. Reports were reviewed for the specific documentation of the presence of steatosis. Baseline clinical and histologic parameters were recorded, and compared for patients with vs. without steatosis. Multiple logistic regression analysis was performed on these baseline variables. Liver biopsies were reviewed by two pathologists, and graded for steatosis. Steatosis on biopsy was compared to steatosis on ultrasound report, and the performance characteristics of ultrasonography were calculated, using biopsy as the gold standard.
Results
Ultrasound reports were available on 164 patients. Patients with steatosis on ultrasound had a higher incidence of the following parameters compared to patients without steatosis: diabetes (12/49 [24%] vs. 7/115 [6%], p < 0.001), fibrosis stage >2 (15/48 [31%] vs. 16/110 [15%], p = 0.02), histologic grade >2 (19/48 [40%] vs. 17/103 [17%], p = 0.002), and ALT (129.5 ± 89.0 IU/L vs. 94.3 ± 87.0 IU/L, p = 0.01). Histologic grade was the only factor independently associated with steatosis with multivariate analysis. When compared to the histologic diagnosis of steatosis (n = 122), ultrasonography had a substantial number of false-positive and false-negative results. In patients with a normal ultrasound, 8/82 (10%) had >30% steatosis on biopsy. Among patients with steatosis reported on ultrasound, only 12/40 (30%) had >30% steatosis on biopsy review.
Conclusion
Steatosis on ultrasound is associated with markers of inflammation and fibrosis in HCV-infected patients, but does not consistently correlate with steatosis on biopsy outside of the context of a controlled study. Clinicians should be skeptical of the definitive diagnosis of steatosis on hepatic ultrasonography.
==== Body
Background
Hepatic steatosis is commonly detected on both radiologic and histologic examination in patients infected with the hepatitis C virus (HCV). In particular, steatosis is often observed in genotype 3 infection [1]. The underlying mechanism of steatosis in hepatitis C is not completely understood, and is most likely multifactorial. Animal models [2,3] and in vitro experimentation [4] suggest that a virologic effect induces steatosis. Clinical observational data suggest that a possible direct virologic effect occurs with hepatitis C genotype 3 infection [1,5]. However, steatosis in genotype 1 and 2 HCV infections is more closely related to risk factors that are known to be associated with non-alcoholic steatohepatitis (i.e. diabetes mellitus and obesity)[5,6]. The assessment of steatosis may have clinical relevance, as some studies have suggested that the detection of steatosis can be an independent predictor of response to therapy [7].
Clinicians routinely order hepatic ultrasonography on patients with hepatitis C as an initial screening test for hepatocellular carcinoma or anatomic abnormality. In ordering this test, the radiologist may also detect and mention steatosis. However, the diagnosis of steatosis on ultrasound can be problematic, as the ultrasonic appearance of steatosis may be that of a 'bright liver,' which can also be observed with hepatic fibrosis and/or inflammation [8]. Even though these overlap syndromes exist, we observed a number of hepatic ultrasonography reports identifying steatosis in our clinical practice, as opposed to echogenic or 'bright liver'.
The sensitivity and specificity of ultrasonography for the detection of steatosis may be very high in the hands of an expert radiologist who consistently applies a particular criteria. However, many clinicians in the United States employ the services of a radiology department or group, in which many radiologists with varying levels of experience actually interpret the images. We were interested in the utility of a report of steatosis in this context.
Only one prior study has compared steatosis on ultrasound to histology in HCV-infected patients [9], and no correlation was observed between histologic and radiologic results. However, the conclusions of this study are limited since the diagnosis of HCV was based on the detection of HCV antibodies, rather than a determination of viremia with polymerase chain reaction (PCR) technology. Additionally, this study was not performed in the context of clinical practice.
We conducted a retrospective observational study to assess the radiology reports of hepatic ultrasonography of HCV-infected patients. Our purpose was to determine demographic, clinical and laboratory characteristics associated with radiology reports of steatosis on ultrasound in HCV-infected patients and to describe the association between steatosis on ultrasound and steatosis on liver biopsy. We specifically wanted to clarify the implications of the identification of steatosis on an ultrasound report for the practicing clinician. Our purpose was not to assess the performance characteristics of hepatic ultrasonography under controlled conditions. Rather, we were interested in the utility of these reports in routine clinical practice, with different radiologists over a three-year period. The first part of our analysis was to examine all patients who had hepatic ultrasonography performed, including patients who did not have a liver biopsy. The second part of our analysis (comparing steatosis on ultrasound report to steatosis on biopsy) included only patients who had hepatic ultrasonography and liver biopsy. We observed that ultrasound reports of hepatic steatosis were particularly associated with histologic inflammation, as well as fibrosis, and the sensitivity and specificity of steatosis on ultrasound was poor, when compared to steatosis on biopsy.
Methods
Patients
A retrospective review was performed of all patients who underwent screening hepatic ultrasonography seen in the hepatitis research clinic at our tertiary care, military academic medical center over a three-year period. Patients were those eligible for care at a military treatment facility, including active duty military personnel, their spouses, and military retirees (serving 20 or more years on active duty). All patients had hepatitis C infection, confirmed by serum HCV RNA PCR testing. Patients were referred from primary care clinics at our institution, and military primary care clinics throughout the region (Texas, Louisiana, Oklahoma, Kansas, Missouri and Colorado) for further assessment of their hepatitis C infection and for discussion of treatment options. The Institutional Review Board at Brooke Army Medical Center approved the study.
Demographic and Clinical Characteristics: Age, gender, body mass index (BMI) and ethnicity were recorded on each patient and utilized for analysis. The following laboratory parameters were noted: alanine aminotransferase (ALT), alpha-fetoprotein (AFP), and HCV genotype. Patients were considered to have diabetes if they were taking medication for diabetes treatment. Liver biopsy reports included a Metavir stage (0=no fibrosis to 4=cirrhosis) [10] and histologic grade (0–4, with 0=no inflammation to 4=severe inflammation), which were also included in the analysis. Each of these parameters was converted into a binomial variable, with 0–2 and 3–4 as the resultant two categories.
Ultrasonography
All patients underwent hepatic ultrasonography during their initial evaluation if they did not have an ultrasound report from a prior assessment. Computerized records were reviewed for the ultrasonography reports. Ultrasonography was primarily performed at Brooke Army Medical Center, the same institution as our hepatitis research clinic. The machines utilized in the Radiology Department were the ATL Ultramark HDI 3000® Ultrasound System and the ATL Ultramark HDI 5000® Ultrasound System. A small number of patients in the hepatitis research clinic (<10%) are referred from Wilford Hall Air Force Medical Center, also located in San Antonio, Texas. If the patient had their ultrasound performed at the Air Force hospital, they were included in the analysis. The Radiology Departments at these two institutions (Brooke Army Medical Center and Wilford Hall Air Force Medical Center) maintain a combined accredited teaching program for radiology residents. These reports were scored as a binomial variable. If the ultrasound report mentioned steatosis as a finding, it was designated positive. If the ultrasound report did not mention steatosis, it was labeled negative. Equivocal studies containing phrases such as "possible steatosis" or "inflammation or steatosis" were excluded from the final analysis. Reports did not distinguish between diffuse or focal fatty liver.
Histologic Examination
A liver biopsy was offered to patients during their initial evaluation in order to assess the patient's severity of disease. Liver biopsies were generally performed within 1–2 months of hepatic ultrasound. There were no liver biopsies in this study utilized that were > 6 months after ultrasonography. The initial histologic examination of the biopsy focused on the histologic stage (fibrosis) and grade (inflammation). For our study, two pathologists (J.A.V. and E.P.F.) retrospectively reviewed these biopsies and a percentage of steatosis was assigned to each biopsy specimen. Any steatosis was denoted as 1%, while a complete absence of steatosis was classified as 0%. If steatosis was considered >2%, it was recorded in 5% increments. These pathologists were blinded to ultrasound results, as well as any clinical characteristics of the patients. In order to explore the relationship between the report of steatosis on ultrasound compared to biopsy, the biopsies were categorized three different ways:
1) grouped into 5 categories (based on the percentage of steatosis on biopsy): 0–2%, 2–10%, 10–30%, 30–60%, and >60%.
2) As a binomial, comparing no steatosis (<2%) vs. any steatosis (>2%).
3) As a binomial, comparing significant steatosis (>30%) vs. not significant steatosis (<30%).
Analysis
Descriptions of the 'operative characteristics' (sensitivity, specificity, positive predictive value, negative predictive value) of steatosis on ultrasound in our clinical setting were calculated, with the assumption that the gold standard for the diagnosis of hepatic steatosis was histologic examination. These calculations were performed for both the determination of any steatosis and significant steatosis on histologic examination. Any steatosis was defined as more than 2% steatosis on biopsy, while significant steatosis was defined as >30% steatosis on biopsy. Prior literature suggests that ultrasonography performs well in the diagnosis of significant steatosis [11].
Demographic, clinical and histologic characteristics were compared between patients with steatosis on biopsy versus patients without steatosis on biopsy. For continuous variables, an independent sample t-test was utilized if their distribution was normal. For continuous variables in which the distribution of values was not normal, the Mann-Whitney rank sum test was used. For categorical variables, a Pearson's chi-square test was employed. To determine the association between steatosis on ultrasound and steatosis on biopsy, a Spearman's rank correlation coefficient was calculated. A kappa-statistic was calculated to assess interobserver variability between steatosis percentages observed by the two pathologists. Agreement between pathologists was determined to be +/- 5% for the liver biopsies examined.
A multiple logistic regression model, with stepwise backward elimination of non-significant variables, was developed to determine factors that were associated with the detection of steatosis on ultrasound. Steatosis on ultrasound was the dependent variable, with the independent variables including: age, gender, histologic stage (as a binomial variable), histologic grade (as a binomial variable), ALT, genotype (genotype 3 compared to all other genotypes), body mass index, and a history of diabetes. All variables were included in the model.
Results
Pre-treatment hepatic ultrasound reports were available on 171 patients. Seven reports were excluded because of equivocal descriptions (see Methods section, reports of "inflammation or fatty infiltration," or "steatosis vs. fibrosis" were excluded). A total of 49/164 (30%) of patients had steatosis documented in the ultrasound report. The comparison of baseline and clinical characteristics between patients with versus without steatosis on ultrasound indicated significant differences in ALT and AFP, a history of diabetes mellitus, and advanced histologic stage and grade (Table 1).
Liver biopsy specimens were available from 122 patients for our pathologists to assess the amount of steatosis. The remaining 42 patients either did not have an initial liver biopsy (n = 6) or their liver biopsy slides were not available at our institution (n = 36), as they had been returned to the patient's referring institution. The kappa statistic for interobserver reliability of steatosis between the two pathologists was calculated to be 0.87. A relationship between steatosis on ultrasound and steatosis on biopsy was detected (Figure 1). The Spearman's rank correlation coefficient was rs = 0.27 (p = 0.003, n = 122). Of the 122 patients that had both histologic review and ultrasound, 40 (33%) had steatosis noted on their ultrasound report. A substantial number of false-positive and false-negative results were observed. For example, 10% (8/82) of patients with a normal ultrasound had >30% steatosis on biopsy. Among patients with steatosis reported on ultrasound, only 12/40 (30%) had >30% steatosis on biopsy review.
We determined the test characteristics for steatosis on ultrasound (in the clinical setting of our study) compared to biopsy as the gold standard, which were calculated for the detection of any steatosis (>2%), or for significant steatosis (>30%) (Table 2). For either of these calculations, the sensitivity and specificity of ultrasonography were not high. We also examined whether or not the diagnosis of steatosis on ultrasound was more useful in detecting substantial fibrosis (Metavir stage 3–4) on biopsy. The sensitivity of steatosis on ultrasound predicting fibrosis on biopsy was 48%, while the specificity was 74%. These performance characteristics for assessing fibrosis were very similar to the effectiveness of ultrasonography to correctly identify significant (>30%) steatosis on biopsy.
A total of 136 patients were included in the multivariate logistic regression model to determine factors associated with steatosis on ultrasound (see Methods for variables included in the model). The other 28 patients were not included due to a missing variable (genotype not performed, liver biopsy not performed, etc). The only statistically significant factor associated with steatosis on ultrasound was histologic grade (OR 3.6, 95% CI 1.3–9.8). A history of diabetes mellitus approached statistical significance (OR 3.8, 95% CI 0.96–14.70). No other factors were independently associated with steatosis on ultrasound.
Discussion
The detection of steatosis on ultrasound in a clinical setting appears to be generally associated with steatosis on biopsy, but also with hepatic inflammation and fibrosis. In particular, steatosis on ultrasound was independently associated with moderate-severe histologic inflammation. The ability of ultrasonography to accurately detect hepatic steatosis is questionable outside of a controlled research setting, as both the sensitivity and specificity of this imaging technique were unacceptably low in our study. An ultrasound report of steatosis was as predictive of fibrosis as it was predictive of >30% steatosis, which reflects the unreliability of this imaging modality in differentiating fibrosis and steatosis. Based on the results of our study, the clinician should understand that an ultrasound report of steatosis could mean the patient has fibrosis, inflammation, significant steatosis, or a combination of these conditions. Alternatively, the patient may have none of these pathologic findings. Practitioners should rely on other diagnostic modalities to assess the liver for steatosis. Magnetic resonance imaging or computerized tomography are two potential techniques that may be clinically useful in the diagnosis of hepatic steatosis, pending further study [12].
Hepatic steatosis, detected by histologic examination, appears to have a multifactorial etiology in patients infected with hepatitis C. Steatosis may be a result of a direct virologic effect, particularly in patients infected with genotype 3 [1,5]. Additionally, steatosis in HCV-infected patients may be associated with accepted steatosis risk factors, including obesity [6,13,14], diabetes mellitus [6] and hypertriglyceridemia [13]. We observed that steatosis on ultrasonography was associated with factors representative of inflammation (histologic grade on biopsy and ALT) and fibrosis (histologic stage on biopsy and alpha-fetoprotein levels [15]) as well as diabetes mellitus. We did not observe an obvious association between genotype or body mass index with steatosis on ultrasound. Lack of association of genotype and BMI with steatosis in our study is likely related to factors confounding the diagnosis of steatosis on ultrasound, such as the additional presence of fibrosis and inflammation.
Increased echogenicity is the characteristic ultrasonographic finding that identifies hepatic steatosis. The increased echogenicity is often compared to the spleen and kidney [16,17]. A loss of definition of the hemi-diaphragm and decreased detail of the intrahepatic architecture (particularly the portal veins) may be supportive findings [11]. The performance characteristics of ultrasonography in the detection of steatosis vary considerably among studies. In the examination of patients with mostly alcoholic liver disease, a specificity of 84% and sensitivity of 94% for the detection of steatosis on ultrasound was described [18]. Using multiple criteria to diagnose steatosis, positive predictive values can be as high as 94% in high-prevalence populations [19]. Performance characteristics tend to improve with the diagnosis of moderate and severe steatosis [11]. One study suggested that 33% steatosis seen on biopsy was an optimal threshold for the radiographic detection of steatosis [20].
However, concomitant liver pathology may complicate the diagnosis of steatosis on ultrasound. Echogenicity on ultrasound may be consistent with either fibrosis or steatosis, and ultrasonography may not effectively differentiate between these two conditions [8]. Due to the overlap in appearance of fibrosis and steatosis, some radiologists opt to utilize the terms "fatty-fibrotic" [16,21] or "steatofibrosis" [22] when describing this echogenic pattern. Fibrosis has been demonstrated to be independently associated with steatosis in hepatitis C patients [6,14,23]. Some authors do not recommend the use of ultrasound as a screening tool for hepatic steatosis due to questions regarding sensitivity and specificity of this test [24].
Our results suggest that ultrasound is an unreliable predictor of steatosis when described on a routine ultrasound report in HCV-infected patients. These findings are consistent with a prior study of hepatic steatosis which documented no correlation between biopsy and ultrasonography in 64 patients with a positive HCV antibody [9]. Our results differed slightly from this previous report in that some correlation between ultrasound and biopsy was demonstrated (rs = 0.27). However, both false-positives and false-negatives were observed in our study. A report of steatosis was equally likely to indicate significant steatosis or fibrosis. Therefore, we conclude that hepatitis C produces significant liver pathology that may confound the diagnosis of steatosis on liver ultrasound in a clinical setting. It may be advisable for radiologists to report 'echogenic liver', 'possible steatosis vs. fibrosis' or 'bright liver', instead of definitively reporting the observation of steatosis.
For the clinician, ultrasound could conceivably be utilized in the documentation of an echogenic liver only. Similarly, any report of steatosis, fibrosis, or inflammation could be understood by the clinician as consistent with the presence of liver pathology. However, we also observed false negatives in our study, in which patients with significant steatosis had normal ultrasound reports. Additionally, the prognostic significance for clinical course and response to therapy may be very different for steatosis compared to fibrosis, and therefore the significance of an echogenic liver on ultrasound may vary substantially between patients.
Certain limitations of the study should be mentioned. The patients in this study had been referred for specialty evaluation in a hepatitis C clinic in which therapeutic clinical trials are emphasized. These patients may be different than the general population with HCV infection, limiting the generalizability of our results. Selection bias that is inherent in retrospective studies may limit the applicability of our results. Also, for some patients, there was time between the dates in which the ultrasound and biopsy were performed. It is possible that small changes in liver histology could have occurred, if the patient had significant weight gain or loss, for example. Ethanol consumption was not measured in our study, which could have impacted some of the baseline variables which were compared between patients with or without steatosis on ultrasound.
The retrospective nature of this study may introduce potential bias in data collection which could limit the clinical applicability of the findings. For example, some of the ultrasounds were not performed within the same 1–2 weeks as the liver biopsy. In addition, incorporating the radiologic interpretations of multiple radiologists (in contrast to a single radiologist) has certain advantages and disadvantages. While multiple radiologists potentially introduce significant variability in interpretation, they more accurately simulate a realistic clinical scenario as opposed to the artificial framework of a study utilizing a single expert radiologist. Since our study focused on the utility of ultrasonography results for clinicians, examining this question in the context of multiple radiologists over time seemed most appropriate. A similar review in other clinical systems, such as medical systems from other countries, would be useful. It is possible that in other systems, radiology do not attempt to define steatosis, but rather only remark on the presence and degree of an echogenic liver. It would also be interesting to survey radiologists to attempt to assess the range of techniques and criteria for the assessment of steatosis on ultrasonography. Additionally, liver biopsy itself is not always a definitive test, as sampling error can occur with this procedure. Differences in the assessment of histologic stage and grade have been observed in patients with hepatitis C who have undergone simultaneous liver biopsies in different lobes [25]. This error may account for some of the false-positive ultrasound reports in the assessment of steatosis in our study.
Conclusion
Unfortunately, routine hepatic ultrasonography does not provide an accurate non-invasive mechanism for the diagnosis of hepatic steatosis in HCV-infected patients in the clinical context of our study. These findings should be examined in other clinical settings, perhaps in other countries. Clinicians should interpret a report of steatosis on ultrasound with caution, and also consider that this report could suggest a combination of inflammation, steatosis and/or fibrosis. Even patients who did not have a liver biopsy were included in the analysis of associations with steatosis, and the description of steatosis on ultrasound was associated with factors reflective of hepatic inflammation (such as ALT). Conversely, lack of ultrasonographic evidence of steatosis does not definitively exclude the presence of steatosis as shown on biopsy. Until a non-invasive modality is proven to be superior in a clinical setting, liver biopsy remains the optimal diagnostic procedure for the determination of steatosis in patients infected with hepatitis C.
List of abbreviations
AFP Alpha-fetoprotein
ALT Alanine aminotransferase
BMI Body mass index
CI Confidence interval
HCV Hepatitis C virus
OR Odds ratio
PCR Polymerase chain reaction
RNA Ribonucleic acid
US Ultrasonography
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
MH and EL developed the initial concept of the study. All authors participated in study design and data analysis. JV and EF conducted the histologic review of the slides. MH conducted the review of ultrasound reports and organization of the data. All authors contributed to the manuscript preparation and completion.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The opinions or assertions contained herein are those of the authors and are not to be construed as official policy or as reflecting the views of the Department of the Army or the Department of Defense. The authors are employees of the U.S. Government. This work was prepared as part of their official duties and, as such, there is no copyright to be transferred. This study was presented in part as a podium presentation at Digestive Diseases Week, Orlando FL, May 19–22, 2003.
Figures and Tables
Figure 1 Comparison of steatosis on biopsy versus ultrasound. For each of the five categories of amount of steatosis on biopsy, the percentage of patients in those categories with a positive ultrasound for steatosis is displayed. rs = 0.27, p = 0.003.
Table 1 Comparison of demographic, clinical and laboratory characteristics of patients with versus without steatosis on hepatic ultrasound. Legend: ALT- alanine aminotransferase; AFP- alpha fetoprotein; BMI- body mass index.
Parameter Steatosis No Steatosis p value
Gender (% male) 31/49 (63%) 77/115 (67%) 0.68
Genotype (3 vs. others) 4/46 (9%) 5/110 (5%) 0.31
Diabetes (% with DM) 12/49 (24%) 7/115 (6%) 0.001
Age 47.9 ± 10.3 46.3 ± 10.3 0.38
ALT (IU/L) 129.5 ± 89.0 94.3 ± 87.0 0.01
AFP (ng/ml) 13.1 ± 25.5 8.0 ± 12.3 0.03
BMI (kg/m2) 28.6 ± 5.1 27.4 ± 4.8 0.17
Stage (% stage 3–4) 15/48 (31%) 16/110 (15%) 0.02
Grade (% stage 3–4)* 19/48 (40%) 17/103 (17%) 0.002
*Seven histology reports scored only stage, and did not report a histologic grade.
Table 2 Performance characteristics of the detection of steatosis on ultrasound (n = 122). PPV = positive predictive value; NPV=negative predictive value. Gold standard was steatosis on histologic examination.
Histologic Examination Sensitivity Specificity PPV NPV
Any Steatosis (>2%) 43% 79% 70% 55%
Steatosis >30% 60% 73% 30% 90%
==== Refs
Ramalho F Hepatitis C virus infection and liver steatosis Antiviral Res 2003 60 125 7 14638408 10.1016/j.antiviral.2003.08.007
Moriya K Yotsuyanagi H Shintani Y Fujie H Ishibashi K Matsuura Y Miyamura T Koike K. Hepatitis C core protein induces hepatic steatosis in transgenic mice J Gen Virol 1997 78 1527 1531 9225025
Perlemuter G Sabile A Letteron P Vona G Topilco A Chretien Y Koike K Pessayre D Chapman J Barba G Brechot C Hepatitis C virus core protein inhibits microsomal triglyceride transfer protein activity and very low density lipoprotein secretion: a model of viral-related steatosis FASEB J 2002 16 185 194 11818366 10.1096/fj.01-0396com
Shi ST Polyak SJ Tu H Taylor DR Gretch DR Lai MM Hepatitis C virus NS5A colocalizes with the core protein on lipid droplets and interacts with apolipoproteins Virology 2002 292 198 210 11878923 10.1006/viro.2001.1225
Kumar D Farrell GC Fung C George J Hepatitis C virus genotype 3 is cytopathic to hepatocytes: reversal of hepatic steatosis after sustained therapeutic response Hepatology 2002 36 1266 1272 12395339 10.1053/jhep.2002.36370
Monto A Alonzo J Watson JJ Grunfeld C Wright TL Steatosis in chronic hepatitis C: relative contributions of obesity, diabetes mellitus, and alcohol Hepatology 2002 36 729 736 12198667 10.1053/jhep.2002.35064
Bjoro K Bell H Hellum KB Skaug K Raknerud N Sandvei P Doskeland B Maeland A Lund-Tonnesen S Myrvang B Effect of combined interferon-α induction therapy and ribavirin on chronic hepatitis C virus infection: a randomized multicentre study Scand J Gastroenterol 2002 37 226 232 11843062 10.1080/003655202753416920
Fontana RJ Lok AS Noninvasive monitoring of patients with chronic hepatitis C Hepatology 2002 36 S57 S64 12407577 10.1053/jhep.2002.36800
Kutcher R Smith GS Sen F Gelman SF Mitsudo S Thung SN Reinus JF Comparison of sonograms and liver histologic findings in patients with chronic hepatitis C virus infection J Ultrasound Med 1998 17 321 325 9586705
The French METAVIR Cooperative Study Group Intraobserver and interobserver variations in liver biopsy interpretation in patients with chronic hepatitis C Hepatology 1994 20 15 20 8020885 10.1016/0270-9139(94)90128-7
Joseph AE Saverymuttu SH al-Sam S Cook MG Maxwell JD Comparison of liver histology with ultrasonography in assessing diffuse parenchymal liver disease Clin Radiol 1991 43 26 31 1999069
Saadeh S Younossi ZM Remer EM Gramlich T Ong JP Hurley M Mullen KD Cooper JN Sheridan MJ The utility of radiological imaging in nonalcoholic fatty liver disease Gastroenterology 2002 123 745 50 12198701 10.1053/gast.2002.35354
Czaja AJ Carpenter HA Santrach PJ Moore SB Host- and disease-specific factors affecting steatosis in chronic hepatitis C J Hepatol 1998 29 198 206 9722200 10.1016/S0168-8278(98)80004-4
Hourigan LF Macdonald GA Purdie D Whitehall VH Shorthouse C Clouston A Powell EE Fibrosis in chronic hepatitis C correlates significantly with body mass index and steatosis Hepatology 1999 29 1215 1219 10094967 10.1002/hep.510290401
Bayati N Silverman AL Gordon SC Serum alpha-fetoprotein levels and liver histology in patients with chronic hepatitis C Am J Gastroenterol 1998 93 2452 2456 9860408
Joy D Thava VR Scott BB Diagnosis of fatty liver disease: is biopsy necessary? Eur J Gastroenterol Hepatol 2003 15 539 543 12702913 10.1097/00042737-200305000-00014
Kurtz AB Middleton WD Ultrasound The Requisites 1996 New York: Mosby 16 18
Saverymuttu SH Joseph AE Maxwell JD Ultrasound scanning in the detection of hepatic fibrosis and steatosis Brit Med J 1986 292 13 15 3080046
Mathiesen UL Franzen LE Aselius H Resjo M Jacobsson L Foberg U Fryden A Bodemar G Increased liver echogenicity at ultrasound examination reflects degree of steatosis but not fibrosis in asymptomatic patients with mild-moderate abnormalities of liver transaminases Digest Liver Dis 2002 34 516 522 10.1016/S1590-8658(02)80111-6
Saadeh S Younossi ZM Remer EM Gramlich T Ong JP Hurley M Mullen KD Cooper JN Sheridan MJ The utility of radiological imaging in nonalcoholic fatty liver disease Gastroenterology 2002 123 745 750 12198701 10.1053/gast.2002.35354
Needleman L Kurtz AB Rifkin MD Cooper HS Pasto ME Goldberg BB Sonography of diffuse benign liver disease: accuracy of pattern recognition and grading Am J Roentgenol 1986 146 1011 1015 3515875
Debongnie JC Pauls C Fievez M Wibin E Prospective evaluation of the diagnostic accuracy of liver ultrasonography Gut 1981 22 130 135 7215943
Castera L Hezode C Roudot-Thoraval F Bastie A Zafrani ES Pawlotsky JM Dhumeaux D Worsening of steatosis is an independent factor of fibrosis progression in untreated patients with chronic hepatitis C and paired liver biopsies Gut 2003 52 288 292 12524415 10.1136/gut.52.2.288
Siegelman ES Rosen MA Imaging of hepatic steatosis Semin Liver Dis 2001 21 71 80 11296698 10.1055/s-2001-12930
Regev A Berho M Jeffers LJ Milikowski C Molina EG Pyrsopoulos NT Feng ZZ Reddy KR Schiff ER Sampling error and intraobserver variation in liver biopsy in patients with chronic HCV infection Am J Gastroenterol 2002 97 2614 2618 12385448
| 15829009 | PMC1087838 | CC BY | 2021-01-04 16:03:26 | no | BMC Gastroenterol. 2005 Apr 13; 5:14 | utf-8 | BMC Gastroenterol | 2,005 | 10.1186/1471-230X-5-14 | oa_comm |
==== Front
BMC GenetBMC Genetics1471-2156BioMed Central London 1471-2156-6-191582320310.1186/1471-2156-6-19Research ArticleAnalysis of sequence variability in the CART gene in relation to obesity in a Caucasian population Guérardel Audrey [email protected] Mouna [email protected] Francis [email protected] Christian [email protected] Vincent [email protected]ément Karine [email protected]é Delphine [email protected] Valérie [email protected] Christopher G [email protected] Pilar [email protected] Serge [email protected] Nicole [email protected] Natascha [email protected] Fritz F [email protected] Philippe [email protected] Philippe [email protected] Institute of Biology-CNRS UMR 8090, Pasteur Institute, Lille, France2 University Hospital, Lille, France3 Department of Nutrition-EA3502, Paris VI University, INSERM "Avenir" Hôtel-Dieu, Paris, France4 Scientific and Technical Institute of Nutrition and Food (ISTNA-CNAM), INSERM U557, INRA U1125, Paris, France5 Service d'Epidémiologie et de Santé Publique-INSERM U.508, Pasteur Institute, Lille, France6 Dr. Horber Adipositas Stiftung, Hornbachstrasse 50, 8034, Zürich, Switzerland7 Imperial College genome Centre and Genomic Medicine, Hammersmith Campus, Imperial College London, UK2005 11 4 2005 6 19 19 23 11 2004 11 4 2005 Copyright © 2005 Guérardel et al; licensee BioMed Central Ltd.2005Guérardel 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 and amphetamine regulated transcript (CART) is an anorectic neuropeptide located principally in hypothalamus. CART has been shown to be involved in control of feeding behavior, but a direct relationship with obesity has not been established. The aim of this study was to evaluate the effect of polymorphisms within the CART gene with regards to a possible association with obesity in a Caucasian population.
Results
Screening of the entire gene as well as a 3.7 kb region of 5' upstream sequence revealed 31 SNPs and 3 rare variants ; 14 of which were subsequently genotyped in 292 French morbidly obese subjects and 368 controls. Haplotype analysis suggested an association with obesity which was found to be mainly due to SNP-3608T>C (rs7379701) (p = 0.009). Genotyping additional cases and controls also of European Caucasian origin supported further this possible association between the CART SNP -3608T>C T allele and obesity (global p-value = 0.0005). Functional studies also suggested that the SNP -3608T>C could modulate nuclear protein binding.
Conclusion
CART SNP -3608T>C may possibly contribute to the genetic risk for obesity in the Caucasian population. However confirmation of the importance of the role of the CART gene in energy homeostasis and obesity will require investigation and replication in further populations.
==== Body
Background
Cocaine and amphetamine regulated transcript (CART) is a potent anorectic peptide that is widely expressed in hypothalamic areas and is involved in the control of feeding behavior [1,2]. Immunohistochemistry studies show that CART peptides co-localize with both anorectic and orexigenic hypothalamic peptides [3], particularly with pro-opiomelanocortin (POMC), in the arcuate nucleus (ARC) [4]. Moreover, CART peptides are distributed in peripheral organs notably in the D cells of the endocrine pancreas [5,6] and throughout the peripheral nervous system within the vagal efferent neurons, where it interacts with cholecystokinin [7]. Leptin regulates CART mRNA expression, since it is reduced in the arcuate nucleus by the disruption of leptin signaling (in ob/ob mice or fa/fa rats) and increased by leptin peripheral injections [8]. Recent studies have shown that in the context of a high fat diet there is a close relationship between CART and leptin which facilitates the regulation of lipid metabolism in order to control body fat [9]. In rodents, intracerebroventricular injection (ICV) of CART peptide fragments inhibits feeding and antagonizes the feeding response induced by the orexigenic neuropeptide Y (NPY), whereas ICV injection of CART antiserum is found to stimulate feeding [3]. However, instead of the expected hyperphagic phenotype, CART-deficient mice are predisposed to obesity only when fed with a calorie-rich diet [10]. Although CART has therefore been shown to be involved in control of feeding behavior, a direct relationship with obesity has not yet been established.
Earlier studies have failed to detect an association between exonic CART gene single nucleotide polymorphisms (SNPs) and obesity or obesity-related phenotypes [11-13]. Recently, however, a putative CART promoter SNP (-156A>G) has been reported as having a possible association with obesity in a Japanese population [14]. Interestingly, CART maps to chromosome 5q13.2, 4.8 Mb from the D5S647 locus, a region previously linked to obesity and serum leptin levels in obese French Caucasian families [15]. CART is located 61.8 kb downstream the MCCC2 gene (methylcrotonoyl-Coenzyme A carboxylase 2) and is distant to 386 Kb from the MAP1B gene (microtubule-associated protein 1B). Only CART gene is recognized as candidate for obesity. Recent data have suggested that restricting SNPs analysis to the coding regions only does not adequately describe all the common haplotypes or the true haplotype block structure observed when all of the common variations within the genetic region are used to infer haplotypes [16]. Thus, in this study's investigation of the genetic contribution of CART to obesity a region that included the three exons, as well as the two introns and the 5' region from the first ATG to 3.7 kb upstream was screened for SNPs. The initial case-control study was performed in 292 morbidly obese French subjects (BMI ≥ 40 kg/m2) and 368 non-obese and normoglycemic controls. In a subsequent extension in the investigation the previously associated SNPs were genotyped in additional sample sets of cases and controls and a haplotype analysis was performed to identify potentially functional SNPs.
Results
Initial case-control study
To investigate variation within CART, a genetic region of 5.4 kb was sequenced in a total of forty-five subjects (39 obese and six non-obese individuals). A total of thirty-four SNPs were identified (Figure 1) [see Additional file 1]. Three SNPs : a missense Glu32Lys mutation (+94G>A) in exon 1, as well as IVS1+114C>T and IVS1-31C>T both of which resided in intron 1, presented with a frequency lower than 1% (data not shown). None of these rare SNPs were found to co-segregate with obesity or obesity related phenotypes in the probands' families (data not shown). Using the analysis of these thirty-one frequent SNPs (frequency>3%), in this group of forty-five subjects, the linkage disequilibrium (LD) was calculated using the GOLD program (Figure 2). Thus, among the thirty-one SNPs, fourteen SNPs were found to be non-redundant and were subsequently genotyped in 292 morbidly obese subjects and 368 controls. In the 5' region, three SNPs (-3608T>C, -1702C>T and -175A>G) were found to be significantly associated with obesity (p = 0.001; p = 0.0015; p = 0.002 ; see Table 1). A weak association was observed for the 3'UTR SNP +1343delA (p = 0.02) and for the 5' SNP-3607C>T (p = 0.046). The 14 SNPs were consistent with the Hardy-Weinberg equilibrium (HWE) in the obese group, although in controls, SNP-3608T>C, SNP-3607C>T, SNP-1702C>T, SNP-175A>G deviated from HWE (p = 0.0017; p = 0.002; p = 0.0016 and p = 0.0017 respectively). In addition, allele frequency differences between obese and non-obese subjects were confirmed with the Armitage's trend test [17], which doesn't rely on the Hardy-Weinberg equilibrium hypothesis. Nevertheless the genotyping accuracy for these SNPs was subsequently confirmed by duplicate direct sequencing in all those individuals.
Figure 1 SNPs map of the CART gene. SNPs are reported assigning +1 to the A of the ATG start codon. The numbering reported in the literature is given in parentheses.
Figure 2 Pairwise linkage disequilibrium between the 31 SNPs of CART gene in 45 subjects. The LD was measured by delta (color scale) using the Gold software. SNPs are indicated along the horizontal and vertical axes, according to the first transcribed nucleotide position. Underlined SNPs were selected for the initial case-control study. Regions of high and low linkage disequilibrium are represented by red and blue shading, respectively.
Table 1 Alleles and genotype distributions for five CART SNPs in 368 French non-obese group and in different French morbid obese groups.
SNPs allelesa p-valueb OR [95%CI] genotypesa p-valuec
-3608T>C T C TT TC CC
French non-obese subjects 311 (42.3) 425 (57.7) 51 (13.9) 209 (56.8) 108 (29.3)
1st set of French morbid obese subjects 299 (51.4) 283 (48.6) 0.001 1.44 [1.16–1.80] 78 (26.8) 143 (49.1) 70 (24.1) 0.0006
2nd set of French morbid obese subjects 287 (47.5) 317 (52.5) 0.05 1.24 [1.00–1.54] 65 (21.5) 157 (52.0) 80 (26.5) 0.04
Pooled set of French morbid obese subjects* 586 (49.4) 600 (50.6) 0.002 1.33 [1.11–1.61] 143 (24.1) 300 (50.6) 150 (25.3) 0.001
-3607C>T C T CC CT TT
French non-obese subjects 392 (53.3) 344 (46.7) 90 (24.5) 212 (57.6) 66 (17.9)
1st set of French morbid obese subjects 342 (58.8) 240 (41.2) 0.046 1.33 [1.11–1.61] 102 (35.1) 138 (47.4) 51 (17.5) 0.037
2nd set of French morbid obese subjects 346 (57.1) 260 (42.9) 0.16 1.17 [0.94–1.45] 100 (33.0) 146 (48.2) 57 (18.8) 0.14
Pooled set of French morbid obese subjects 688 (57.9) 500 (42.1) 0.046 1.21 [1.004–1.45] 202 (34.0) 284 (47.8) 108 (18.2) 0.04
-1702C>T C T CC CT TT
French non-obese subjects 438 (63.3) 254 (36.7) 125 (36.1) 188 (54.3) 33 (9.6)
1st set of French morbid obese subjects 306 (54.4) 256 (45.6) 0.0015 1.44 [1.15–1.81] 82 (29.2) 142 (50.5) 57 (20.3) 0.0009
2nd set of French morbid obese subjects 345 (57.7) 253 (42.3) 0.04 1.27 [1.01–1.58] 94 (31.4) 157 (52.5) 48 (16.1) 0.028
Pooled set of French morbid obese subjects 651 (56.1) 509 (43.9) 0.002 1.35 [1.11–1.64] 176 (30.3) 299 (51.6) 105 (18.1) 0.0015
-175A>G A G AA AG GG
French non-obese subjects 294 (42.4) 400 (57.6) 48 (13.8) 198 (57.1) 101 (29.1)
1st set of French morbid obese subjects 279 (51.1) 267 (48.9) 0.0022 1.42 [1.13–1.78] 72 (26.4) 135 (49.4) 66 (24.2) 0.0014
2nd set of French morbid obese subjects 289 (46.8) 329 (53.2) 0.10 1.19 [0.96–1.49] 69 (22.3) 151 (48.9) 89 (28.8) 0.095
Pooled set of French morbid obese subjects 568 (48.8) 596 (51.2) 0.007 1.30 [1.07–1.57] 141 (24.2) 286 (49.1) 155 (26.7) 0.006
+1343delA A delA AA AdelA delAdelA
French non-obese subjects 656 (92.7) 52 (7.3) 304 (85.9) 48 (13.6) 2 (0.5)
1st set of French morbid obese subjects 520 (89.1) 64 (10.9) 0.02 1.53 [1.06–2.28] 232 (79.4) 56 (19.2) 4 (1.4) 0.02
2nd set of French morbid obese subjects 590 (89.6) 68 (10.4) 0.05 1.45 [1.00–2.12] 263 (80.0) 64 (19.4) 2 (0.6) 0.048
Pooled set of French morbid obese subjects 1110 (89.4) 132 (10.6) 0.017 1.50 [1.07–2.10] 495 (79.7) 120 (19.3) 6 (1.0) 0.017
aNumber of alleles or genotypes (frequencies, %)
b p-value corresponds to association analysis between different French morbidly obese groups (BMI ≥ 40 kg/m2) and non-obese subjects (BMI<27 kg/m2).
c p-value is calculated by the Armitage's trend test.
"Risk" alleles underlined are those encountered more frequently in obese groups.
* Pooled set of 621 French morbidly obese subjects were composed of the first set of 292 French morbidly obese patients (BMI = 47.5 ± 7.8 kg/m2, age = 44.7 ± 10.8 years) and of a second set of 329 French morbidly obese patients (BMI = 48.1 ± 7.1 kg/m2, age = 48.5 ± 10.5 years).
The genotyping of the five SNPs potentially associated with obesity (5' SNPs : -3608T>C, -3607C>T, -1702C>T, -175A>G and 3'UTR SNP: +1343delA) was then extended to an additional set of 329 morbidly obese subjects. This second French population was not significantly different in mean age or BMI in comparison to the first and therefore enabled pooling of the two case cohorts and the comparison of the SNP allele frequencies between a total of 621 morbidly obese subjects and 368 controls. The -3608T, the -3607C, the -1702T, the -175A and the +1343delA alleles were still more frequent in the morbid obese subjects than in the controls : OR = 1.33, 95%CI = [1.11–1.61], p = 0.002 ; OR = 1.21, 95%CI = [1.004–1.45], p = 0.046 ; OR = 1.35, 95%CI = [1.11–1.64], p = 0.002 ; OR = 1.30, 95%CI = [1.07–1.57], p = 0.007 and OR = 1.50, 95%CI = [1.07–2.10], p = 0.017 respectively, see Table 1). The results with the most significant p-value of 0.002, which was observed for the SNPs -3608T>C and -1702C>T, was only reached seven times on 1000 permutations, confirming the strength of this result, even though multiple testing was performed.
Haplotype analysis
The fourteen SNPs were further submitted to haplotype analysis. Rare haplotypes (frequencies < 0.02) were removed. Considering 94.4 % of the existing haplotypes, seven SNPs were identified, which are structured in two combinations of six SNPs (SNPs -3608T>C/ -1705C>T/ -1702C>T/ -1474T>C/ -271C>T/ +1343delA) and (-175A>G/ -1705C>T/ -1702C>T/ -1474T>C/ -271C>T/ +1343delA) represented the haplotype information the best. Haplotype frequencies were compared between the 292 morbidly obese subjects and the 368 controls using the THESIAS (Testing Haplotype Effects in Association Studies) software [18]. Both combinations displayed association with obesity (p = 0.0003). Then, investigating the effect of individual or group of SNPs on the general haplotypic model, it was observed that the significant effect on obesity was mainly due to i) two combinations of promoter SNPs : (-3608T>C/ -1702C>T) and (-1702C>T/ -175A>G) (p = 0.0002 for both combinations) and ii) the presence of the SNP+1343delA (p = 0.002). The two combinations (-3608T>C/ -1702C>T/ +1343delA and -1702C>T/ -175A>G/ +1343delA) were analyzed in the 621 morbidly obese subjects and in the 368 controls and were found to show an association with obesity (global p = 6.10-5). The haplotypes including SNPs -3608T (or -175A), -1702T, +1343A alleles or -3608C (or -175G), -1702C, +1343delA were more frequent in obese subjects than in controls (10.9% vs. 7.2%, OR = 1.91, 95%CI = [1.18–3.12], p = 0.009; 44.0% vs. 35.6%, OR = 1.60, 95%CI = [1.26–2.03], p = 0.0001 respectively).
These analyses were then implemented in the COCAPHASE program to exclude putative errors in haplotype determination induced by the Hardy Weinberg disequilibrium observed in the controls, which confirmed the results obtained with THESIAS (data not shown). Altogether, these data highlighted the role of SNPs -3608T>C (or -175A>G), and -1702C>T combined to SNP+1343delA in the SNP haplotype structure associated with obesity in French Caucasians.
Study of SNPs -3608T>C, -1702C>T,-175A>G and +1343delA in a French general population
In order to support further our initial findings, we analyzed an independent control group consisting of 732 non obese subjects ascertained from the general French population. All SNPs did not deviate from HWE in this group. When this control group was compared to the 621 French morbid obese subjects, SNP-3608T>C showed a nominal association with morbid obesity (OR = 1.18, 95%CI = [1.01–1.39], p = 0.038), although the SNPs -1702C>T, -175A>G and +1343delA were not found to be associated with obesity (data not shown). The prevalence of the SNP -3608C>T T allele was 49.4% in morbidly obese subjects, 45.2% in controls derived from the French general population and 42.3% in the non obese and normoglycemic subjects (Figure 3).
Figure 3 Histogram summarized the -3608T allele frequency in different groups of controls and obese subjects. Compared to French controls from type 2 Diabetes pedigrees and French controls from general population ; -3608T allele frequency was more frequent in the French moderate obese subjects (p = 0.011 and p = 0.14, respectively) and in the Swiss morbidly obese subjects (p = 0.002 and p = 0.078, respectively).
Study of SNP-3608T>C in morbid Swiss and moderate French obese subjects
To further investigate the impact of the 3608T>C CART SNP on obesity in populations of European origin, we extended the study to 619 additional French subjects with moderate obesity (30 kg/m2<BMI<40 kg/m2) and to 385 morbidly obese Swiss subjects (BMI>40 kg/m2). As shown in Figure 3, no significant difference in the T allele frequency was found between the two obese groups or with the initial 621 morbidly obese subjects, thus confirming the high prevalence of this allele in association with obesity. When the three obese group (n = 1,625) and the two control groups (n = 1,100) were analyzed together, an association with obesity and the CART 3608T>C T allele was confirmed (OR = 1.22, 95%CI = [1.09–1.37], p = 0.0005).
Electrophoretic mobility shift assays (EMSAs) analysis
Therefore genetic studies suggested an association between CART SNP-3608C>T and obesity. These data then promoted the investigation of this polymorphism's potential functional effect. Potential effects of plausible promoter SNP -3608T>C on binding affinity for nuclear proteins was evaluated by electrophoretic mobility shift assays (EMSA) in RIN-1027-B2 cells. This cell line was chosen because CART expression has been localized in the somatostatin producing islet D pancreatic cells. As SNP -3607C>T is adjacent to SNP -3608T>C, two pairs of DNA oligonucleotide probes were required. In the presence of -3607C, RIN-1027-B2 nuclear proteins (Figure 4, band 2) have a lower affinity for the C allele than the T allele of SNP-3608T>C (Figure 4, lane 6), indicating that the presence of both the -3608T and the -3607C alleles are required for binding. With at least one C allele present for SNP -3608T>C or SNP -3607C>T, a binding (band 1) was observed and this was found to be then decreased two-fold in the -3608T-3607T configuration (Figure 4, lane 9). To identify which SNP loci correspond to which band, homologous and heterologous competition with non-labeled probes were used in EMSA (Figure 4, lanes 3, 4, 7, 8, 10, 11, 13, 14). The addition of either competitor -3608T or -3608C induced a decrease of the two signals for all configurations.
Figure 4 Electrophoretic mobility shift assays for SNP-3608T>C with nuclear extracts prepared from RIN-1027-B2 cells. Four probes used for EMSA contained respectively -3608T-3607C, -3608C-3607C, -3608T-3607T and -3608C-3607T. Specific complex formations (compared lines 2 and 5) are indicated by two arrows. Lane 1 is radiolabeled probes without nuclear extract. In the presence of the -3608T allele and the -3607C allele, two bands (bands 1 and 2) were observed corresponding to the fixation of two different factors (lane 2). The intensity of the band 2 decreased when either of these two alleles was changed (lanes 6, 9 and 12 compared to lane 2). The band 2 decreased by 1.5 fold in a TT or a CT configuration (lanes 9 and 12) and completely disappeared in a CC configuration at both loci (lane 6). The intensity of the band 1 and 2 was decreased when non radioactive competitors (either C or T alleles at the -3608 locus) were added to the reaction (lanes 3 and 4 compared to lane 2). In the TT configuration at the two loci, the addition of the -3608C non labeled probe did not decrease the band 2 signal unlike the -3608T probe (lanes 10 and 11 compared to lane 9). These observations suggest that the band 2 corresponds to the binding of a putative nuclear protein when both the -3608T and the -3607C alleles are present. The band 1 was observed when at least one C allele was present at either locus, decreased by 2 fold in a TT configuration (lanes 2, 6, 12 compared to lane 9).
In silico analysis suggested that allele -3608T could be located within at least three putative consensus sequences for transcription factor binding: CHOP/GADD153 (C/EBP homologous protein), GATA-3 and OCT-1. Therefore, the presence of these transcription factors in nuclear extracts from RIN-1027-B2 cells was ascertained by performing western blot analysis with specific antibodies (Figure 5). As expression of CHOP was increased by tunicamycin [19], EMSA experiments were performed on nuclear extracts of treated cells. Results were similar to those presented in Figure 4 (data not shown). Then super-shift assays were performed with anti-GATA-3, anti-OCT-1 antibodies on nuclear extracts from untreated RIN-1027-B2 cells. For anti-CHOP antibody, nuclear extracts were used from tunicamycin-treated and untreated RIN-1027-B2 cells. In the presence of antibodies the patterns of the super-shift assays did not differ from those of the shift assays described in Figure 4 (four replicates, data not shown). These results indicate that none of these three transcription factors seemed to be involved in the constitution of nuclear proteins – DNA complexes including the -3608 locus observed in the EMSA experiments.
Figure 5 Expression of nuclear proteins in untreated and tunicamycin-treated RIN-1027-B2 cells. GATA-3 and OCT-1 were specifically detected in RIN-untreated cells (lanes 3 and 4). For CHOP, a multiple band pattern was observed (lane 1). As tunicamycin induces endoplasmic stress in cells and consequently enhances CHOP expression, a western blot was performed with tunicamycin treated RIN-1027-B2 cells to determine which band corresponds to CHOP. As pointed by the arrow, the intensity of the band of 27 KDa increased in nuclear extracts from RIN-treated cells, indicating that the amount of CHOP nuclear protein was increased (lane 2 compared to lane 1).
Discussion
This is the first extensive study of the CART gene that includes an extensive analysis of the 5' region of the gene (3.7 kb upstream) in association with human obesity. Indeed, Dominguez et al. characterized the region encompassing 3.4 kb of 5' upstream sequence in the mouse CART gene and identified several regulation sites for CART mRNA levels within this region [20]. CART is very polymorphic in human : we found thirty-one common SNPs and three rare variants (Figure 1). Of these, the +1343delA (ΔA1457) and +1361A>G (A1475G) located in the 3'UTR have been previously reported in Danish subjects without association with obesity [11]. The +1361A>G (A1475G) SNP was associated with a lower waist-to-hip ratio in British Caucasian non-obese men from the Isle of Ely type 2 diabetes cohort [12], but in our French population showed no association with either obesity or obesity-related phenotypes. In contrast to Fu et al [21], no association between any CART SNPs and obesity-related quantitative traits, like lipid levels, was detected in our obese population (data not shown). Six SNPs in the 5'flanking sequence covering a region of 1.1 Kb have been reported in Japanese obese subjects [14]. Instead of the -175G (-156G) allele being associated with obesity in the Japanese obese population, the -175A (-156A) wild type allele was more prevalent in obese than in non-obese French subjects. This discordant result between these two populations might be explained by ethnic differences or a particular environmental influence. Alternatively, it may only show that this SNP is not functional, as suggested by the EMSA data that showed no protein-binding at this site (data not shown). Nevertheless, discordance in allelic frequency between Caucasian and Japanese populations has been previously observed for other diabesity susceptibility genes, such as adiponectin/APM1 [22].
Our study of the genetic variability of the CART gene in various obese and control populations of European origin suggests a possible association with obesity which is mainly due to the effect of SNP -3608 T>C. We are aware that the existence of Hardy-Weinberg disequilibrium (in the first cohort of 368 French controls only) may lead to an incorrect estimation of haplotype frequencies and thus of the haplotype effect on obesity. Therefore we used haplotype reconstruction algorithms that rely on the hypothesis of random gamete association minimizing the possible errors induced by absence of HWE. The deviation from Hardy Weinberg equilibrium observed may have several explanations. A random chance may account for deviation from HWE in 1 out of 20 markers (5%) [23]. In these present data, 50% of genotyped markers from the 5' sequences of CART deviated from HWE thereby excluding this hypothesis. The full replication of the initial SNP genotypes by direct sequencing makes significant genotyping errors unlikely. This deviation is also unlikely to be due to hidden population structure since an excess of heterozygote subjects was observed, whereas the effect of population structure usually increases homozygote frequency in the whole population [24]. To test whether departure from HWE is specific to CART, we reanalyzed previous genotyping data obtained in the 368 controls for eight genes, some of them associated with obesity or associated phenotypes such as UCP3, APM1, PGC1 and GAD2 genes. No deviation from HWE was observed for any of the 27 SNPs from these genes [[22,25,26] and unpublished data]. This would suggest a specific deviation of HWE in the CART gene in the non-obese subjects of set 1. Deviation from HWE may also be interpreted as a signal of association [23,27]. Unlike the 732 controls who are representative of the French general population, the initial cohort of 368 non-obese individuals resulted from a selection of non-obese normoglycemic subjects chosen from French type 2 diabetes pedigrees. These subjects live in families with type 2 diabetes where obesity is frequent, and may therefore represent subjects more "resistant" to both type 2 diabetes and obesity in spite of a shared common environment with their diabetic and/or obese family members. The more significant association of SNPs -3608T>C with obesity in this population compared to the more general control population advocates the interest of such a sample for an association study [28]. In addition, the more robust association with morbid obesity highlights that the probability to detect genetic mutations influencing BMI is higher in an obese population with a more extreme phenotype.
EMSA experiments on SNP -3608T>C showed a functional difference between the two alleles at the locus where DNA variation modulates binding affinity in RIN-1027-B2 pancreatic cells. However, we cannot clearly ascertain the functional role of the SNP -3608T>C, since the nucleotide at position -3607 seems to interact with the binding of nuclear factors. Moreover, haplotype analysis comparing the 368 French controls and the 621 French morbidly obese subjects, showed that the -3608T/-3607C combination is associated with obesity (p = 0.0013). Therefore, from our genetic and functional results, we hypothesize that these two SNPs, located one base pair apart, jointly modulate the binding of the same nuclear factor on CART gene 5' sequences. Nuclear factors involved in the protein DNA complexes remain to be identified. It also must be considered that functional study in this cell line may or may not reflect all relevant conditions in brain. So, cells from hypothalamic neurons would allow confirmation of this observation and a better understanding of plausible gene regulation.
Conclusion
This result concludes a possible modest contribution of variation in the CART gene to the genetic susceptibility to obesity. Replication in further populations will be required to further strengthen this association.
Methods
Subjects
According to the previously reported linkage results at the 5q locus [15], we selected 45 subjects (39 obese and six non-obese subjects) in families that contributed to the linkage with obesity. The association study was performed using at first a set of 368 non-obese and normoglycemic subjects ([mean ± SD] age, 57.1 ± 13.5 years, BMI = 22.9 ± 2.3 kg/m2, sex ratio : women/men 221/147) and a set of 292 morbidly obese French individuals (age, 44.7 ± 10.8 years, range 24 to 74, BMI = 47.5 ± 7.8 kg/m2, sex ratio : women/men 232/60). SNPs showing a significant association were genotyped in an extended set of 329 morbidly obese French subjects (age, 48.5 ± 10.5 years, range 24 to 74, BMI = 48.1± 7.1 kg/m2, sex ratio : women/men 229/100) from the same population as the first set of morbidly French obese subjects were extracted from. The 385 severely obese subjects from Zurich, Switzerland that were studied were consecutive, unrelated, Caucasian subjects ([mean ± SEM] age, 43.5 ± 0.5 years, range 24 to 69; BMI = 43.4 ± 0.4 kg/m2; sex ratio : women/men 302/83) referred to the ? clinic for refractory obesity from January 1999 to December 2000; informed written consent was obtained [29]. The set of 619 moderately obese French subjects were characterized as following: age 49.9 ± 13.4 years, range 24 to 88, BMI = 34.5 ± 2.9 kg/m2, sex ratio: women/men 354/265. Two additional sets of control subjects from the French general populations were studied : 546 subjects from the SUVIMAX population [30] ([mean ± SD] age 55 ± 6 years, BMI = 22 ± 1.8 kg/m2, sex ratio women/men 246/300) and 186 subjects from the WHO-MONICA Lille population [31] ([mean ± SD] age, 60.6 ± 3.1 years, BMI = 24.7 ± 2.9 kg/m2, sex ratio : women/men 98/88). Further analysis was performed by pooling these data.
Screening and SNP map of the CART gene
We screened for SNPs in 3.7 kb of the plausible promoter region, as well as the exons and the introns of the CART gene by direct sequencing. PCR primers and annealing temperatures are available from authors on request. The protocol was carried out using the 96 capillary ABI PRISM® 3700 DNA Analyzer (Applied Biosystems, Foster City, CA) with the Big Dye Terminator Cycle Sequencing Ready Reaction Kit, as previously described [32]. SNP positions were assigned according to the A of the translation initiation codon (ATG ; Figure 1). The list of identified SNPs and corresponding rs numbers are presented in the additional file (Table 2).
Genotyping
Several genotyping methods were used. E32K (+94G>A), IVS1+114C>T, IVS1+224G>A and IVS1-31C>T were genotyped with PCR-RFLP using the Sac1, Blp1, Apa1 and Mae3 restriction enzymes respectively (New England Biolabs). Promoter fragments containing more than two SNPs were genotyped by direct sequencing. All other SNPs were genotyped with the LightCycler™ assay (Roche, Mannheim, Germany) based on hybridization of probes labeled by two different dyes allowing Fluorescence Resonance Energy Transfer (FRET) [33]. A genotyping quality control was performed by introducing duplicates in the PCR plates and by genotyping all individuals twice. Sequences of primers and conditions of LightCycler assays are available on request.
LD analysis
Pairwise Δ (correlation coefficient between identified CART SNPs) was estimated from genotypes for the 45 subjects and the results were visualized by the GOLD program .
Statistical analysis for association studies
Hardy-Weinberg proportions for cases and controls were tested by the χ2 test. Allelic and genotypic frequencies differences between cases and controls were assessed by χ2. A region-wide empirical p-value was calculated through permutation. This involved the individual genotype as a whole and the individual's status being shuffled. This method preserves the correlation between SNPs (LD) while breaking the relation between status and the genotypes. For each replicate or permutation each SNP was tested for association and the most significant p-value was stored. We could then compare this p-value to the best observed p-value.
For the haplotype analysis, identification of the minimal set of SNPs that could account for the genotypic diversity was made by systematic enumeration in each block. Haplotypes frequencies were calculated and, after skipping the haplotypes with a frequency lower than 0.02, each SNP and set of SNPs in turn were removed. Thus we found the SNP combination that preserves the marginal haplotype frequencies. This method is implemented in the STRATEGY software. Effects of haplotype were tested using the THESIAS (Testing Haplotype Effects in Association Studies) software. The objective of the program is to perform haplotype-based association analysis in unrelated individuals. This program is based on the maximum likelihood model described in Tregouet et al. (2002) and is linked to the SEM algorithm [18]. The effect of haplotypes with a frequency lower than 1% was not included in the analysis. THESIAS allows the simultaneous estimation of haplotype frequencies and of their effects on the phenotype of interest. It is possible to get the log-likelihood of the data under a specific hypothesis concerning haplotype effects by setting some appropriate constraints on regression parameters. The notation β(h) will refer to the regression parameter characterizing the effect of haplotype h. This option is useful for testing for the equality of haplotype effects and to observe the SNP effects on a haplotype. For example, to test the effect of the second SNP among the four existent haplotypes, we could note two equations β(11) = β(12) and β(21) = β(22). If the difference between global likelihood and likelihood for tested SNP is significant, then the SNP tested had an important role in the haplotype combination. Significance of the model was confirmed through permutation with disease. As the Hardy Weinberg disequilibrium observed in controls could induce errors for haplotype analysis, we tested the robustness of this analysis. As a result, the permutation of status among individuals allowed us to confirm the significance of the result. Secondly, the analysis using the UNPHASED/COCAPHASE program [34] on the individuals having unambiguous haplotypes, which does not rely on Hardy Weinberg equilibrium hypotheses was carried out and confirmed a strong association.
Cell line and treatment
Rat islet somatostatin-producing RIN-1027-B2 cells were grown in Dulbecco's modified Eagles medium (DMEM) supplemented with 10% fetal bovine serum, penicillin (10 U/ml), streptomycin (10 mg/ml) and incubated at 37°C under a 5% CO2 atmosphere. To analyse CHOP DNA-binding activity in stressed cells, a set of confluent cells was treated for six hours with 2 μg/ml tunicamycin (Sigma), as previously described [35]. Nuclear extracts were prepared as previously described [36].
Electrophoretic Mobility Shift Assay (EMSA) experiment and Western blot
For EMSA, protein concentrations were determined by the Bio-Rad protein assay with bovine serum albumin as a standard. Double-stranded DNA probes of 23 bp (forward strand : 5'-gctcactgcaaT/Cctctgccctgc-3') containing the -3608T>C polymorphisms were labeled with T4 polynucleotide kinase using [γ32P]ATP, purified with the mini-Quick Spin Columns system (Roche Applied Science, Basel, Switzerland). The labelled probes had a specific activity of ~1 × 106 cpm/pmol DNA. Binding reactions were performed at least three times to replicate results and in the presence of homologous, heterologous and unrelated control competitors They were carried out in a total volume of 20 μl, containing 35000 cpm of radio labeled probe, 10 μg of nuclear extracts, 2 μg of poly (dI-dC) and a buffer with 20 mM potassium phosphate (pH 7.9), 70 mM KCL, 1 mM DTT, 0.3 mM EDTA and 10% glycerol. Gels were exposed to X-ray films (Kodak, Rochester, New York, United States). Quantitation of the label was performed with the NIH Image software . Nuclear extracts were subjected to SDS-Page, Western blotting and immunolabeling using rabbit anti-CHOP (R-20), rabbit anti-GATA-3 (H-18) or rabbit anti-OCT-1 (C-21) polyclonal antibodies (200 μg/ml) from Santa Cruz Biotechnology, Inc., California, U.S.A. When EMSA was performed with antibodies, the binding mix (listed above) was incubated with antibody for 30 min; where after the radiolabeled probe was incubated for 30 min.
List of abbreviations
ARC, arcuate nucleus; CART, cocaine and amphetamine regulated transcript; CHOP/GADD153 (C/EBP homologous protein 10); CI, confidence interval; GATA-3, GATA-binding protein 3; OCT-1, octamer-binding transcription factor 1 ; EMSA, electrophoretic mobility shift assay; FRET, fluorescence resonance energy transfer; HWE, Hardy-Weinberg equilibrium; ICV, intracerebroventricular injection; LD, linkage disequilibrium; NPY, neuropeptide Y; OR, odds ratio; POMC, pro-opiomelanocortin; SNPs, single nucleotide polymorphisms.
Authors' contributions
Audrey Guérardel and Mouna Barat-Houari have screened the CART gene for SNPs, have selected the subsequent SNPs for genotyping and genotyped these in the different populations, with the technical help of Vincent Vatin. They have performed most of the statistical analyses (with the technical help of Valérie Vasseur-Delannoy, Francis Vasseur and Nicole Helbecque) under the supervision of Christian Dina. Christopher Bell provided scientific editing of the manuscript. Philippe Froguel has directed the study and Philippe Boutin and Philippe Froguel have directed the redaction of the paper. DNA was provided by Karine Clément, Delphine Eberlé, Pilar Galan, Serge Hercberg, Natascha Potoczna, Fritz F. Horber and Nicole Helbecque.
Supplementary Material
Additional File 1
List of identified SNPs and rs numbers (UCSC Genome Browser on Human May 2004 Assembly).
Click here for file
Acknowledgements
The authors wish to thank Frédéric Allegaert, Yamina Benmezroua, Marianne Deweirder, Annie LeGall, Geneviève Bonhomme (EA 3502, Paris) and Céline Wahl for their technical assistance, and Karine Seron for her help concerning functional experiments. In addition, we thank JF. Habener's team for RIN-1027-B2 cells. This research was supported in part by Pasteur Institute Lille and Region Nord-Pas de Calais with regards Audrey Guérardel and in part by the European Community Project NUGENOB (QLRT-CT-2000-00618, ). The recruitment of French morbidly obese subjects was supported by the Direction de la Recherche Clinique/Assistance Publique-Hopitaux de Paris, the Programme Hospitalier de Recherche Clinique (AOM 96088).
==== Refs
Vrang N Tang-Christensen M Larsen PJ Kristensen P Recombinant CART peptide induces c-Fos expression in central areas involved in control of feeding behaviour Brain Res 1999 818 499 509 10082837 10.1016/S0006-8993(98)01349-3
Beaudry G Zekki H Rouillard C Levesque D Clozapine and dopamine D3 receptor antisense reduce cocaine- and amphetamine-regulated transcript expression in the rat nucleus accumbens shell Synapse 2004 51 233 240 14696011 10.1002/syn.10302
Lambert PD Couceyro PR McGirr KM Dall Vechia SE Smith Y Kuhar MJ CART peptides in the central control of feeding and interactions with neuropeptide Y Synapse 1998 29 293 298 9661247 10.1002/(SICI)1098-2396(199808)29:4<293::AID-SYN1>3.0.CO;2-0
Elmquist JK Elias CF Saper CB From lesions to leptin: hypothalamic control of food intake and body weight Neuron 1999 22 221 232 10069329 10.1016/S0896-6273(00)81084-3
Koylu EO Couceyro PR Lambert PD Ling NC DeSouza EB Kuhar MJ Immunohistochemical localization of novel CART peptides in rat hypothalamus pituitary and adrenal gland J Neuroendocrinol 1997 9 823 833 9419833 10.1046/j.1365-2826.1997.00651.x
Jensen PB Kristensen P Clausen JT Judge ME Hastrup S Thim L Wulff BS Foged C Jensen J Holst JJ Madsen OD The hypothalamic satiety peptide CART is expressed in anorectic and non- anorectic pancreatic islet tumors and in the normal islet of Langerhans FEBS Lett 1999 447 139 143 10214934 10.1016/S0014-5793(99)00291-4
Broberger C Holmberg K Kuhar MJ Hokfelt T Cocaine- and amphetamine-regulated transcript in the rat vagus nerve: A putative mediator of cholecystokinin-induced satiety Proc Natl Acad Sci U S A 1999 96 13506 13511 10557351 10.1073/pnas.96.23.13506
Kristensen P Judge ME Thim L Ribel U Christjansen KN Wulff BS Clausen JT Jensen PB Madsen OD Vrang N Larsen PJ Hastrup S Hypothalamic CART is a new anorectic peptide regulated by leptin Nature 1998 393 72 76 9590691 10.1038/29993
Wortley KE Chang GQ Davydova Z Fried SK Leibowitz SF "Cocaine- and amphetamine-regulated transcript in the arcuate nucleus stimulates lipid metabolism to control body fat accrual on a high-fat diet." Regul Pept 2004 117 89 99 14700744 10.1016/j.regpep.2003.08.005
Asnicar MA Smith DP Yang DD Heiman ML Fox N Chen YF Hsiung HM Koster A Absence of cocaine- and amphetamine-regulated transcript results in obesity in mice fed a high caloric diet Endocrinology 2001 142 4394 4400 11564703 10.1210/en.142.10.4394
Echwald SM Sorensen TI Andersen T Hansen C Tommerup N Pedersen O Sequence variants in the human cocaine and amphetamine-regulated transcript (CART) gene in subjects with early onset obesity Obes Res 1999 7 532 536 10574510
Challis BG Yeo GS Farooqi IS Luan J Aminian S Halsall DJ Keogh JM Wareham NJ O'Rahilly S The CART gene and human obesity: mutational analysis and population genetics Diabetes 2000 49 872 875 10905499
Walder K Morris C Ravussin E A polymorphism in the gene encoding CART is not associated with obesity in Pima Indians Int J Obes Relat Metab Disord 2000 24 520 521 10805512 10.1038/sj.ijo.0801196
Yamada K Yuan X Otabe S Koyanagi A Koyama W Makita Z Sequencing of the putative promoter region of the cocaine- and amphetamine-regulated-transcript gene and identification of polymorphic sites associated with obesity Int J Obes Relat Metab Disord 2002 26 132 136 11791158 10.1038/sj.ijo.0801848
Hager J Dina C Francke S Dubois S Houari M Vatin V Vaillant E Lorentz N Basdevant A Clement K Guy-Grand B Froguel P A genome-wide scan for human obesity genes reveals a major susceptibility locus on chromosome 10 Nat Genet 1998 20 304 308 9806554 10.1038/3123
Crawford DC Carlson CS Rieder MJ Carrington DP Yi Q Smith JD Eberle MA Kruglyak L Nickerson DA Haplotype diversity across 100 candidate genes for inflammation lipid metabolism and blood pressure regulation in two populations Am J Hum Genet 2004 74 610 622 15015130 10.1086/382227
Slager SL Schaid DJ Case-control studies of genetic markers: power and sample size approximations for Armitage's test for trend Hum Hered 2001 52 149 153 11588398 10.1159/000053370
Tregouet DA Barbaux S Escolano S Tahri N Golmard JL Tiret L Cambien F Specific haplotypes of the P-selectin gene are associated with myocardial infarction Hum Mol Genet 2002 11 2015 2023 12165563 10.1093/hmg/11.17.2015
Oyadomari S Mori M Roles of CHOP/GADD153 in endoplasmic reticulum stress Cell Death Differ 2004 11 381 389 14685163 10.1038/sj.cdd.4401373
Dominguez G Lakatos A Kuhar MJ "Characterization of the cocaine- and amphetamine-regulated transcript (CART) peptide gene promoter and its activation by a cyclic AMP- dependent signaling pathway in GH3 cells." J Neurochem 2002 80 885 893 11948252 10.1046/j.0022-3042.2002.00775.x
Fu M Cheng H Chen L Wu B Cai M Xie D Fu Z "[Association of the cocaine and amphetamine-regulated transcript gene with type 2 diabetes mellitus]" Zhonghua Nei Ke Za Zhi 2002 41 805 808 12654232
Vasseur F Helbecque N Dina C Lobbens S Delannoy V Gaget S Boutin P Vaxillaire M Lepretre F Dupont S Hara K Clement K Bihain B Kadowaki T Froguel P Single-nucleotide polymorphism haplotypes in the both proximal promoter and exon 3 of the APM1 gene modulate adipocyte-secreted adiponectin hormone levels and contribute to the genetic risk for type 2 diabetes in French Caucasians Hum Mol Genet 2002 11 2607 2614 12354786 10.1093/hmg/11.21.2607
Lewis CM Genetic association studies: design analysis and interpretation Brief Bioinform 2002 3 146 153 12139434
Hartl DL Clark AG Principles of Population Genetics 1997 3 Sinauer Associates 111 162
Otabe S Clement K Dina C Pelloux V Guy-Grand B Froguel P Vasseur F A genetic variation in the 5' flanking region of the UCP3 gene is associated with body mass index in humans in interaction with physical activity Diabetologia 2000 43 245 249 10753049 10.1007/s001250050037
Boutin P Dina C Vasseur F Dubois S Corset L Seron K Bekris L Cabellon J Neve B Vasseur-Delannoy V Chikri M Charles MA Clement K Lernmark A Froguel P GAD2 on Chromosome 10p12 Is a Candidate Gene for Human Obesity PLoS Biol 2003 1 E68 14691540 10.1371/journal.pbio.0000068
Nielsen DM Ehm MG Weir BS Detecting marker-disease association by testing for Hardy-Weinberg disequilibrium at a marker locus Am J Hum Genet 1998 63 1531 1540 9867708 10.1086/302114
Morton NE Collins A Tests and estimates of allelic association in complex inheritance Proc Natl Acad Sci U S A 1998 95 11389 11393 9736746 10.1073/pnas.95.19.11389
Branson R Potoczna N Kral JG Lentes KU Hoehe MR Horber FF Binge eating as a major phenotype of melanocortin 4 receptor gene mutations N Engl J Med 2003 348 1096 1103 12646666 10.1056/NEJMoa021971
Hercberg S Preziosi P Briancon S Galan P Triol I Malvy D Roussel AM Favier A A primary prevention trial using nutritional doses of antioxidant vitamins and minerals in cardiovascular diseases and cancers in a general population: the SU.VI.MAX study – design methods and participant characteristics. SUpplementation en VItamines et Mineraux AntioXydants Control Clin Trials 1998 19 336 351 9683310 10.1016/S0197-2456(98)00015-4
Lepretre F Cheyssac C Amouyel P Froguel P Helbecque N A promoter polymorphism in CD36 is associated with an atherogenic lipid profile in a French general population Atherosclerosis 2004 173 375 377 15064117 10.1016/j.atherosclerosis.2003.12.029
Boutin P Wahl C Samson C Vasseur F Laget F Froguel P Big Dye terminator cycle sequencing chemistry: accuracy of the dilution process and application for screening mutations in the TCF1 and GCK genes Hum Mutat 2000 15 201 203 10649499 10.1002/(SICI)1098-1004(200002)15:2<201::AID-HUMU11>3.0.CO;2-8
Blomeke B Sieben S Spotter D Landt O Merk HF Identification of N-acetyltransferase 2 genotypes by continuous monitoring of fluorogenic hybridization probes Anal Biochem 1999 275 93 97 10542113 10.1006/abio.1999.4288
Dudbridge F "Pedigree disequilibrium tests for multilocus haplotypes" Genet Epidemiol 2003 25 115 121 12916020 10.1002/gepi.10252
Wang XZ Kuroda M Sok J Batchvarova N Kimmel R Chung P Zinszner H Ron D Identification of novel stress-induced genes downstream of chop Embo J 1998 17 3619 3630 9649432 10.1093/emboj/17.13.3619
Schreiber E Matthias P Muller MM Schaffner W Rapid detection of octamer binding proteins with 'mini-extracts' prepared from a small number of cells Nucleic Acids Res 1989 17 6419 2771659
| 15823203 | PMC1087839 | CC BY | 2021-01-04 16:38:18 | no | BMC Genet. 2005 Apr 11; 6:19 | utf-8 | BMC Genet | 2,005 | 10.1186/1471-2156-6-19 | oa_comm |
==== Front
BMC GenomicsBMC Genomics1471-2164BioMed Central London 1471-2164-6-461579042810.1186/1471-2164-6-46Research ArticleOsteo-Promoter Database (OPD) – Promoter analysis in skeletal cells Grienberg Inbal [email protected] Dafna [email protected] Department of Cell and Developmental Biology, Sackler School of Medicine, Tel-Aviv University, Israel2005 25 3 2005 6 46 46 8 11 2004 25 3 2005 Copyright © 2005 Grienberg and Benayahu; licensee BioMed Central Ltd.2005Grienberg and Benayahu; 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
Increasing our knowledge about the complex expression of genes in skeletal tissue will provide a better understanding of the physiology of skeletal cells. The study summarizes transcriptional regulation factors interacting and cooperating at promoter regions that regulate gene expression. Specifically, we analyzed A/T rich elements along the promoter sequences.
Description
The Osteo-Promoter Database (OPD) is a collection of genes and promoters expressed in skeletal cells. We have compiled a new viewer, OPD, as unique database developed and created as an accessible tool for skeletal promoter sequences. OPD can navigate to identify genes specific to skeletal cDNA databases and promoter analysis sites. OPD offers exclusive access to facilitate a dynamic extraction of promoters' gene-specific analyses in skeletal tissue. The data on promoters included in OPD contains cloned promoters or predicted promoters that were analyzed by bioinformatics tools. OPD offers MAR-analysis, which allocates A/T rich elements along these promoter sequences.
Conclusion
The analysis leads to a better insight of proteins that bind to DNA, regulate DNA, and function in chromatin remodeling. The OPD is a distinctive tool for understanding the complex function of chromatin remodeling and transcriptional regulation of specific gene expression in skeletal tissue.
==== Body
Background
The Human Genome Project has provided large-scale sequencing and a multitude of genetic databases, based on information from bacteria, plant, Drosophila, vertebrates and mammals [1-5]. These databases represent evolutionary information that is summarized in the public databases NCBI, EMBL, and HUGO. In silico biology explores various tools that enable the translation of raw data into workable models and provides guidance for high-throughout gene analysis to 'make sense' of the genomic data. Computational tools analyzed the genetic information and data mining aided the knowledge interpretation in the functional cellular context. Genes expression from tissue-specific libraries can be developed to new models and add insight into biological information in a context of gene networks, protein pathways ect. Genes in eukaryotic cells are regulated in "active regions," in which chromatin structure is "open" and accessible to DNA-binding proteins and "silent regions" where "packed" chromatin renders the DNA inaccessible. The regulation is performed on several epigenetic levels that include DNA methylation, nucleosome positioning, histone modification, and additional components of higher-order structure. These modifications can activate or repress RNA synthesis that is associated with tissue-specific genes in the absence of transcriptional enhancers. The details of how chromatin structures modified are unresolved, but it is clear that DNA methylation has a direct effect on both histone acetylation and on higher-order chromatin structure that regulates DNA-chromatin structure [6].
Transcriptional regulation factors interacting and cooperating at promoter regions determine the fate of cells and their function. The promoter is a cis-acting element immediately upstream of the transcription start site (TSS) that controls the rate of the initiation of transcription. The promoter region is important in regulation of various processes in development, morphogenesis, cell differentiation; hormonal communication and cellular stress responses. To date, only a small fraction of promoters is identified and most are not yet cloned. To better understand tissue specificity, there is a need to explore the interactions that occur between chromatin remodeling proteins and the promoters. Such interactions take place next to transcription-factor (TF) binding sites. An important motif presented in the promoter is an A/T-rich region that binds proteins with an A/T-hook motif. Numerous chromatin-remodeling proteins identified by presence of single or multiple A/T hook motifs that form complex of the protein with the DNA [7]. The A/T rich regions are characteristic by MAR frequently observed at the genomic regions and affect the interactions with the A/T hook motif of the proteins [8].
The knowledge evolve transcription factor binding sites on the promoter will allow the creation of the regulation map of each gene and will enable the recognition of other transcription regulation systems that control gene expression. To learn how promoters operate, it is first necessary to identify them. To date, a relatively small number of promoters have been identified, and, thus, it is possible to predict the location of the promoter upstream to the coding regions of the gene sequence. Gene-structure prediction programs, such as BCM Gene Finder tool, ConPro [9], Core Promoter [10], FirstEF [11], Gene2Promoter [12], FunSiteP [13] and MatInspector [14] enable to classify promoters. Promoters are being analyzed on species levels as well, for example, the Saccharomyces promoter database (SCPD) [14,15] and C. elegans promoter database (CEPDB) , EPD is a Eukaryotic Promoter Database [16], and LSPD is a liver-specific tissue promoter database . Transcription factors bind to regulatory elements upstream of transcription start sites and interact with other factors to create an abundant regulatory environment. Each regulatory protein binds to a specific element on the promoter, in response to an appropriate signal. The complexity of the regulatory system can be explored by the use of software that searches for specific transcription elements on the promoter. Genomatix offers S/MARt DB [17] and S/MARs [18], aid in the analysis of scaffold/matrix-attached regions and the nuclear matrix proteins.
Construction and content
The OPD homepage contains a list of genes, where each gene has one entry link that enables the user to obtain the specific gene description. Skeletal genes which are cataloged in OPD are obtained from Human Bone libraries such as: Skeletal Gene Database (SGD) [19] and Human Bone Marrow stroma (lib. 931) [20]. Bone tissue specific expression was defined using the Unigene database . Information about gene products, their cellular function and chromosomal locus imported from the NCBI database. Accession numbers were used to link the gene entries to information in the resource databases. Accession numbers were obtained from NCBI and HUGO. Information about human genetic disorders and their associated genes is derived from the Online Mendelian Inheritance in Man database, OMIM . All the references were structured with hyperlinked both to the original Web servers and to the NCBI and HUGO database and OMIM. Searching for a gene's promoter was performed using NCBI and PUBMED databases. The accession number of each promoter is linked to the promoter sequence page in the NCBI database. MAR-Analysis in the promoter region is carried out using the MAR-FINDER program [21]. The MAR Finder program algorithm scores A/T nucleotides along promoters using a 200 bp window at 10 bp intervals with MAR potential of 0.1 as minimum significance threshold.
Utility; Osteopromoter Database (OPD) description
The Osteo-Promoter Database (OPD) summarizes the information about genes and promoters were designed based on analysis of resources covering skeletal cells. The functional genes presented in OPD were annotated and catalogued from human bone marrow stroma [19] and the skeletal gene database (SGD) [20]. The OPD is a novel bone-tissue promoter database that includes hundreds of genes that mediate osteoblast cells proliferation and differentiation [19,20,22-25]. The genes and promoters included in OPD from the PUBMED, bibliographic references databases, and cDNA sequences centers and more promoters that were identified using Gene2Promoter. The data is organized in alphabetical order and have cross-references with other databases.
In OPD, we summarized series of genes known to play a role in the regulation of skeletal cells. The information explores the transcription controlled by various factors on skeletal cells [22-25]. OPD presents distribution of transcription-factor sites in promoter regions and the connection with chromatin structure [26,27]. Promoters were analyzed and enlighten on the connections between transcription factors and chromatin remodeling factors. We choose to focus on A/T-rich region that binds A-T hook proteins and serves as interface between transcription factors and chromatin remodeling proteins. MAR regions that contain A/T-rich elements are embedded along the genome and recognized for their role in the regulation of gene transcription by creating an environment that controls gene expression at the chromatin remodeling level.
Specifically, examples are given to genes related to progenitors proliferate and differentiate to osteoblasts that form the calcified matrix. Table 1 summarizes genes related to cell proliferation, differentiation and matrix deposition in the formed bone. Transcription factors and hormones control the expression of genes responsible for cell proliferation. The cell's differentiation requires sequential activation of regulatory proteins and signaling molecules such as BMPs and hormones. The osteoblast maturation is defined by the biosynthesis and organization of extracellular matrix proteins, such as osteocalcin, bone sialoprotein and osteonectin [23]. We identified from the literature two hundred promoters' that were cloned, and the rest of the promoters included in OPD, were not cloned but identified using bioinformatics tools.
Table 1 Genes related to cells stages of differentiation
Proliferation Differentiation Mineralization
Dlx5 Collagen 2A1, Collagen 1A1 phosphatase Alkaline
jun-c, fos-c BMP7, BMP4, BMP2 Osteopontin
Cbfa1 TGFR beta3, TGFR beta1 Osteocalcin
Vitamin D IGF2, IGF1 sialoprotein Bone
Prostaglandin E2 Fibrillin 1, 2 Osteonectin
PTH FGER 1-3 Biglycan
Table summarized gene expressed by cells at different stages of differentiation. Genes are related to various families of functional genes; growth factors, transcription factors and hormones are presented according their expression throughout osteoblastic differentiation.
Database organization
Home Page
OPD catalogs hundreds of genes and promoters that are organized in alphabetical order, with direct access from the OPD homepage to each specific gene possible via the list. To view a specific gene, one can view the list and choose the gene or alternatively, identify the gene through a search engine. To search for the gene, the user types the full name of the gene/protein into the search engine box and clicks on the "search" button (Fig. 1). Entry into the gene page will provide a description for each gene, including expression analysis and accession numbers structured as specific links. The information for each gene/protein on the promoter includes chromosomal location, catalog international number of sequences, cellular function, mutations (OMIM), and involvement in differentiation or diseases (Fig. 2, Table 2). A/T-rich region analyses are available through the "see picture" link next to the name of the gene. Absence of a link means that promoter of the gene has not been investigated yet. On the left side of the page, there are useful links to the promoter analysis and to the genomic databases sites (Fig. 1). The OPD presents promoter sequences identified and readable from positions within sequences of the HUGO and NCBI databases. Other links to skeletal cDNA databases and promoter analysis sites are presented.
Figure 1 OPD catalogs hundreds of genes and promoters that are organized in alphabetical order, with direct access from the OPD homepage to each specific gene possible via the list. A search for a specific protein can be carried out by paging the list or by using the search engine on the top of the Homepage. Each name of gene is linked to gene presentation. Clicking on "see picture" link enables entry to MAR analysis of each promoter. From the page are links to the promoter analysis and to the other genomic databases sites.
Figure 2 Entry into the gene page will provide a description for specific gene, including expression analysis and accession numbers structured as specific links. The description includes also cellular function and linked disease for mutations when available. An example of the Cbfa1 gene is presented and the promoter sequence linked to the NCBI page is shown. MAR analysis of this promoter is performed through "A/T RICH graph" link. Accession numbers are linked to NCBI and UNIGENE pages, which present gene /protein sequences and bibliographic references.
Table 2 Detailed information related to genes in osteogenic differentiation
GENE PROTEIN CELLULAR FUNCTION DISEASE LOCUS ACCESSION NUMBER PROMOTER
BGLAP Bone gamma-carboxyglutamate (gla) protein (osteocalcin) Gamma-carboxyglutamic acid residues are formed by vitamin K dependent carboxylation. These residues are essential for the binding of calcium. 1q25-q31 P-NP_000702
N-NM_000711
U-Hs.2558 AY147065
BGN Biglycan Binding to collagen fibrils and transfering growth factor-beta. It may promote neuronal survival. Happle syndrome Xq28 P-NP_001702
N-NM_001711
U-Hs.821 U82940
X83526
BMP4 Bone morphogenetic protein 4 Involved in bone induction & tooth development Fibrodysplasia ossificans progressiva 14q22-q23 P-NP_001193
N-NM_001202
U-Hs.68879 AH007194
AF210053
CBFA1 Core-binding factor alpha-subunit 1 Osteoblast-specific transcription factor Cleidocranial dysplasia murine Cbfa1+/- cleidocranial dysplasia murine Cbfa1-/- perinatal lethal with deficient ossification of skeleton 6p21 P-NP_004339
N-NM_004348
U-Hs.121895 AB013356
ESR1 Estrogen receptor 1 (ER alpha) Activities of this protein have been demonstrated in the regulation of a variety of genes including lactoferrin, osteopontin, medium-chain acyl coenzyme A dehydrogenase (MCAD) and thyroid hormone receptor genes. 6q25.1 P-NP_004442
N-NM_004451
U-Hs.110849 X62462
PRLR Prolactin receptor This is a receptor for the anterior pituitary hormone prolactin 5p14-p13 P-NP_000940
N-NM_000949
U-Hs.1906 AF091859
List of factors such as transcription factors, hormones and matrix proteins play a role mediating of osteoblast differentiation.
Management System
The database is constructed based on Access database and is presented on the Web pages were created dynamically by ASP scripts written in html and VB script language. The A/T-rich graphs designed in Bitmap format.
Promoter analysis
OPD summarizes genes that are important in the control of osteoblast differentiation. OPD includes hundreds of promoters that were identified by cloning and the study of the promoter function was performed at the molecular level. Other promoters included were not cloned, but are related to genes relevant to skeletal-cell differentiation and function, which we investigated based on each promoter's prediction from the genomic region. The task of OPD focused on resolving A/T-rich sequences in a defined sequence, using the MAR-FINDER algorithm. The analysis by MAR-FINDER is based on the integration of the frequency of the motif in a defined sequence in relation to the probability of their random occurrence in the same sequence. This software scores A/T nucleotides in a given sequence, and statistical significance is associated with the frequency of occurrence of the pattern influenced by A-T content.
Using bioinformatics we predicted promoter structures from genomic contigs that had not been cloned and followed with the analyzed of A/T-rich elements. Herein we present few examples presented at the OPD; TGFβ3 and TGFRβ3 promoters were predicated. TGFβ3 is constructed from 7 exons and the promoter region upstream to TSS in contig NT_026437 identified A/T rich at -1180-1970 bp. TGFRβ3 gene was identified in contig NT_034388 includes 18 exons and the promoter was identified from the same genomic region, includes three A/T-rich elements at -420-1120 bp, -1250-1330 bp, and 1690-1970 bp. An analysis for BMP factors that are part of the TGF-β superfamily know for their role in regulating the growth and differentiation of osteoblasts. Promoters for BMP2, BMP4, and BMP7 were analyzed and identified the A/T-rich element for the BMP2 at -250-820 bp, and for BMP4 promoter three A/T-rich elements were identified at 250–430 bp, 670–930 bp, and 1170–2150 bp, but none were identified for BMP7. The BMP4 promoter includes three A/T-rich elements with high MARs potentials and thus suggests a significant potential to bind protein with an A-T hook motif.
The differences among MAR-potential are reflected by the percentage of A/T-rich sequences, higher MAR regions suggests a higher potential for binding of a protein with an A-T hook domain. MARs play a role in the context of long-distance interactions mediated between locus control regions (LCR) and the target promoters by creating loops of DNA. The DNA loop keeps the flexibility in A/T-rich length for its functionality. The presence of several copies of A/T-rich sites and longer stretches of A/T-rich sequence create a wider surface area for bonding a specific protein [28,29]. HMG-I/Y protein contains three A-T hooks responsible for multivalent interactions between the A-T hooks and appropriately spaced A/T-rich tracts are required for high-affinity binding. A single A-T hook in a polypeptide or a single A-T tract on the target DNA results in reduced interaction [28]. Furthermore, the C-terminal acidic region and the spacer regions between A-T hooks also affect the protein – DNA binding [29].
Fibroblast growth factors (FGFs) are regulators of cells division and differentiation [26] also control alkaline phosphatase; and regulate matrix proteins synthesis such as collagen, osteonectin, and osteocalcin. Four FGF receptors (FGFR 1–4) members of the tyrosine kinase receptors family were identified. FGFR1 and FGFR2 promoters were predicted and analyzed for MAR region results with two A/T-rich sites with high MAR potential for FGFR1 and lower for FGFR2. These different MAR potentials expressed in the promoters of functional genes assumed for the potential of protein to bind at a different affinity in these promoters. The analysis of promoters related to functional genes in osteogenic differentiation results with annotated genes extended for analysis by MAR-FINDER regions (Table 2). The results have shown that 73% of these genes expressing MARs regions suggesting that MAR element are a common regulatory element in transcriptional control. Thus, it indicates for A/T-rich regions a role in control and guidance of activation and repression of osteoblastic differentiation. The accessibility of proteins to A/T-rich sites on promoters of tissue-specific genes along with other factors will promote the comprehensive specificity of the regulatory process of functional genes in bone tissue.
The presence of several A/T-rich sites on promoters and their high MAR potential implies for their involvement in regulation by creating the boundaries for protein interactions. Along the human genome, 100,000 MAR regions contain A/T-rich elements. The MAR regions were recognized along with other regulatory elements at the promoters, such as enhancers and transcriptional-factors, binding sites that form protein-complex regulators for DNA transcription [18]. MAR regions control expression of genes, since they subdivide the chromatin into DNA loops, which mediate between Locus controls regions (LCRs) to distant promoters [30]. Frequency of motifs in MAR elements are influenced by the A/T content [8] that participates in global transcription regulation by binding RNA polymerase to the nuclear matrix and creating the regulatory structure necessary for transcriptional initiation. Thus, the interaction between A/T-rich sites on the DNA to A/T-hook motif of the protein is an important component of the regulatory complex that controls transcription of genes [7].
Conclusion
The data presented in OPD will assist in the study of gene regulation by proteins that possess A/T-hook motifs. The OPD provides a broad analysis of the promoters essential for skeletal cell regulation and an accessible resource for the skeletal research community. This analysis will deepen and facilitate our understanding of the mechanism of transcriptional regulation of osteogenic cells. The construction of OPD, a bone-specific promoter database, will lead to the development of a novel skeletal-enhanced promoter-based microarray for studying gene expression in skeletal tissues at different stages of growth, development and the pathological state of malignancies and metabolic bone diseases.
Availability and requirements
Osteo-Promoter Database (OPD), available through web at
Authors' contributions
GI carried out the collections of genes in OPD, made the predication of genes' promoters, constructed the web site and drafted the manuscript. BD participated in the design of the study, conceived of the study, coordination and helped to draft the manuscript. All authors read and approved the final manuscript.
Acknowledgements
This study is fulfillment for the requirement of MsC thesis for GI. Grants supported this study Greidinger for cancer research and the Chief Scientist, Ministry of Health to BD.
==== Refs
Rombauts S Dehais P Van Montagu M Rouze P PlantCARE, a plant cis-acting regulatory element database Nucleic Acids Res 1999 27 295 296 9847207 10.1093/nar/27.1.295
Higo K Ugawa Y Iwamoto M Korenaga T Plant cis-acting regulatory DNA elements (PLACE) database: 1999 Nucleic Acids Res 1999 27 297 300 9847208 10.1093/nar/27.1.297
The FlyBase database of the Drosophila Genome Projects and community literature. The FlyBase Consortium Nucleic Acids Res 1999 27 85 88 9847148 10.1093/nar/27.1.85
Blake JA Eppig JT Richardson JE Davisson MT The Mouse Genome Database (MGD): expanding genetic and genomic resources for the laboratory mouse. The Mouse Genome Database Group Nucleic Acids Res 2000 28 108 111 10592195 10.1093/nar/28.1.108
Strausberg RL Feingold EA Klausner RD Collins FS The mammalian gene collection Science 1999 286 455 457 10521335 10.1126/science.286.5439.455
Geiman TM Robertson KD Chromatin remodeling, histone modifications, and DNA methylation-how does it all fit together? J Cell Biochem 2002 87 117 125 12244565 10.1002/jcb.10286
Aravind L Landsman D AT-hook motifs identified in a wide variety of DNA-binding proteins Nucleic Acids Res 1998 26 4413 4421 9742243 10.1093/nar/26.19.4413
Liebich I Bode J Reuter I Wingender E Evaluation of sequence motifs found in scaffold/matrix-attached regions (S/MARs) Nucleic Acids Res 2002 30 3433 3442 12140328 10.1093/nar/gkf446
Liu R States DJ Consensus promoter identification in the human genome utilizing expressed gene markers and gene modeling Genome Res 2002 12 462 469 11875035 10.1101/gr.198002
Zhang MQ Identification of human gene core promoters in silico Genome Res 1998 8 319 326 9521935
Davuluri RV Grosse I Zhang MQ Computational identification of promoters and first exons in the human genome Nat Genet 2001 29 412 417 11726928 10.1038/ng780
Scherf M Klingenhoff A Werner T Highly specific localization of promoter regions in large genomic sequences by PromoterInspector: a novel context analysis approach J Mol Biol 2000 297 599 606 10731414 10.1006/jmbi.2000.3589
Hehl R Wingender E Database-assisted promoter analysis Trends Plant Sci 2001 6 251 255 11378466 10.1016/S1360-1385(01)01954-9
Quandt K Frech K Karas H Wingender E Werner T MatInd and MatInspector: new fast and versatile tools for detection of consensus matches in nucleotide sequence data Nucleic Acids Res 1995 23 4878 4884 8532532
Zhu J Zhang MQ SCPD: a promoter database of the yeast Saccharomyces cerevisiae Bioinformatics 1999 15 607 611 10487868 10.1093/bioinformatics/15.7.607
Praz V Perier R Bonnard C Bucher P The Eukaryotic Promoter Database, EPD: new entry types and links to gene expression data Nucleic Acids Res 2002 30 322 324 11752326 10.1093/nar/30.1.322
Liebich I Bode J Frisch M Wingender E S/MARt DB: a database on scaffold/matrix attached regions Nucleic Acids Res 2002 30 372 374 11752340 10.1093/nar/30.1.372
Frisch M Frech K Klingenhoff A Cartharius K Liebich I Werner T In silico prediction of scaffold/matrix attachment regions in large genomic sequences Genome Res 2002 12 349 354 11827955 10.1101/gr.206602. Article published online before print in January 2002
Jia L Young MF Powell J Yang L Ho NC Hotchkiss R Robey PG Francomano CA Gene expression profile of human bone marrow stromal cells: high-throughput expressed sequence tag sequencing analysis Genomics 2002 79 7 17 11827452 10.1006/geno.2001.6683
Jia L Ho NC Park SS Powell J Francomano CA Comprehensive resource: Skeletal gene database Am J Med Genet 2001 106 275 281 11891679 10.1002/ajmg.10227
Singh GB Kramer JA Krawetz SA Mathematical model to predict regions of chromatin attachment to the nuclear matrix Nucleic Acids Res 1997 25 1419 1425 9060438 10.1093/nar/25.7.1419
Katagiri T Takahashi N Regulatory mechanisms of osteoblast and osteoclast differentiation Oral Dis 2002 8 147 159 12108759 10.1034/j.1601-0825.2002.01829.x
Aubin JE Regulation of osteoblast formation and function Rev Endocr Metab Disord 2001 2 81 94 11704982 10.1023/A:1010011209064
McCarthy TL Ji C Centrella M Links among growth factors, hormones, and nuclear factors with essential roles in bone formation Crit Rev Oral Biol Med 2000 11 409 422 11132763
Karsenty G Minireview: transcriptional control of osteoblast differentiation Endocrinology 2001 142 2731 2733 11415989 10.1210/en.142.7.2731
Bucher P Trifonov EN Compilation and analysis of eukaryotic POL II promoter sequences Nucleic Acids Res 1986 14 10009 10026 3808945
Bucher P Weight matrix descriptions of four eukaryotic RNA polymerase II promoter elements derived from 502 unrelated promoter sequences J Mol Biol 1990 212 563 578 2329577 10.1016/0022-2836(90)90223-9
Maher JF Nathans D Multivalent DNA-binding properties of the HMG-1 proteins Proc Natl Acad Sci U S A 1996 93 6716 6720 8692884 10.1073/pnas.93.13.6716
Frank O Schwanbeck R Wisniewski JR Protein footprinting reveals specific binding modes of a high mobility group protein I to DNAs of different conformation J Biol Chem 1998 273 20015 20020 9685339 10.1074/jbc.273.32.20015
Heng HH Krawetz SA Lu W Bremer S Liu G Ye CJ Re-defining the chromatin loop domain Cytogenet Cell Genet 2001 93 155 161 11528105 10.1159/000056977
| 15790428 | PMC1087840 | CC BY | 2021-01-04 16:39:52 | no | BMC Genomics. 2005 Mar 25; 6:46 | utf-8 | BMC Genomics | 2,005 | 10.1186/1471-2164-6-46 | oa_comm |
==== Front
BMC Infect DisBMC Infectious Diseases1471-2334BioMed Central London 1471-2334-5-241581998310.1186/1471-2334-5-24Research ArticleOutcome of infections due to pandrug-resistant (PDR) Gram-negative bacteria Falagas Matthew E [email protected] Ioannis A [email protected] Sofia K [email protected] George [email protected] Panayiota [email protected] Argyris [email protected] Alfa Institute of Biomedical Sciences (AIBS), Athens, Greece2 Tufts University School of Medicine, Boston, Massachusetts, USA3 Alfa HealthCare, Athens, Greece4 Department of Medicine, University Hospital, University of Crete, School of Medicine, Heraklion, Greece5 Intensive Care Unit, "Henry Dunant" Hospital, Athens, Greece2005 8 4 2005 5 24 24 11 2 2005 8 4 2005 Copyright © 2005 Falagas et al; licensee BioMed Central Ltd.2005Falagas 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 increasing problem of infections due to multidrug-resistant Gram-negative bacteria has led to re-use of polymyxins in several countries. However, there are already clinical isolates of Gram-negative bacteria that are resistant to all available antibiotics, including polymyxins.
Methods
We present a case series of patients with infections due to pathogens resistant to all antimicrobial agents tested, including polymyxins. An isolate was defined as pandrug-resistant (PDR) if it exhibited resistance to all 7 anti-pseudomonal antimicrobial agents, i.e. antipseudomonal penicillins, cephalosporins, carbapenems, monobactams, quinolones, aminoglycosides, and polymyxins.
Results
Clinical cure of the infection due to pandrug-resistant (PDR) Gram-negative bacteria, namely Pseudomonas aeruginosa or Klebsiella pneumoniae was observed in 4 out of 6 patients with combination of colistin and beta lactam antibiotics.
Conclusion
Colistin, in combination with beta lactam antibiotics, may be a useful agent for the management of pandrug-resistant Gram-negative bacterial infections. The re-use of polymyxins, an old class of antibiotics, should be done with caution in an attempt to delay the rate of development of pandrug-resistant Gram-negative bacterial infections.
==== Body
Background
The incidence of multiresistant bacterial infections increases worldwide. This made the medical community to re-evaluate older agents such as colistin, for infections due to Gram-negative multiresistant bacteria, including Pseudomonas aeruginosa, Klebsiella pneumoniae, and Acinetobacter baumannii. Colistin was used for about two decades after its discovery in 1950, but the reported nephrotoxicity and neurotoxicity led to gradual decrease of its use. However, the antibiotic recently regained some popularity in several countries as a salvage antimicrobial agent against bacteria resistant to all other classes of antibiotics. Fortunately, several studies during the last five years did not verify the high rates of toxicity of the medication reported in the old literature [1,2].
The reuse of colistin is associated with a possible significant therapeutical problem, namely the advent of bacteria resistant to all classes of available antimicrobial agents, including the polymyxins. These bacteria are really pandrug-resistant. It should be noted that the definition of pandrug-resistant Gram-negative bacteria does not include always testing for colistin in many countries. For example, a high mortality rate (60%) was reported in a study of Acinetobacter baumannii infections from Taiwan; however, no colistin was used in the in vitro susceptibility testing and, most important, the drug was not given to patients [3,4]. There are recent reports from patients with cystic fibrosis with pandrug-resistant Gram-negative bacterial infections [5,6]. In addition, microbiological data about clinical isolates of Gram-negative bacteria resistant to colistin were included in some recent reports [7-12]. However, no study up to now has presented the characteristics of such patients, clinical information about the infections as well as the outcomes. The objective of this study was to present data regarding the clinical characteristics, therapeutic management, and clinical outcome of a group of patients with infections caused by pandrug-resistant (PDR) Gram-negative microorganisms.
Methods
Design of the study – data collection
Patients with infections caused by pandrug-resistant Gram-negative bacteria (resistant to all tested antibiotics, including colistin), who were hospitalized in the intensive care unit, during the period from 1/October/2000 to 31/August/2004 (hospital 1) were identified by the ICU's electronic database, and during the same period in hospital 2. The medical records of these patients were reviewed. Data for several variables, including demographic and clinical information, as well as results of laboratory and imaging tests of the patients were collected using a specially designed case report form and entered in a computer database.
Definitions
Multidrug-resistant (MDR) Acinetobacter baumannii, Pseudomonas aeruginosa, or Klebsiella pneumoniae strains were defined if resistance of the isolates was observed to 5 out of the 7 anti-pseudomonal classes of antimicrobial agents, i.e. antipseudomonal penicillins, cephalosporins, carbapenems, monobactams, quinolones, aminoglycosides, and colistin. Clinically important subsets of MDR Gram-negative isolates are those sensitive only to colistin and carbapenems (CCOS-colistin-carbapenem-only-sensitive). An isolate was defined as colistin-only-sensitive (COS) if it was resistant to all anti-pseudomonal agents, except colistin, and as PDR if it exhibited resistant to all 7 anti-pseudomonal antimicrobial agents, including colistin. Intermediate sensitivity was considered as resistance.
The definition of the site of the infection and of the responsible pathogen were based on the clinical symptoms and signs of individual patients, imaging findings, and on the isolation of a possible pathogen from evaluable clinical specimens. Specifically, diagnosis of pneumonia required two or more serial chest radiographs with at least one of the following: new or progressive and persistent infiltrate, consolidation, cavitation, or pleural effusion. In addition, patients must have had fever >38°C with no other recognized cause, or abnormal white blood cell count [leukopenia (< 4000 WBC/mm3) or leukocytosis (≥ 12.000 WBC/mm3)], and at least two of the following: new onset of purulent sputum or change in character of sputum, increased respiratory secretions or increased suctioning requirements, new onset or worsening of cough or dyspnea or tachypnea, rales or bronchial breath sounds, or worsening gas exchange [13].
Bacteremia required either growth of a recognized pathogen from one or more blood specimen cultures or at least one of the following signs or symptoms: fever (>38°C), chills, or hypotension and at least one of the following: a) common skin contaminant (e.g., diphtheroids, Bacillus sp., Propionibacterium sp., coagulase-negative staphylococci, or micrococci) grown from two or more blood cultures drawn on separate occasions or b) common skin contaminant (e.g., diphtheroids, Bacillus sp., Propionibacterium sp., coagulase-negative staphylococci, or micrococci) grown from at least one blood culture from a patient with an intravascular line and physician- instituted antimicrobial therapy [13].
Infections at other body sites or fluids, such as urinary tract infections, meningitis, and central venous catheter-related infections were defined based on guidelines from the Centers for Disease Control and Prevention [13].
Microbiological testing
Identification of all causative microorganisms was performed by classic microbiologic methods. Susceptibility testing was performed both by the disk diffusion method and according to an automated broth microdilution method (bioMerieux, Vitek II, Hazelwood, MO). The breakpoints were those defined by the Clinical and Laboratory Standards Institute (CLSI) [14]. The MIC breakpoint used to identify bacteria susceptible to colistin was 2 mg/l. Bacteria for which MIC was 2 mg/l or less were considered susceptible while bacteria with MIC 4 mg/l or more were considered resistant. Susceptibility to colistin was tested by means of the disk diffusion method with the use of 10 μg of colistin sulfate disk (Oxoid, Basingstoke, Hants, England). Isolates were considered sensitive if the inhibition zone was 11 mm or more. The results were verified by the E-test technique using the same cut points to determine the resistant isolates (sensitive: 2 mg/l or less, resistant: 4 mg/l or more). No molecular biology testing was performed to clarify whether isolates of the same microbial species represented the same or different strains.
Results
Seven patients were identified to have infection due to pandrug-resistant Gram-negative bacteria during the study periods. The isolated organisms were found to be resistant to the following antibiotics; amikacin, aztreonam, cefepime, ceftazidime, ciprofloxacin, gentamicin, imipenem, meropenem, netilmicin, pefloxacin, piperacillin/tazobactam, ticarcillin/clavulanic acid, trimethoprim/sulfamethoxazole, cefpirome, colistin, and isepamicin. The clinical features of patients, including the isolated pandrug-resistant strains, the antimicrobial agents used for the management of the infections, as well as their outcome, are shown in the table. The colistin formulation that was administered to the patients was sodium colistimethate (Colomycin, Forest Laboratories®, Kent, UK). One milligram of the colistin used is approximately equal to 12,500 IU. Other data related to the hospitalization of these patients are summarized below.
Patient 1
A 47-year-old male was transferred from another ICU due to multiple chest injuries after a car accident, respiratory tract infection, and ICU polyneuropathy. Based on clinical and microbiological evidence, therapy for severe pneumonia was initiated on the 2nd day of the ICU stay, with a combination of imipenem/cilastatin, gentamicin, ofloxacin and vancomycin that were given for 10, 22, 6 and 14 days respectively. Imipenem/cilastatin was substituted with piperacillin/tazobactam given for 12 days, and piperacillin/tazobactam was then substituted with ceftriaxone given for 10 days. On the 30th day teicoplanin was added to the regimen, given for 21 days. On the 40th day of the ICU stay, a PDR Pseudomonas aeruginosa strain was isolated from his bronchial secretions. He was subsequently treated, due to the persistence of the infection, with a prolonged course of intravenous colistin for a total of 52 days, combined with imipenem/cilastatin for 12 days, then meropenem for 23 days and then ofloxacin for 16 days. On the 54th day a strain of Pseudomonas aeruginosa sensitive only to colistin (COS) was isolated from his bronchial secretions. The infection was gradually controlled and, finally, the patient was discharged from the hospital in a good condition.
Patient 2
A 47-year-old female was admitted to the ICU due to possible meningitis. She had been operated for meningioma and had a ventriculo-peritoneal shunt. The shunt was removed at the day of admission with simultaneous insertion of external cerebrospinal fluid drainage. The initial cultures of CSF specimens were negative. The patient was treated with intravenous ceftriaxone and vancomycin. A CSF specimen, taken on the 24th day of ICU stay, grew a MDR Pseudomonas aeruginosa strain (sensitive only to colistin and piperacillin/tazobactam, intermediate sensitive to imipenem and meropenem) and ceftriaxone was substituted with piperacillin/tazobactam and meropenem. Due to deterioration of her clinical condition several combinations of antimicrobial agents were subsequently given, including ciprofloxacin for 19 days, chloramphenicol for 11 days, colistin for 23 days, and gentamicin for 6 days. On the 62nd day of the ICU stay, a PDR Pseudomonas aeruginosa strain was isolated from a cerebrospinal fluid specimen. Until the isolation of the PDR strain, repeated CSF specimen cultures grew MDR Pseudomonas aeruginosa. A gradual change in the Pseudomonas susceptibility was observed from multidrug-resistance to pandrug-resistance. The patient died on the 66th day of hospitalisation due to Pseudomonas meningitis.
Patient 3
An 18-year-old male was admitted to the ICU with fractures of the 2nd and 3rd cervical vertebrae and concomitant acute respiratory failure type I. Pneumonia was diagnosed clinically and confirmed microbiologically. Staphylococcus aureus and COS Acinetobacter baumannii were repeatedly isolated from bronchial secretions during the second and the third week of his hospitalization, respectively. Different combinations of antibiotics were provided, due to the persistence of severe pneumonia. The antibiotics were vancomycin, broad-spectrum cephalosporins, clindamycin, piperacillin/tazobactam, aminoglycosides, quinolones, meropenem and colistin (given for a total of 21, 15, 9, 7, 25, 7, 6 and 11 days respectively, up to the 34th day of ICU stay). On the 34th day of ICU stay PDR Pseudomonas aeruginosa grew from his bronchial secretions. At that time, he was receiving colistin, meropenem, ofloxacin and gentamicin. He continued to receive the same antimicrobial combination. PDR Pseudomonas aeruginosa was not isolated again from subsequent cultures of bronchial secretions. Instead, Acinetobacter baumannii grew again from bronchial secretions on the 35th day of hospitalization and in subsequent cultures by the 45th day of ICU stay. Meanwhile a gradual improvement of the infection was noted. The patient was transferred to a specialized orthopedic center after 78 days of ICU stay, in good general condition, without evidence of pneumonia or other infection.
Patient 4
A 72-year-old female was admitted to the ICU with a complicated, severe urinary tract infection (UTI) due to COS Pseudomonas aeruginosa and acute respiratory distress syndrome (ARDS). During the last trimester, she had been hospitalized twice for episodes of UTIs. Treatment with intravenous colistin and meropenem was initiated on ICU admission and 15 days later urine specimen cultures were negative. However, due to persistence of fever the treatment was continued. Vancomycin was added to the regimen after isolation of Enterococcus faecium from cultures of the tip of the central venous catheter (CVC) on the 8th day of ICU stay. On the 33rd day, PDR Klebsiella pneumoniae was isolated from the tip of the CVC. Due to unavailability of antimicrobial agents with in vitro effect against the pathogen, no change in the antimicrobial regimen was performed. Her clinical condition was complicated on the 37th day with lower respiratory tract infection and bacteremia due to COS Pseudomonas aeruginosa, for which a short course of gentamicin was added to the treatment. From the tip of the CVC, MDR Klebsiella pneumoniae resistant to all tested drugs (not tested for colistin) was isolated again on the 48th day, and COS Klebsiella pneumoniae on the 63rd day of ICU stay. Her clinical condition improved gradually, although COS Pseudomonas aeruginosa kept being isolated in cultures of bronchial secretions up to the 78th day of ICU stay. She was transferred in good condition to another hospital for continuation of her care, without fever or other clinical and laboratory evidence of infection, after 85 days of ICU stay.
Patient 5
A 56-year-old male was admitted to the ICU after elective operation for a thoraco-abdominal aortic aneurysm and intra-operative hemorrhage. On the 24th day of ICU stay the patient was re-operated for infection of the thoracotomy with pus collection, which required drainage. On the 52nd day PDR Klebsiella pneumoniae was isolated from bronchial secretions. He had already received 31 days of colistin treatment, combined with ampicillin/sulbactam (for 22 days) and trimethoprim/sulphamethoxazole (for 8 days), all given for treatment of polymicrobial pneumonia caused by MDR strains of Acinetobacter baumannii and Klebsiella pneumoniae and bacteremia by MDR Klebsiella pneumoniae strain. By the time of isolation of the PDR microorganism the patient's condition was steadily deteriorating; eventually irreversible septic shock and multiple organ failure led to his death on the 68th day of ICU stay.
Patient 6
A 35-year-old male was transferred to our hospital from another medical center with infection of the lower respiratory tract and UTI due to COS Pseudomonas aeruginosa. The patient had a history of rupture of an aneurysm of the circle of Willis that was surgically managed. During his prolonged hospitalization, episodes of obstructive hydrocephalus and CNS infection complicated his clinical condition, which necessitated admission to the ICU. On the 32nd day of ICU stay, PDR Pseudomonas aeruginosa was isolated from a urine specimen. By that time, he had received several combinations of antimicrobial agents, including 6 days of treatment with colistin, for lower respiratory tract infection with COS Pseudomonas aeruginosa. After isolation of the PDR pathogen he received combination treatment with colistin and ceftazidime leading to clinical cure of the UTI. The clinical condition of the patient gradually improved and he became afebrile. However, the patient's respiratory tract remained constantly colonized with MDR and COS Pseudomonas aeruginosa strains. He was transferred to the neurosurgical ward after 134 days of ICU stay, in improved condition. Thirty days later, pneumonia relapsed and progressed into septic shock and multiple organ failure, and subsequent death.
Patient 7
An 82-year-old male was admitted in the nephrology clinic due to cellulitis of the right lower limb. The patient had a history of chronic renal failure and a recent episode of bronchiolitis obliterans organizing pneumonia (BOOP). Blood specimen cultures performed on admission and 24 hours later, grew both PDR Pseudomonas aeruginosa strain. Treatment with intravenous piperacillin/tazobactam was initiated. A quick improvement of the cellulitis was noted. The therapeutic regimen was not changed when a result for isolation of a PDR Pseudomonas aeruginosa strain from the blood cultures became available. The patient was discharged to home after 10 days of treatment, with the cellulitis cured. In addition, subsequent blood cultures did not reveal any isolates.
Discussion
We present the first case series that focuses on clinical information and outcome of infections due to pandrug-resistant Gram-negative bacteria. Our study shows that the isolation of a pandrug-resistant Gram-negative rod from clinical specimens does not necessarily mean a bad outcome. This may be explained by several reasons. First, the achieved concentration of several antimicrobial agents in some body fluids, including the urine may exceed the minimal inhibitory concentration of the isolated pathogen, even if the in vitro susceptibility result would place the isolate in the resistant category. Second, infections, even severe, sometimes are self-limited without the use of antimicrobial agents, as the pre-antibiotic era taught us. Third, pandrug-resistant bacteria may be colonizers of the respiratory tract in patients receiving several classes of antimicrobial agents for a long period of time, while the real pathogen may not be isolated. Fourth, it has been shown that occasionally MDR organisms may exhibit decreased virulence compared to other more sensitive organisms of the same species.
In addition, some of our patients received combinations of colistin with antimicrobial agents such as carbapenems, third generation cephalosporins, and quinolones. This fact represents an important practical observation of our study. Specifically, 4 patients achieved complete cure from the pandrug-resistant Gram-negative bacterial infection, with combinations of colistin with other antimicrobial agents, with which may have a synergistic effect. Previous experimental and clinical studies showed promising results regarding the synergistic effect of colistin with antibiotics from other classes such as β-lactams (piperacillin, ceftazidime, aztreonam, meropenem), aminoglycosides (amikacin, gentamicin), sulfonamides (trimethoprim/sulfamethoxazole), rifamycins (rifampicin), and quinolones (ciprofloxacin) against MDR Gram-negative bacteria [15-17]. However, clinical trials are needed to evaluate the in vivo results of these combinations.
Colistin resistance among Gram-negative organisms has been reported up today in some in vitro studies, as well as in surveillance studies for antimicrobial resistance [7,8,18]. In most of these studies the isolated strains of Pseudomonas aeruginosa, Klebsiella pneumoniae, and Acinetobacter baumannii exhibited a greater than 98% susceptibility to colistin or polymyxin B. However, reports about clinical parameters and outcome of infections due to colistin-resistant Gram-negative bacteria are rare in the literature. A search of PubMed and Current Contents databases revealed only 2 such reports. Both of them referred to cystic fibrosis patients [5,6]. In the first study, 5 out of 44 adult patients with cystic fibrosis and Pseudomonas aeruginosa chronic bronchial infection developed resistance to colistin. In the second study, colistin resistant Pseudomonas aeruginosa was found in 5 out of 150 children with cystic fibrosis over a 5-year period of follow up. Our study confirms the rarity in the incidence of such infections since in a study period of several years we recorded only seven such infections in two hospitals.
Gram-negative bacteria can develop resistance to polymyxins through mutation or adaptation mechanisms [19]. Mutation is inherited, low-level, and independent of the continuous presence of the antibiotic, whereas adaptation is the opposite (non-inherited, high level, and requires the continuous presence of the antibiotic). Almost complete cross-resistance exists between colistin and polymyxin B [20,21]. It has been suggested that a basic mechanism of resistance in Pseudomonas aeruginosa strains is that the outer membrane protein OprH blocks the self-promoted uptake pathway of polymyxins by replacing Mg2+ on the LPS molecule. Thus, the overexpression of OprH caused by mutation or as a result of adaptation to an Mg2+-deficient medium can be associated with resistance to polymyxins [22,23].
Studies on polymyxin-resistant Pseudomonas aeruginosa strains have suggested that cell surface changes are related with the development of resistance. These include alterations of the outer membrane of the bacteria cell (reduction of LPS, reduced levels of specific outer membrane proteins, reduction in cell envelope Mg2+ and Ca2+ contents, and lipid alterations). Specifically, the absence of 2-hydroxylaurate, the presence of 4-aminoarabinose, and increase in the palmitate content of lipid A has been associated with resistance of Pseudomonas aeruginosa to polymyxins [5,24]. In addition, studies in Salmonella spp. have shown that the two component systems PhoP-PhoQ and PmrA-PmrB regulate polymyxin resistance. The transcription of PmrA-PmrB is activated by the PhoP-PhoQ or by growth in mild acidic conditions (pH = 5.8) and micromolar concentrations of Mg2+ (10 μM). The products of PmrA-PmrB activated genes are responsible for the modifications on the LPS molecules of Gram-negative bacteria. Consequently, the affinity of polymyxins to the bacterial cell surface is reduced [24].
In addition, a recent study in Yersinia spp. demonstrated that an efflux pump/potassium antiporter system may be associated with the mediation of resistance to cationic antimicrobial peptides in general, including polymyxin B [25]. Although enzymatic resistance of bacteria to polymyxins has not been reported, it is interesting that Bacillus polymyxa subsp.colistinus produces colistinase that inactivates colistin [26].
Guidelines from Clinical and Laboratory Standards Institute (CLSI) the about the in vitro determination of the minimum inhibitory concentrations (MICs) of different microorganisms to colistin using broth dilution and agar dilution techniques were established in 1970. However, due to the rare use of colistin in the United States and most countries, the CLSI guidelines were not modified after 1981 and were withdrawn in 2000. The re-use of polymyxins during the last years will make necessary the re-evaluation of the susceptibility breakpoints. A common method for susceptibility testing of colistin has been the disk diffusion method that uses 10 micrograms of colistin sulfate disk (Oxoid, Basingstoke, Hants, England). Isolates are considered sensitive if the inhibition zone is ≥ 11 mm. It is important to emphasize that in clinical practice it is colistimethate sodium, not colistin sulfate that is used for intravenous administration. There is generally agreement in the results obtained from agar dilution and broth microdilution methods regarding testing of colistin sulfate [27]. However, it has been suggested that the disk diffusion test should be confirmed with a dilution method, because the disk diffusion method may reveal false-susceptible microorganisms [27].
We should acknowledge several limitations of this study. First, molecular typing was not performed thus the possibility that PDR isolates derived from the previous isolated microorganisms cannot be determined. Second, MDR or PDR bacteria were markers of the patients' condition and the effect of the infection and of the treatment on the outcome of the patients cannot be adequately evaluated.
Conclusion
In an era of alarmingly increasing bacterial resistance, our study adds clinical information about the outcome of infections due to pandrug-resistant Gram-negative bacteria. Further work is needed to elucidate the possible in vitro synergistic effect of combinations of colistin with antibiotics from several classes and their effectiveness in clinical practice.
Competing interests
The author(s) declare that they have no competing interests.
Authors' Contributions
MEF conceived the idea for the study; IAB, SKK and PA collected the data; MEF drafted the manuscript. All authors contributed in the writing and preparation of the manuscript. All authors read and approved the final manuscript.
Table 1 Characteristics of the reported patients and outcomes of the infections.
Patient no. 1 2 3 4 5 6 7
Isolated organism Pseudomonas aeruginosa Pseudomonas aeruginosa Pseudomonas aeruginosa Klebsiella pneumoniae Klebsiella pneumoniae Pseudomonas aeruginosa Pseudomonas aeruginosa
Site of isolation Bronchial secretions Cerebrospinal fluid Bronchial secretions Tip of central venous catheter Bronchial secretions Urine Blood
Date of isolation (mo/dd/year) 08/07/2001 07/17/2002 07/29/2003 01/16/2004 01/30/ 2004 04/16/ 2004
Days in the ICU until the isolation of the pandrug-resistant strain 40 62 34 33 52 32 0
Prior colistin use (days) 0 23 11 33 31 6 0
Antibiotic combination used after isolation Colistin + imipenem/ cilastatin, Colistin + meropenem, Colistin + ofloxacin Two days prior to the isolation, all antibiotics had been stopped Colistin + meropenem + ofloxacin + gentamicin Colistin + meropenem Colistin + ampicillin/sulbactam + trimethoprim/sulfamethoxazole Colistin + ceftazidime Piperacillin/tazobactam
Dosage/Duration of colistin 2 million IU q8h iv for 35 days, 1 million IU q8h iv for 17 days 1 million IU q8h iv for 23 days 3 million IU q8h iv for 37 days, 1 million IU q8h nebulized for 11 days 2 million IU q8h iv for 16 days, 1 million IU q8h iv for 12 days, 1 million IU q12h iv for 19 days 1 million IU q8h iv for 34 days 2 million IU q8h in for 35 days, 1 million IU q8h iv for 17 days 2.25 gr q8h iv for 10 days
Outcome of the infection Cure (clinical + microbiological). No response.
Patient died due to meningitis. Cure (clinical + microbiological). Clinical cure (microbiological persistence for 1 month). Deterioration.
Superinfection of the respiratory tract by an Acinetobacter baumannii strain. Patient died due to sepsis. Cure (clinical + microbiological). Cure (clinical + microbiological).
Pre-publication history
The pre-publication history for this paper can be accessed here:
==== Refs
Garnacho-Montero J Ortiz-Leyba C Jimenez-Jimenez FJ Barrero-Almodovar AE Garcia-Garmendia JL Bernabeu-WittelI M Gallego-Lara SL Madrazo-Osuna J Treatment of multidrug-resistant Acinetobacter baumannii ventilator-associated pneumonia (VAP) with intravenous colistin: a comparison with imipenem-susceptible VAP Clin Infect Dis 2003 36 1111 1118 12715304 10.1086/374337
Michalopoulos AS Tsiodras S Rellos K Mentzelopoulos S Falagas ME Colistin treatment in patients with ICU-acquired infections caused by multiresistant Gram-negative bacteria: the renaissance of an old antibiotic Clin Microbiol Infect 2005 11 115 121 15679485 10.1111/j.1469-0691.2004.01043.x
Hsueh PR Teng LJ Chen CY Chen WH Yu CJ Ho SW Luh KT Pandrug-resistant Acinetobacter baumannii causing nosocomial infections in a university hospital, Taiwan Emerg Infect Dis 2002 8 827 832 12141969
Kuo LC Yu CJ Lee LN Wang JL Wang HC Hsueh PR Yang PC Clinical features of pandrug-resistant Acinetobacter baumannii bacteremia at a university hospital in Taiwan J Formos Med Assoc 2003 102 601 606 14625603
Denton M Kerr K Mooney L Keer V Rajgopal A Brownlee K Arundel P Conway S Transmission of colistin-resistant Pseudomonas aeruginosa between patients attending a pediatric cystic fibrosis center Pediatr Pulmonol 2002 34 257 261 12205566 10.1002/ppul.10166
Tamm M Eich C Frei R Gilgen S Breitenbucher A Mordasini C [Inhaled colistin in cystic fibrosis] Schweiz Med Wochenschr 2000 130 1366 1372 11059027
Catchpole CR Andrews JM Brenwald N Wise R A reassessment of the in-vitro activity of colistin sulphomethate sodium J Antimicrob Chemother 1997 39 255 260 9069549 10.1093/jac/39.2.255
Manikal VM Landman D Saurina G Oydna E Lal H Quale J Endemic carbapenem-resistant Acinetobacter species in Brooklyn, New York: citywide prevalence, interinstitutional spread, and relation to antibiotic usage Clin Infect Dis 2000 31 101 106 10913404 10.1086/313902
Giamarellos-Bourboulis EJ Sambatakou H Galani I Giamarellou H In vitro interaction of colistin and rifampin on multidrug-resistant Pseudomonas aeruginosa J Chemother 2003 15 235 238 12868548 10.1159/000069498
Pitt TL Sparrow M Warner M Stefanidou M Survey of resistance of Pseudomonas aeruginosa from UK patients with cystic fibrosis to six commonly prescribed antimicrobial agents Thorax 2003 58 794 796 12947141 10.1136/thorax.58.9.794
Spence RP Towner KJ Henwood CJ James D Woodford N Livermore DM Population structure and antibiotic resistance of Acinetobacter DNA group 2 and 13TU isolates from hospitals in the UK J Med Microbiol 2002 51 1107 1112 12466410
Valero E Sevillano D Calvo A Garcia R Leturia A Gomez-Lus ML [Activity of new fluoroquinolones against clinical isolates of Acinetobacter baumannii] Rev Esp Quimioter 2001 14 358 363 11856982
Gaynes RP Horan TC Mayhall CG Surveillance of nosocomial infections. Appendix A: CDC definitions of nosocomial infections Hospital epidemiology and infection control 1996 Baltimore: Williams & Wilkins 1 14
National Committee for Clinical Laboratory Standards Performance standards for antimicrobial disc susceptibility tests. Approved standard M2-A2 S2 Waynee, PA NCCLS 1981
Rynn C Wootton M Bowker KE Alan HH Reeves DS In vitro assessment of colistin's antipseudomonal antimicrobial interactions with other antibiotics Clin Microbiol Infect 1999 5 32 36 11856210
Giamarellos-Bourboulis EJ Karnesis L Giamarellou H Synergy of colistin with rifampin and trimethoprim/sulfamethoxazole on multidrug-resistant Stenotrophomonas maltophilia Diagn Microbiol Infect Dis 2002 44 259 263 12493173 10.1016/S0732-8893(02)00443-1
Gunderson BW Ibrahim KH Hovde LB Fromm TL Reed MD Rotschafer JC Synergistic activity of colistin and ceftazidime against multiantibiotic-resistant Pseudomonas aeruginosa in an in vitro pharmacodynamic model Antimicrob Agents Chemother 2003 47 905 909 12604520 10.1128/AAC.47.3.905-909.2003
Ciofu O Fussing V Bagge N Koch C Hoiby N Characterization of paired mucoid/non-mucoid Pseudomonas aeruginosa isolates from Danish cystic fibrosis patients: antibiotic resistance, beta-lactamase activity and RiboPrinting J Antimicrob Chemother 2001 48 391 396 11533004 10.1093/jac/48.3.391
Falagas ME Kasiakou SK Colistin: the revival of polymyxins for the managment of multidrug-resistant Gram-negative bacterial infections Clin Infect Dis 2005
Moore RA Chan L Hancock RE Evidence for two distinct mechanisms of resistance to polymyxin B in Pseudomonas aeruginosa Antimicrob Agents Chemother 1984 26 539 545 6097168
Groisman EA Kayser J Soncini FC Regulation of polymyxin resistance and adaptation to low-Mg2+ environments J Bacteriol 1997 179 7040 7045 9371451
Nicas TI Hancock RE Outer membrane protein H1 of Pseudomonas aeruginosa: involvement in adaptive and mutational resistance to ethylenediaminetetraacetate, polymyxin B, and gentamicin J Bacteriol 1980 143 872 878 6259125
Moore RA Hancock RE Involvement of outer membrane of Pseudomonas cepacia in aminoglycoside and polymyxin resistance Antimicrob Agents Chemother 1986 30 923 926 3028253
Gunn JS Lim KB Krueger J Kim K Guo L Hackett M Miller SI PmrA-PmrB-regulated genes necessary for 4-aminoarabinose lipid A modification and polymyxin resistance Mol Microbiol 1998 27 1171 1182 9570402 10.1046/j.1365-2958.1998.00757.x
Bengoechea JA Skurnik M Temperature-regulated efflux pump/potassium antiporter system mediates resistance to cationic antimicrobial peptides in Yersinia Mol Microbiol 2000 37 67 80 10931306 10.1046/j.1365-2958.2000.01956.x
Ito-Kagawa M Koyama Y Selective cleavage of a peptide antibiotic, colistin by colistinase J Antibiot (Tokyo) 1980 33 1551 1555 7019177
Gales AC Reis AO Jones RN Contemporary assessment of antimicrobial susceptibility testing methods for polymyxin B and colistin: review of available interpretative criteria and quality control guidelines J Clin Microbiol 2001 39 183 190 11136768 10.1128/JCM.39.1.183-190.2001
| 15819983 | PMC1087841 | CC BY | 2021-01-04 16:28:15 | no | BMC Infect Dis. 2005 Apr 8; 5:24 | utf-8 | BMC Infect Dis | 2,005 | 10.1186/1471-2334-5-24 | oa_comm |
==== Front
BMC Infect DisBMC Infectious Diseases1471-2334BioMed Central London 1471-2334-5-281582900610.1186/1471-2334-5-28Case ReportDifficulties in assessing cytomegalovirus-associated gastric perforation in an HIV-infected patient Mégarbane Bruno [email protected]ésière Dabor [email protected] Jacqueline [email protected] Laurent [email protected] Kouroche [email protected] Frédéric J [email protected] Medical and Toxicological Intensive Care Unit, Lariboisière Hospital, Paris, France2 Department of pathology, Lariboisière Hospital, Paris, France3 Department of Bacteriology and Virology, Lariboisière Hospital, Paris, France4 Department of Gastro-enterology, Lariboisière Hospital, Paris, France2005 13 4 2005 5 28 28 2 2 2005 13 4 2005 Copyright © 2005 Mégarbane et al; licensee BioMed Central Ltd.2005Mégarbane 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
Active Cytomegalovirus (CMV) infection is a common complication in advanced symptomatic Human Immunodeficiency Virus (HIV) infection. CMV-induced intestinal perforations are hard to diagnose and may be observed throughout the gastrointestinal tract. Isolated stomach perforation is exceptional.
Case presentation
A 47-year-old man was admitted to our intensive care unit with multiorgan failure. Gastrointestinal endoscopic examination showed erythematous gastritis but normal duodenum and colon. CMV blood culture was positive. Histologic examination of a gastric biopsy showed inflammatory infiltrate and immunostaining typical intranuclear CMV inclusion bodies. Concomitant abdominal CT scan disclosed large peripancreatic hypodensities without pneumoperitoneum. The patient died despite supportive therapies and ganciclovir infusion. Postmortem examination showed a 4-cm gastric perforation adhering to the transverse colon and liver, with a thick necrotic inflammatory coating around the pancreas. The whole GI tract, except the stomach, was normal. As other causes, especially Helicobacter pylori infection could be ruled out, a causal relationship between CMV and gastric disease was assumed.
Conclusion
CMV may be responsible for gastric perforations, with difficulties in assessing the diagnosis. Early diagnosis based on cautious endoscopy and histopathologic examination is needed to make a favorable outcome possible.
==== Body
Background
Cytomegalovirus (CMV) is frequently isolated from immunosuppressed patients with HIV infection, organ or bone marrow transplants, malignancies or immunosuppressive medications. CMV may cause disseminated diseases and is generally predictive, in HIV-infected patients, of poor long-term survival reflecting severe immunosuppression [1]. In these patients, CMV gastrointestinal (GI) infection, including colitis, esophagitis, and gastritis [2,3]. Clinical signs are usually sparse, making the diagnosis difficult. Although CMV-induced intestinal perforation has been observed in the patients with advanced immunodeficiency, isolated gastric perforation remains a rare event. We report here the fatal case of an HIV-infected patient, who presented a CMV-associated gastric perforation.
Case presentation
A 47-year-old man was hospitalized in our intensive care unit (ICU) for generalized seizures and hypotension. Three days before admission, he suffered from fatigue, fever, vomiting, abdominal pain and lower GI hemorrhage. He was homeless and a chronic alcoholic. Physical examination showed general status impairment, with a temperature of 36.5°C and a Glasgow Coma Score of 10. No focal neurological deficit was noted. His blood pressure was 80/26 mmHg, his heart rate 106 /min and his respiratory rate 36 /min. His abdomen was tender with hepatomegaly. Usual blood tests showed a creatinine of 413 μmol/l (N <120), white blood cell count of 15.1 × 106 /l (N: 4 × 106-10 × 106) with 85% polymorphonuclear leukocytes, hemoglobin of 8.5 g/dl (N: 13–18) and platelets 178 × 106 /l (N: 150–400). He presented with severe lactic acidosis (arterial pH 7.21 (N: 7.34–7.45), PaCO2 3.7 kPa (N: 4.3–6.0), blood bicarbonate 11 mmol/l (N: 20–26), and plasma lactate 22 mmol/l (N: 1–2)). His prothrombin time was 9% (N>75%), fibrinogen 3.72 g/l (N: 2–4), factor V 25% (N >75%), and factors II + VII 5% (N >75%). Other tests showed an AST of 4,480 UI/l (N<50), an ALT of 2,340 UI/l (N<50), bilirubin 67 μmol/l (N<17), LDH 24,500 UI/l (N<275), lipase 11,820 UI/l (N<200), and creatine phosphokinase 3,330 UI/l (N<170). Abdominal ultrasonography showed an enlarged liver without ascites. The chest X-ray, electrocardiogram, cerebral CT scan were unremarkable. The patient was intubated. He received IV infusions of norepinephrine (6 mg/h), dobutamine (15 μg/kg/min), omeprazole (80 mg/day), N-acetylcysteine (300 mg/kg over 20 h), cefotaxime (3 g/day) and metronidazole (1.5 g/day). Continuous hemodiafiltration was started. Cultures of blood, urine, stools and tracheal aspiration were negative. Toxicological screening tests, including acetaminophen, ethanol, ethylene glycol, and methanol were negative. HIV-1 ELISA serology was positive, CD4+ lymphocytes count 160 × 106 /l (N >700), and HIV-1 RNA level was 5.8 × 105 copies/ml. Bronchoalveolar lavage (BAL) showed 2.5 × 108 cells/l with 58% lymphocytes. A diagnosis of Mycobacterium tuberculosis infection was rapidly assessed by an amplification assay (Amplified Mycobacterium Tuberculosis Direct Test, Gen Probe, California) and retrospectively confirmed by BAL cultures. Isoniazid, rifampicin, ethambutol, and pyrazinamide were started on day 3. Endoscopic examination showed grade II esophageal varices, moderate peptic esophagitis, erythematous gastritis with cardial superficial ulcerations without active bleeding, and normal duodenum and colon. CMV culture using MRC5 cells inoculation was positive in blood but not in BAL. CMV amplification assay was negative in the cerebrospinal fluid. Histopathologic analysis of a gastric biopsy showed edema, inflammatory infiltrate and typical intranuclear cytomegalic inclusion bodies in endothelial cells in mucosa. Immunohistochemistry with a specific anti-CMV antibody (E13, Argène, France) confirmed active CMV infection of the stomach (Fig. 1). Specific staining and culture were negative for Helicobacter pylori. As soon as these results were known (on day 10), ganciclovir was started. Concomitant abdominal CT scan disclosed large and heterogeneous peripancreatic hypodensities, mildly enhanced following contrast infusion (Fig. 2). However, their exact etiology remained undetermined. The patient developed hospital-acquired Pseudomonas aeruginosa pneumonia and his condition worsened. He died 30 days after admission. Postmortem examination showed a large, well-circumscribed gastric perforation with a diameter of 4 cm. The stomach was solidly adhered to the transverse colon and to the lower face of the liver. There was a thick necrotic and hemorrhagic inflammatory coating around the pancreas, which, in contrast, appeared macroscopically normal. The whole GI tract, except the stomach, was normal.
Figure 1 Biopsy specimen of gastric glandular epithelial mucosa showing the presence of intranuclear cytomegalic inclusion bodies (1A, arrow), together with edema and inflammatory infiltrate. CMV infection was confirmed using immunohistochemistry, with a specific anti-CMV antibody (DAB-peroxydase) (1B).
Figure 2 Abdominal CT scan showing the presence of large and heterogeneous hypodensities (open arrow), surrounding the pancreas (filled arrow) and mildly enhanced following contrast infusion. The spleen (gray arrow) was patchy, demonstrating ischemic zones.
CMV infection of the GI tract can cause severe damage. Isolated involvement of the stomach is possible, although rare [3,4]. Symptoms are not specific, including unexplained fever, dysphagia, sharp or postural epigastric pain, vomiting, diarrhea, and GI bleeding. Characteristic upper endoscopic findings are edema, congestion, mucosal ulcerations, multiple punched-out gastric ulcers, erosions, and upper GI hemorrhage [5]. However, findings may be mild showing congestion and thickening or be atypical, with hemorrhaged necrotic gastritis or pseudo-tumors [6]. Multiple mucosal biopsies are needed to detect CMV [7]. In HIV-infected patients, CMV is significantly associated with chronic active gastritis or gastroduodenal ulcers [7-9], in contrast with healthy adults, in whom symptomatic CMV gastric infection is exceptional [10]. Thus, early recognition of CMV GI infection, including blood cultures and cautious GI endoscopic evaluation, may allow for adequate therapy, preventing complications, such as life-threatening bleeding or perforation [3,11].
CMV infection activates endothelial cells and leukocytes, altering GI microcirculation and inducing extensive vasculitis, thrombotic vascular occlusion, necrosis, and ischemic perforation [12]. The most common reported locations of intestinal perforation are the colon (53%) [13-15], the distal ileum (40%) [16,17], and the appendix (7%) [18]. CMV-associated perforation of the stomach is a rare presentation [19]. To our knowledge, isolated CMV-induced gastric perforation has only been reported in non-HIV organ transplant recipients [20]. In our patient, CMV-associated gastric perforation, as a presenting manifestation of HIV infection, was responsible for multiorgan failure. The diagnosis was difficult, in the absence of acute abdomen and pneumoperitonium. Consistently, initial endoscopic examination missed the gastric perforation, since the hole in the stomach was probably filled in by peritoneum. Moreover it is possible that a suboptimal endoscopic evaluation of the upper GI tract was due to the patient's initial unstable condition and the performance of the gastroscopy in such a critical situation. In contrast, concomitant abdominal CT scan showed large hypodense images surrounding the pancreas, attributed after postmortem examination, to the hemorrhagic and necrotic inflammatory materials that filled in the gastric perforation.
Alternative causes of gastric ulcer, including Helicobacter pylori infection, should be discussed. Indeed, CMV is an opportunistic virus and may thus appear in previously damaged tissues. However, a recent study suggested that CMV, rather than Helicobacter pylori, may be the main causative pathogen of peptic ulcers in HIV-infected patients, in comparison to non HIV-infected ones [9]. In our case, histopathologic findings and cultures were negative for Helicobacter pylori. Moreover, CMV inclusion bodies, which were observed on initial pathological findings, clearly indicated the existence of a CMV disease at the patient's admission.
Usually, resolution of symptoms and endoscopic findings is obtained with IV ganciclovir or foscarnet. Combined antiretroviral therapies may also be effective, even without specific treatment of CMV [21]. However, in the case of perforation, despite immediate surgical resection and antiviral therapy, mortality rate remains elevated (> 80%), due to elevated operative mortality and increased postoperative complications [17,18]. In our case, ganciclovir therapy successfully reduced CMV viral load before the occurrence of death, which resulted from misdiagnosed gastric perforation.
Conclusion
This dramatic presentation demonstrates that perforation may complicate CMV gastric infection in HIV-infected patients. Physicians should be aware that early diagnosis, based on cautious GI endoscopy and histopathologic examination of GI biopsy specimen is needed to make a favorable outcome possible.
Competing interests
The author(s) declare that they have no competing interest.
Authors' contributions
BM, DR and FJB carried out the clinical study and drafted the manuscript. JF carried out the autopsy and the postmortem findings. LR carried out the microbiological studies. KV performed the gastrointestinal endoscopy. All the authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
Written consent was obtained from the patient relative for publication of study.
==== Refs
Gallant JE Moore RD Richman DD Keruly J Chaisson RE Incidence and natural history of cytomegalovirus disease in patients with advanced human immunodeficiency virus diseases treated with zidovudine. The Zidovudine Epidemiology Study Group J Infect Dis 1992 166 1223 1227 1358986
Wilcox CM Straub RF Schwartz DA Cytomegalovirus esophagitis in AIDS: a prospective evaluation of clinical response to ganciclovir therapy, relapse rate, and long-term outcome Am J Med 1995 98 169 176 7847433 10.1016/S0002-9343(99)80400-8
Bobak DA Gastrointestinal Infections Caused by Cytomegalovirus Curr Infect Dis Rep 2003 5 101 107 12641994
Vachon GC Brown BS Kim C Chessin LN CMV gastric ulcer as the presenting manifestation of AIDS Am J Gastroenterol 1995 90 319 321 7847313
Ruiz AR JrBorum ML Cytomegalovirus hemorrhagic gastritis AIDS Patient Care STDS 2001 15 1 5 11177582 10.1089/108729101460047
Laguna F Garcia-Samaniego J Alonso MJ Alvarez I Gonzalez-Lahoz JM Pseudotumoral appearance of cytomegalovirus esophagitis and gastritis in AIDS patients Am J Gastroenterol 1993 88 1108 1111 8391211
Murray RN Parker A Kadakia SC Ayala E Martinez EM Cytomegalovirus in upper gastrointestinal ulcers J Clin Gastroenterol 1994 19 198 201 7806828
Varsky CG Correa MC Sarmiento N Bonfanti M Peluffo G Dutack A Maciel O Capece P Valentinuzzi G Weinstock D Prevalence and etiology of gastroduodenal ulcer in HIV-positive patients: a comparative study of 497 symptomatic subjects evaluated by endoscopy Am J Gastroenterol 1998 93 935 940 9647023
Chiu HM Wu MS Hung CC Shun CT Lin JT Low prevalence of Helicobacter pylori but high prevalence of cytomegalovirus-associated peptic ulcer disease in AIDS patients: Comparative study of symptomatic subjects evaluated by endoscopy and CD4 counts J Gastroenterol Hepatol 2004 19 423 428 15012780 10.1111/j.1440-1746.2003.03278.x
Forne M Panes J Viver J Sierra M Garau J Gastric ulcer as a unique and symptomatic manifestation of infection by cytomegalovirus in a healthy adult Med Clin (Barc) 1989 93 743 744 2560102
Lai IR Chen KM Shun CT Chen MY Cytomegalovirus enteritis causing massive bleeding in a patient with AIDS Hepatogastroenterology 1996 43 987 991 8884325
Persoons MC Stals FS van dam Mieras MC Bruggeman CA Multiple organ involvement during experimental cytomegalovirus infection is associated with disseminated vascular pathology J Pathol 1998 184 103 109 9582535 10.1002/(SICI)1096-9896(199801)184:1<103::AID-PATH964>3.0.CO;2-C
Kram HB Hino ST Cohen RE DeSantis SA Shoemaker WC Spontaneous colonic perforation secondary to cytomegalovirus in a patient with acquired immune deficiency syndrome Crit Care Med 1984 12 469 471 6325091
Tatum ET Sun PC Cohn DL Cytomegalovirus vasculitis and colon perforation in a patient with the acquired immunodeficiency syndrome Pathology 1989 21 235 238 2561192
Wexner SD Smithy WB Trillo C Hopkins BS Dailey TH Emergency colectomy for cytomegalovirus ileocolitis in patients with the acquired immune deficiency syndrome Dis Colon Rectum 1988 31 755 761 2844477
Maguer D Bouleau-Desbordes O Guibert L Amusini P Godde F Delaporte B Ileal perforation with CMV enteritis in a newborn infant with acquired immunodeficiency syndrome Arch Pediatr 1995 2 900 7581795 10.1016/0929-693X(96)81276-0
Meza AD Bin-Sagheer S Zuckerman MJ Morales CA Verghese A Ileal perforation due to cytomegalovirus infection J Natl Med Assoc 1994 86 145 148 8169991
Kram HB Shoemaker WC Intestinal perforation due to cytomegalovirus infection in patients with AIDS Dis Colon Rectum 1990 33 1037 1040 2173658
Frank D Raicht RF Intestinal perforation associated with cytomegalovirus infection in patients with acquired immune deficiency syndrome Am J Gastroenterol 1984 79 201 205 6322576
Kaplan CS Petersen EA Icenogle TB Copeland JG Villar HV Sampliner R Minnich L Ray CG Gastrointestinal cytomegalovirus infection in heart and heart-lung transplant recipients Arch Intern Med 1989 149 2095 2100 2549896 10.1001/archinte.149.9.2095
Monkemuller KE Wilcox CM Esophageal ulcer caused by cytomegalovirus: resolution during combination antiretroviral therapy for acquired immunodeficiency syndrome South Med J 2000 93 818 820 10963519
| 15829006 | PMC1087842 | CC BY | 2021-01-04 16:28:16 | no | BMC Infect Dis. 2005 Apr 13; 5:28 | utf-8 | BMC Infect Dis | 2,005 | 10.1186/1471-2334-5-28 | oa_comm |
==== Front
BMC ImmunolBMC Immunology1471-2172BioMed Central London 1471-2172-6-81583679010.1186/1471-2172-6-8Research ArticleAccumulation of marginal zone B cells and accelerated loss of follicular dendritic cells in NF-κB p50-deficient mice Ferguson Andrew R [email protected] Ronald B [email protected] Department of Microbiology, Boston University School of Medicine, 715 Albany Street, Boston, MA 02118 USA2005 18 4 2005 6 8 8 18 11 2004 18 4 2005 Copyright © 2005 Ferguson and Corley; 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
Marginal zone (MZ) B cells play important roles in the early phases of humoral immune responses. In addition to possessing an inherent capacity to rapidly differentiate into antibody secreting cells, MZ B cells also help to regulate the fate of both T-independent and T-dependent blood-borne antigens in the spleen. For T-dependent antigens, MZ B cells bind IgM-antigen complexes in a complement-dependent manner. Once MZ B cells bind IgM-containing immune complexes (IgM-IC), they transport them into B cell follicles for deposition onto follicular dendritic cells (FDCs), an important component of secreted IgM's ability to enhance adaptive immune responses. To further define the requirement for MZ B cells in IgM-IC deposition, mice deficient in the NF-κB protein p50, which have been reported to lack MZ B cells, were analyzed for their ability to trap IgM-IC onto FDCs.
Results
Mice (2 months of age) deficient in p50 (p50-/-) had small numbers of MZ B cells, as determined by cell surface phenotype and localization in the splenic MZ. These cells bound high levels of IgM-IC both in vivo and in vitro. Subsequent to the binding of IgM-IC by the MZ B cells in p50-/- mice, small amounts of IgM-IC were found localized on FDCs, suggesting that the MZ B cells retained their ability to transport these complexes into splenic follicles. Strikingly, MZ B cells accumulated with age in p50-/- mice. By 6 months of age, p50-/- mice contained normal numbers of these cells as defined by CD21/CD23 profile and high level expression of CD1d, CD9, and IgM, and by their positioning around the marginal sinus. However, FDCs from these older p50-/- mice exhibited a reduced capacity to trap IgM-IC and retain complement components.
Conclusion
These results demonstrate that while the p50 component of the NF-κB transcription complex plays an important role in the early development of MZ B cells, MZ B cells can develop and accumulate in mice lacking this protein. These results highlight the interface between genetic deficiencies and age, and suggest that different transcription factors may play distinct roles in the development and maintenance of cell populations at different ages.
==== Body
Background
Secreted IgM antibodies play important roles in the initial phases of humoral immune responses. Unlike other antibody isotypes, secreted IgM uniquely functions as an adjuvant, promoting primary antibody responses to sub-immunogenic doses of T-dependent antigens [1-7]. The adjuvanticity of IgM is completely dependent on an intact complement system [7-9]. While mice deficient in secreted IgM make protective antibody responses to pathogens, these responses are delayed compared to the immune response of IgM sufficient mice [10-13] Interestingly, pathogens more readily disseminate into vital organs in secreted IgM-deficient mice, rather than concentrate them into secondary lymphoid organs [14]. This suggests that an important role for IgM is to serve as an arbiter of antigen localization, which may contribute to its adjuvant function(s).
To investigate how IgM affects the localization of foreign antigen, a system was developed to follow the fate of IgM-containing immune complexes (IgM-IC) following their intravenous (i.v.) administration [7,15] IgM-IC first localize to the splenic marginal zone (MZ), where they are phagocytized by MZ macrophages and are also bound to MZ B cells. Binding of IgM-IC to MZ B cells is strictly complement dependent. The MZ B cells subsequently transport the IgM-IC into the B cell follicles where they are deposited onto follicular dendritic cells (FDCs) [15]. The important role of MZ B cells in this process was established by monitoring IgM-IC deposition onto FDCs in mice that lacked MZ B cells, either as a result of a genetic lesion (CD19-deficient mice) that inhibited MZ B cell survival, or by their depletion using pertussis toxin. In these mice, little if any IgM-IC was deposited onto FDCs. In addition, adoptive transfer of B cells with bound IgM-IC into mice led to IgM-IC deposition onto FDCs. In these mice, transferred MZ B cells in close proximity to FDCs were found to have IgM-IC capped and directed toward the FDC dendrites, strongly suggesting IgM-IC was being transferred to FDCs by MZ B cells [15].
To further investigate the role of MZ B cells in IgM-IC transport, we used mice deficient in p50, a member of the NF-κB family of transcription factors [16]. These mice have been reported to lack MZ B cells, suggesting an unambiguous role for this component of the NF-κB transcription factor complex in the development of this subset of B cells [17]. However, we found that small numbers of MZ B cells are present in young p50-/- mice, and these cells are capable of binding high levels of IgM-IC in vivo and in vitro. Interestingly, MZ B cells were found to accumulate with age in p50-/- mice such that by 6 months a normal sized MZ B cell compartment was present. We also found that p50-/- mice exhibited an accelerated loss of FDC networks, as assessed by intense CD21 staining. These data indicate that MZ B cells can develop in the complete absence of p50, and suggest that the transcriptional requirements for the development and/or survival of MZ B cells might differ in young and old mice.
Results
Presence of MZ B cells in p50-deficient mice
Two month old p50-/- or control C57BL/6 (B6) mice were injected with IgM-IC and the spleens removed 16 h later. Based on the reported lack of MZ B cells in p50-/- mice [17] and the requirement for MZ B cells in the transport and deposition of IgM-IC onto FDCs [15], we expected that IgM-IC would not be found localized on FDCs in the splenic follicles of these mice. However, immunohistochemistry (IH) revealed the presence of IgM-IC on FDCs in p50-/- mice (Figure 1). The amount of IgM-IC localized in the spleen follicles of p50-/- mice was substantially less than in control B6 mice, but none-the-less readily detectable. These results could indicate either that IgM-IC can traffic to FDCs by a previously unanticipated mechanism independent of MZ B cells, or p50-/- mice may possess small numbers of MZ B cells that are still functional for transporting IgM-IC into follicles.
To help discriminate between these alternatives, splenic B cells from p50-/- mice were analyzed for the presence of MZ B cells by FACS. MZ B cells can be phenotypically distinguished from other B cell subsets in the spleen based on CD21/23 [18,19] and B220/CD1d [20,21]. profiles. As shown in Figure 2, MZ B cells were, as expected [17], severely reduced in p50-/- mice. However, small numbers of B cells fell within the CD21hi/CD23lo and B220+/CD1dhi gates, at ~10% of the frequency found in the B cell compartment of control B6 mice. To determine if these MZ B cells could bind high levels of IgM-IC, spleen cells from mice that had been injected 1 h earlier were isolated and the presence of IgM-IC on B cell subsets was determined by FACS. When IgM-IC were injected into B6 mice, the complexes were found associated with MZ B cells at high levels (Figure 2A), as expected from their high level expression of the complement receptor CR2 (CD21) and localization at the marginal sinus, which facilitates exposure to blood-borne IgM-IC [15]. In contrast, little if any of the IgM-IC were present on follicular (FO) B cells. In p50-/- mice, there was discernable binding of IgM-IC to MZ B cells (as defined by a shift in the FACS profile) while, as seen in B6 mice, IgM-IC did not bind to FO B cells (Figure 2A). This suggests that the cells in the CD21hi gate in p50-/- mice were authentic MZ B cells. To confirm this, the capacity of these cells to bind IgM-IC was determined using in vitro binding assays (Figure 2B). When p50-/- B cells were incubated with IgM-IC in vitro, MZ B cells bound higher levels of IgM-IC than did FO B cells (3 to 5 times as assessed by mean fluorescence intensity), similar to results using B cells from B6 mice.
Note that the spleens of p50-/- mice contain an increased number of CD21loCD23lo B cells (Fig. 2A; see also Fig. 3A). While we have not explored the identity of these cells in detail. they are likely to reflect increased numbers of immature "transitional" B cell subsets, since the number of AA4.1+ B cells [22] in the spleens of these mice are increased (up to 2-fold) over those in age-matched control B6 mice. As expected from their low levels of CD21, however, this population of cells does not bind IgM-IC (data not shown).
MZ B cells can be distinguished as IgMhi B cells positioned outside of the marginal sinus. To determine whether the MZ B cells in p50-/- mice were positioned correctly around the marginal sinus of the spleen, immunofluorescence (IF) was performed on spleen sections using anti-IgM and anti-MOMA-1, which stains marginal metallophilic macrophages and delineates the MZ [23]. In control B6 mice, IgMhi cells (in red) were present in the MZ, outside of the MOMA-1+ cells (Figure 2C). IgMhi MZ B cells were also found outside of the MOMA-1 ring in p50-/- mice, but in reduced numbers, as expected from the drastic reductions in MZ B cells in these mice (Figure 2A). Taken together, these data indicate that p50-/- mice contain small numbers of MZ B cells that are correctly positioned around the marginal sinus, where they can encounter blood-borne IgM-IC and transport them into splenic follicles for deposition onto FDCs.
Accumulation of MZ B cells with age in p50-deficient mice
When older p50-/- mice were analyzed for the presence of MZ B cells, increased frequencies of MZ B cells were detected. At 4 months of age, some p50-/- mice contained increased numbers of MZ B cells, whereas others did not, suggesting that these cells might be emerging around this time period (data not shown). By 6 months of age, a phenotypically distinct MZ B cell population was detected in all of the more than 10 mice analyzed, as determined by the CD21/23 profile. In these mice, MZ B cells were present at frequencies comparable to those found in B6 mice (Figure 3A, left panels), although the total number of splenic B cells did not appear to change. At all ages tested, p50-/- and B6 mice had similar frequencies of T and B cells in their spleens, and their spleen sizes were also similar (unpublished results), and the cellularity of the spleens (as determined by IH; see for example Fig. 5B) were also similar, indicating that the increased number of MZ B cells was selective for this subset of cells. To confirm the phenotypic identity of the CD21hi/CD23lo B cells as MZ B cells, the expression of other markers of MZ B cells, including CD1d [20] and CD9 [24], was assessed. As shown in Figure 4A, populations of B220+ B cells with CD1dhi and CD9hi phenotypes were present in p50-/- mice, consistent with the presence of MZ B cells. The frequencies of these cells were similar to those in aged-matched control B6 mice. We also assessed the levels of IgM on the CD21hiCD23lo cells, since MZ B cells express higher levels of IgM compared with FO B cells [25]. MZ B cells from p50-/- mice clearly expressed higher levels of IgM than on FO B cells from these same mice, and at levels similar to those in control mice (Figure 4B). These B cells were also positioned within the MZ in p50-/- mice (Figure 4C), as indicated by the presence of IgM+ cells outside of the marginal sinus as revealed by laminin staining [26].
As further evidence that the accumulating cells were functional MZ B cells, cells were tested for their capacity to bind IgM-IC in vivo and in vitro. MZ B cells from 2 and 6 month old B6 mice or from 6 month old p50-/- mice had high levels of IgM-IC bound following i.v. injection, whereas FO B cells bound little or no IgM-IC (Figure 3A). An in vitro binding assay also demonstrated the capacity of the MZ B cells from p50-/- mice to bind high levels of IgM-IC compared to FO B cells (Figure 3B). Taken together, the data demonstrate that while few MZ B cells are present in young p50-/- mice, they do develop and accumulate with age, such that by 6 months they have reached normal levels and can functionally bind IgM-IC at high levels, as would be expected from their increased expression of CR1/2 (CD21/35).
Accelerated loss of FDC networks in p50-deficient mice
In an effort to determine whether the repopulation of older p50-/- mice with MZ B cells enabled efficient IgM-IC trapping, mice were injected with IgM-IC and analyzed for FDC deposition. Despite the normal levels and positioning of MZ B cells, no IgM-IC deposition was observed in the spleens of 6 month old p50-/- mice (Figure 5A). The failure of the IgM-IC to localize in the follicles of these mice could be due to the absence of FDCs. However, as shown in Figure 5B, FDCs were present in these mice, as assessed by CD21 staining. However, they appeared to be much smaller and the dendrites less pronounced than in age-matched B6 mice, as evident from the diminished area of intense CD21 staining (Figure 5B, right panels; see arrows). In addition, the number of follicles with intense CD21 staining was reduced in the 6 month old p50-/- mice compared with age matched B6 mice (Figure 5B, 4× view in left panels). This suggests that the ability of FDCs from p50-/- mice to retain complement fragments might be diminished at 6 months of age. Interestingly, 6 month old B6 mice also exhibited some reduced capacity to trap IgM-IC (Figure 5A) and slightly reduced CD21 staining compared with younger mice (Figure 5B), suggesting a loss of FDC function may also be occurring naturally with age, as previously observed [27]. Consistent with this, FDCs in 6 month old p50-/- and B6 mice could not be detected by FDC-M2 staining (data not shown), in contrast with younger mice (Figure 1). FDC-M2 is a FDC specific antibody [28] that detects the complement component C4 on FDCs [29]. The data indicate the loss of FDC function may be accelerated in mice lacking p50. Together, the data imply that the lack of FDC localization of IgM-IC in 6 month old p50-deficient mice is not due to functional inactivity of the MZ B cells that have accumulated in these mice, but reflects a loss of FDCs and/or FDC function by this age.
Discussion
Our data demonstrate that while NF-κB p50-/- mice are clearly deficient in MZ B cells for the first 2 months of life, MZ B cells appear with age, such that by 6 months a normal splenic MZ B cell compartment has been reconstituted. There are at least four distinct but not mutually exclusive mechanisms that could account for the late appearance of cells within the MZ of p50-/- mice. First, these B cells could be memory B cells, which have been reported to accumulate in the MZ [30]. While we cannot formally rule out this possibility, B cells from p50-/- mice respond poorly to antigenic challenge, especially T-dependent antigens [16]. Consequently, it is unlikely that memory B cells would be generated in these mice, irrespective of their age. Alternatively, the emergence of MZ B cells in p50-/- mice could represent the accumulation of these cells through homeostatic proliferation, in which cells proliferate to fill niches in response to deficits in specific subsets of cells [31]. While splenic B cells from p50-/- mice will undergo homeostatic proliferation as a consequence of their transfer into SCID recipients [32], the ability of the few MZ B cells in the p50-/- donor spleen to expand following transfer was not examined [32]. Nevertheless, we view it as unlikely that homeostatic proliferation accounts for the accumulation of MZ B cells in the spleens of older p50-/- mice, since the accumulation of cells brought on by homeostatic proliferation is generally observed over a much shorter period of time, days to a couple of weeks, rather than the months required to reconstitute MZ B cells in p50-/- mice. A third possibility is that a NF-κB p50-dependent signal transduction pathway that activates a developmentally required transcription program in MZ B cells or their precursors is overcome by compensatory mechanisms with aging. NF-κB p50 is involved in a number of pathways crucial for B cell development and maturation, including B cell receptor (BCR) signalling and responses to members of the TNF family [33-35] Signal strength through the BCR has been proposed to be an important determinant of B cell lineage fate decisions, including the MZ B cell pool [36,37] BAFF signalling is also important for the selection of B cells into the MZ B cell pool [38], while an intact LT-β R signalling pathway is required for the proper positioning of these cells [39]. The accumulation of MZ B cells in older p50-/- mice could reflect compensatory mechanisms in one or more of these signal transduction pathways important for the emergence or maintenance of MZ B cells. A fourth possibility is that the small numbers of MZ B cells that successfully emerge in p50-/- mice survive for extended periods of time, resulting in the slow accumulation of this subset of cells with age. Hao and Rajewsky [40] have demonstrated that the half-life of MZ B cells is greater than that of FO B cells. Thus, a simple explanation for the accumulation of MZ B cells in p50-/- mice with age is that these cells preferentially survive, and thus their frequency relative to the FO B cell pool and other B cells in the spleen increases slowly with age.
B cells from p50-/- mice are compromised in their ability to respond to specific antigen challenge, and are completely unresponsive to mitogenic substances such as LPS [16]. However, our data suggests that MZ B cells from p50-/- mice display some functionality, including the binding, transport and deposition of IgM-IC onto FDCs, a function unique to MZ B cells [15]. Like IgM-containing T-dependent immune complexes, MZ B cells also bind T-independent type 2 (TI-2) antigens in a complement and CR1/2-dependent manner [41]. Despite the fact that p50-/- mice are widely considered to be devoid of MZ B cells, the presence of small numbers of these B cells in p50-/- mice were in fact noted in the initial studies of Pillai and colleagues, and these appeared to bind low levels of TI-2 antigens [17]. However, further analysis of the consequences of this binding, such as the transfer of immune complexes to FDCs, was not evaluated. The results of the current study not only extend these previous studies, but demonstrate that one consequence of the small amount of IgM-IC binding manifests itself by the limited, but none-the-less detectable, deposition of these complexes onto FDCs. This suggests that this function of MZ B cells is completely independent of the activation of NF-κB, at least those complexes containing p50. These results extend other evidence suggesting that the activation of MZ B cells through engagement of the BCR may not be required for transport and deposition of IgM-IC, since MZ B cells from CD19-deficient mice, which also have deficits in MZ B cells, are also functional in this assay [15].
Together, these results raise the question of whether MZ B cells require activating signals to migrate into B cell follicles following IgM-IC binding to CR1/2. MZ B cell migration can be detected in 1 h or less after the binding of IgM-IC [15]. The binding of IgM-IC to these cells may interfere with one or more of the mechanisms proposed to play a role in MZ B cell positioning, including retention by macrophages within the MZ [42], integrin-mediated retention [39], or responsiveness to lysophospholipids in the blood [43]. Once released, the MZ B cells would be expected to migrate into follicles in response to chemoattractants such as BLC [44]. These considerations suggest that follicular migration of MZ B cells may be the "default" pathway, whereas BCR engagement, which is known to direct MZ B cells not into the follicles but to the T-B interface [45-47], actively alters the fate of antigen specific MZ cells away from follicular migration.
Despite the increase in frequency (and numbers) of MZ B cells in p50-/- mice, the deposition of IgM-IC within the follicles does not improve with age, but declines severely. This appears to be due to a decrease in function of FDCs, which are fewer in number and are less pronounced, as evidenced by CD21 staining (Figure 5). Thus, the decreased ability of FDCs to retain complement containing immune complexes observed in old mice [27], appear to emerge with accelerated kinetics in p50-/- mice.
NF-κB p50 is only one of a growing number of genes, including a number of transcription factors, whose expression influences the development of MZ B cells (reviewed in [48]), highlighting the complex interactions that must converge to generate this B cell compartment. The fact that the requirements for individual proteins, such as p50, could change with age suggests that compensatory mechanisms exist which can overcome cellular defects, and/or that more than one pathway to generate these cells may exist.
Conclusion
These experiments demonstrate the profound influence age can exert on cellular populations in genetically modified mice. The appearance of large numbers of MZ B cells in 6 month old p50-/- mice suggests that compensatory mechanisms overcome the deficits caused by p50-deficiency. Both in young and old p50-/- mice, these MZ B cells appear to retain some functionality, as assessed by their ability to bind, transport and deposit IgM-IC onto FDCs. On the other hand, FDCs are unable to maintain their functional capacity to trap IgM-IC in aged p50-/- mice indicating that, while MZ B cells recover with age, other immune defects become more pronounced as a result of deficiency in this member of the NF-κB family of transcription factors.
Methods
Mice
C57BL/6J and NF-κB p50 deficient (B6;129P2-Nfkb1tm1Bal) [16] mice were purchased from Jackson Laboratories (Bar Harbor, ME). Mice were maintained in the Laboratory Animal Sciences Center, Boston University Medical Center. All studies were reviewed by an Institutional Animal Care and Use Committee and carried out in accordance with established guidelines according to National Institutes of Health. Mice of both sexes were used at 2 to 6 months of age as described.
Preparation and injection of IgM-containing immune complexes (IgM-IC)
Mice were injected i.v. with a mixture of 15 to 20 μg each of NP12-BSA-biotin (Biosearch Technologies, Novato, CA) and affinity purified pentameric IgM anti-NP monoclonal antibody (B1-8) exactly as previously described [7,15].
In vitro binding assay
In vitro assays to assess the binding of IgM-IC to MZ B cells were performed as previously described [15]. Briefly, IgM-IC were formed by mixing 1 μg each of NP-BSA-biotin and B1-8 IgM in RPMI in the presence of 10% fresh mouse serum as a source of complement and incubated for 30 minutes at 37°C. Splenocytes (107) were resuspended in the IgM-IC and incubated at 37°C for an additional 30 minutes. Cells were washed and analyzed by flow cytometry.
Immunohistochemistry (IH)
Staining of spleen sections was performed exactly as previously described [7,15]. Briefly, spleens were removed and snap frozen in O.C.T. Compound (Tissue Tek, Torrance, CA) and stored at -80°C. 6 μm sections were cut (IEC Microtome, Needham, MA) and fixed in acetone at -20°C for 10 min. Sections were incubated with 0.3% H2O2/PBS for 5 min and blocked with 3% normal rabbit or rat serum. All antibodies were diluted in 5% BSA in PBS. Primary antibodies used included: anti-CD3-FITC (1/200), anti-CD3-biotin (1/200), anti-CD21/35-FITC (1/100), (all from BD Biosciences, San Diego), ER-TR9 (1/200; BMA Biomedical, Augst, Switzerland), and anti-FDC-M2 (1/200; Immunokontact, Frankfurt, Germany). Secondary antibodies included anti-FITC-HRP (1/100; Roche Molecular Biochemicals, Indianapolis, IN), goat anti-rat IgG-HRP (1/500; Jackson Immunoresearch, West Grove, PA), and goat anti-rat IgM (Fab')2-HRP (1/1000; Jackson Immunoresearch). NP12-BSA-biotin and anti-CD3-biotin was visualized by alkaline phosphatase (AP)-streptavidin (1/400; Molecular Probes, Eugene, OR) and developed with napthol AS-MX phosphate/Fast blue BB base (Sigma, St. Louis, MO). Visualization of antibody stains was accomplished using horse radish peroxidase (HRP) conjugated secondary antibodies developed with 3-amino-9-ethylcarbazole (Sigma) and H202. After staining, sections were covered with crystal mount (Biomedia, Foster City, CA) before examining by digital microscopy (Zeiss, Axiovert 200 M, Oberkochen, Germany). Unless otherwise noted, all images were obtained using a 10× objective (initial magnification of 100×) unless otherwise noted (see Fig. 4).
Immunofluorescence (IF)
Sections were initially treated as for IH. Antibodies utilized included anti-MOMA-1 (1/25; Serotec, Raleigh NC) visualized with a goat anti-rat IgG-AMCA (1/100; Jackson Immunoresearch), anti-IgM-Texas Red (1/100; BD Biosciences), and anti-laminin (1/500; Sigma) visualized with an anti-rabbit IgG-FITC (1/100; Jackson Immunoresearch). All sections were covered with Fluoromount G (Jackson Immunoresearch) and a cover slip before examining by digital microscopy (Zeiss, Axiovert 200 M, Oberkochen, Germany).
Flow cytometry
Splenocytes were isolated at the indicated times and stained as follows. The lymphocyte gate was utilized for all analysis. Cells were analyzed using a FACScan (Becton Dickinson, Franklin Lakes, NJ) and FlowJo software (Tree Star, Ashland, OR). Antibodies utilized were anti-B220-FITC (1/200), anti-B220-biotin (1/200), anti-B220-PE (1/200), anti-CD21/35-FITC (1/200), anti-CD23-PE (1/200), and anti-CD1d-FITC or PE (1/100) (BD Biosciences). Streptavidin-Cy5 (1/200; BD Biosciences) was used to identify the NP antigen (NP12-BSA-biotin) associated with lymphocytes, as well as to visualize biotin-labelled antibodies.
Authors' contributions
A.R.F. performed the experiments reported in this manuscript and participated in the data interpretation and analysis. R.B.C. originally conceived of the study, participated in data interpretation and analysis, and takes overall responsibility for the conduct of the studies. Both authors have read and approve of the final manuscript.
Acknowledgements
The authors would like to thank members of our laboratory for thoughtful discussions and K. Daley for critical reading of the manuscript. This work was supported by USPHS grant AI31209 from the National Institutes of Health, and the Aid for Cancer Research for instrumentation grants. A.R.F. was supported in part by NIH training grants AI07903 and HL007501.
Figures and Tables
Figure 1 p50-/- mice trap IgM-IC less efficiently on FDCs. Localization of IgM-IC in B6 or p50-/- mice was determined by IH 16 h after injection. ER-TR9 identifies marginal zone macrophages (MZM) and FDC-M2 identifies FDCs. IgM-IC (which appear blue in this and all subsequent figures and are identified by arrows) are localized to FDCs in spleens from both groups of mice, but FDCs from B6 mice show more intense staining, indicating that more IgM-IC are trapped in these mice. All IH stains were performed on serial sections and were developed side-by-side for identical periods of time.
Figure 2 p50-/- mice contain reduced numbers of MZ B cells that bind high levels of IgM-IC A. B cell subsets and IgM-IC binding. Splenocytes from B6 or p50-/- mice were analyzed for B cell subsets by FACS using CD21/CD23 or B220/CD1d expression. Histograms on the right show binding of IgM-IC to MZ and FO B cells 1 h after i.v. injection (shaded region); clear profiles represent uninjected control mice. Although spleen cells from p50-/- mice contained reduced numbers of MZ B cells, the cells present in the MZ B cell gate bind IgM-IC at high levels over background, similar to levels of binding of IgM-IC to B6 MZ B cells. B. In vitro binding assay showing binding of IgM-IC to MZ B cells (shaded region) compared with FO B cells (clear region). MZ B cells from both B6 and p50-/- mice bind high levels of IgM-IC compared to FO B cells. C. Immunofluorescence shows the presence of IgMhi cells (red) in the MZ as demarcated by MOMA-1 (marginal metallophilic macrophages, blue) staining. p50-/- mice possess some IgM+ cells in the MZ outside of the MOMA-1 ring, consistent with the presence of appropriately positioned MZ B cells in these mice.
Figure 3 p50-/- mice accumulate MZ B cells with age that have the capacity to bind high levels of IgM-IC A. FACS profile of splenocytes from 2 and 6 month old B6 and 6 month p50-/- mice reveal the presence of normal numbers of MZ B cells in spleens of the p50-/- mice. Histograms illustrate binding of IgM-IC to p50-/- MZ B cells as described in Figure 2A. B. In vitro IgM-IC binding assay. Binding to MZ B cells (shaded profiles); binding to FO B cells (clear profiles).
Figure 4 Phenotypic and histological analysis of MZ B cells in p50-/- mice A. FACS profile of splenocytes from 2 and 6 month old B6 and 6 month old p50-/- mice revealed by B220/CD1d and B220/CD9 profiles. B. FACS histograms showing IgM levels on MZ and FO B cells (shaded and clear regions, respectively) as determined by CD21/23 profiles. Cells present in the MZ B cell gate in p50-/- mice express higher levels of IgM compared with FO B cells, similar to B6 mice. C. IF reveals well demarcated IgMhi cells (in red) in the MZ outside of the laminin staining (which identifies the marginal sinus; in green) in 6 month p50-/- mice.
Figure 5 Deteriorated FDCs in aged p50-/- mice correlates with reduced IgM-IC localization A. 2 month and 6 month old B6 mice and 6 month old p50-/- mice were injected with IgM-IC and analyzed by IH for IgM-IC localization 16 h later. IgM-IC were present on FDCs in 2 and 6 month B6 mice (see arrows), but were not deposited onto FDCs in p50-/- mice. B. Intense CD21 staining (designated by arrows) was less visible in the spleens of p50-/- mice, indicating that FDC networks were poorly developed in these mice. This suggests that reduced IgM-IC localization may be attributed to the loss of functional FDC networks in p50-/- mice. The magnifications noted refer to the objective used, not the absolute magnification.
==== Refs
Henry C Jerne NK Competition of 19S and 7S antigen receptors in the regulation of the primary immune response J Exp Med 1968 128 133 152 5662011 10.1084/jem.128.1.133
Dennert G The mechanism of antibody-induced stimulation and inhibition of the immune response J Immunol 1971 106 951 995 4928749
Heyman B Andrighetto G Wigzell H Antigen-dependent IgM-mediated enhancement of the sheep erythrocyte response in mice. Evidence for induction of B cells with specificities other than that of the injected antibodies J Exp Med 1982 155 994 1009 6174678 10.1084/jem.155.4.994
Harte PG Cooke A Playfair JHL Specific monoclonal IgM is a potent adjuvant in murine malaria vaccination Nature 1983 302 256 258 6835362 10.1038/302256a0
Heyman B Wigzell H Specific IgM enhances and IgG inhibits the induction of immunological memory in mice Scand J Immunol 1985 21 255 266 3992195
Basalp A Cirakoglu B Bermek E Enhancement of the immune response to hepatitis B virus vaccine by antigen specific IgM Immunol Lett 2000 73 1 6 10963803 10.1016/S0165-2478(00)00193-0
Youd ME Ferguson AR Corley RB Synergistic roles of IgM and complement in antigen trapping and follicular localization Eur J Immunol 2002 32 2328 2337 12209646 10.1002/1521-4141(200208)32:8<2328::AID-IMMU2328>3.0.CO;2-T
Heyman B Pilström L Shulman MJ Complement activation is required for IgM-mediated enhancement of the antibody response J Exp Med 1988 167 1999 2004 3385360 10.1084/jem.167.6.1999
Applequist SE Dahlstrom J Jiang N Molina H Heyman B Antibody production in mice deficient for complement receptors 1 and 2 can be induced by IgG/Ag and IgE/Ag, but not IgM/Ag complexes J Immunol 2000 165 2398 2403 10946263
Boes M Esau C Fischer MB Schmidt T Carroll M Chen J Enhanced B-1 cell development, but impaired IgG antibody responses in mice deficient in s ecreted IgM J Immunol 1998 160 4776 4787 9590224
Boes M Prodeus AP Schmidt T Carroll MC Chen J A critical role of natural immunoglobulin M in immediate defense against systemic bacterial infection J Exp Med 1998 188 2381 2386 9858525 10.1084/jem.188.12.2381
Ehrenstein MR O'Keefe TL Davies SL Neuberger MS Targeted gene disruption reveals a role for natural secretory IgM in the maturation of the primary immune response Proc Natl Acad Sci USA 1998 95 10089 10093 9707605 10.1073/pnas.95.17.10089
Baumgarth N Herman OC Jager GC Brown LE Herzenberg LA Chen J B-1 and B-2 cell-derived immunoglobulin M antibodies are nonredundant components of the protective response to influenza virus infection J Exp Med 2000 192 271 280 10899913 10.1084/jem.192.2.271
Ochsenbein AF Fehr T Lutz C Suter M Brombacher F Hengartner H Zinkernagel RM Control of early viral and bacterial distribution and disease by natural antibodies Science 1999 286 2156 2159 10591647 10.1126/science.286.5447.2156
Ferguson AR Youd ME Corley RB Marginal zone B cells transport and deposit IgM-containing immune complexes onto follicular dendritic cells Int Immunol 2004 16 1411 1422 15326094 10.1093/intimm/dxh142
Sha WC Liou HC Tuomanen EI Baltimore D Targeted disruption of the p50 subunit of NF-κB leads to multifocal defects in immune responses Cell 1995 80 321 330 7834752 10.1016/0092-8674(95)90415-8
Cariappa A Liou HC Horwitz BH Pillai S Nuclear factor κB is required for the development of marginal zone B lymphocytes J Exp Med 2000 192 1175 1182 11034607 10.1084/jem.192.8.1175
Oliver AM Martin F Gartland GL Carter RH Kearney JF Marginal zone B cells exhibit unique activation, proliferative and immunoglobulin secretory responses Eur J Immunol 1997 27 2366 2374 9341782
Takahashi K Kozono Y Waldschmidt TJ Berthiaume D Quigg RJ Baron A Holers VM Mouse complement receptors type 1 (CR1;CD35) and type 2 (CR2;CD21): expression on normal B cell subpopulations and decreased levels during the development of autoimmunity in MRL/lpr mice J Immunol 1997 159 1557 1569 9233655
Amano M Baumgarth N Dick MD Brossay L Kronenberg M Herzenberg LA Strober S CD1 expression defines subsets of follicular and marginal zone B cells in the spleen: beta 2-microglobulin-dependent and independent forms J Immunol 1998 161 1710 1717 9712035
Makowska A Faizunnessa NN Anderson P Midtvedt T Cardell S CD1high B cells: a population of mixed origin Eur J Immunol 1999 29 3285 3294 10540340 10.1002/(SICI)1521-4141(199910)29:10<3285::AID-IMMU3285>3.3.CO;2-G
Allman D Lindsley RC DeMuth W Rudd K Shinton SA Hardy RR Resolution of three nonproliferative immature splenic B cell subsets reveals multiple selection points during peripheral B cell maturation J Immunol 2001 167 6834 6840 11739500
Kraal G Janse M Marginal metallophilic cells of the mouse spleen identified by a monoclonal antibody Immunology 1986 58 665 669 3733156
Won WJ Kearney JF CD9 is a unique marker for marginal zone B cells, B1 cells, and plasma cells in mice J Immunol 2002 168 5605 5611 12023357
Kumararatne DS MacLennan IC Cells of the marginal zone of the spleen are lymphocytes derived from recirculating precursors Eur J Immunol 1981 11 865 869 6976895
Wang SL Kutsche M DiSciullo G Schachner M Bogen SA Selective malformation of the splenic white pulp border in L1-deficient mice J Immunol 2000 165 2465 2473 10946272
Aydar Y Balogh P Tew JG Szakal AK Follicular dendritic cells in aging, a "bottle-neck" in the humoral immune response Ageing Res Rev 2004 3 15 29 15163101 10.1016/j.arr.2003.08.002
Kosco-Vilbois MH Zentgraf H Gerdes J Bonnefoy JY To 'B' or not to 'B' a germinal center? Immunol Today 1997 18 225 230 9153954 10.1016/S0167-5699(97)01048-7
Taylor PR Pickering MC Kosco-Vilbois MH Walport MJ Botto M Gordon S Martinez-Pomares L The follicular dendritic cell restricted epitope, FDC-M2, is complement C4; localization of immune complexes in mouse tissues Eur J Immunol 2002 32 1888 1896 12115608
Liu YJ Oldfield S MacLennan IC Memory B cells in T cell-dependent antibody responses colonize the splenic marginal zones Eur J Immunol 1988 18 355 362 3258564
Surh CD Sprent J Regulation of naive and memory T-cell homeostasis Microbes Infect 2002 4 51 56 11825775 10.1016/S1286-4579(01)01509-X
Cabatingan MS Schmidt MR Sen R Woodland RT Naive B lymphocytes undergo homeostatic proliferation in response to B cell deficit J Immunol 2002 169 6795 6805 12471111
Gold MR Ingham RJ McLeod SJ Christian SL Scheid MP Duronio V Santos L Matsuuchi L Targets of B-cell antigen receptor signaling: the phosphatidylinositol 3-kinase/Akt/glycogen synthase kinase-3 signaling pathway and the Rap1 GTPase Immunol Rev 2000 176 47 68 11043767 10.1034/j.1600-065X.2000.00601.x
Gugasyan R Grumont R Grossmann M Nakamura Y Pohl T Nesic D Gerondakis S Rel/NF-κB transcription factors: key mediators of B-cell activation Immunol Rev 2000 176 134 410 11043773 10.1034/j.1600-065X.2000.00615.x
Aggarwal BB Signalling pathways of the TNF superfamily: a double-edged sword Nat Rev Immunol 2003 3 745 756 12949498 10.1038/nri1184
Cariappa A Tang M Parng C Nebelitskiy E Carroll M Georgopoulos K Pillai S The follicular versus marginal zone B lymphocyte cell fate decision is regulated by Aiolos, Btk, and CD21 Immunity 2001 14 603 615 11371362 10.1016/S1074-7613(01)00135-2
Casola S Otipoby KL Alimzhanov M Humme S Uyttersprot N Kutok JL Carroll MC Rajewsky K B cell receptor signal strength determines B cell fate Nat Immunol 2004 5 317 327 14758357 10.1038/ni1036
Mackay F Woodcock SA Lawton P Ambrose C Baetscher M Schneider P Tschopp J Browning JL Mice transgenic for BAFF develop lymphocytic disorders along with autoimmune manifestations J Exp Med 1999 190 1697 1710 10587360 10.1084/jem.190.11.1697
Lu TT Cyster JG Integrin-mediated long-term B cell retention in the splenic marginal zone Science 2002 297 409 412 12130787 10.1126/science.1071632
Hao Z Rajewsky K Homeostasis of peripheral B cells in the absence of B cell influx from the bone marrow J Exp Med 2001 194 1151 1164 11602643 10.1084/jem.194.8.1151
Guinamard R Okigaki M Schlessinger J Ravetch JV Absence of marginal zone B cells in Pyk-2-deficient mice defines their role in the humoral response Nat Immunol 2000 1 31 36 10881171 10.1038/76882
Karlsson MC Guinamard R Bolland S Sankala M Steinman RM Ravetch JV Macrophages control the retention and trafficking of B lymphocytes in the splenic marginal zone J Exp Med 2003 198 333 340 12874264 10.1084/jem.20030684
Cinamon G Matloubian M Lesneski MJ Xu Y Low C Lu T Proia RL Cyster JG Sphingosine 1-phosphate receptor 1 promotes B cell localization in the splenic marginal zone Nat Immunol 2004 5 713 20 15184895 10.1038/ni1083
Cyster JG Ansel KM Reif K Ekland EH Hyman PL Tang HL Luther SA Ngo VN Follicular stromal cells and lymphocyte homing to follicles Immunol Rev 2000 176 181 193 11043777 10.1034/j.1600-065X.2000.00618.x
Liu YJ Zhang J Lane PJ Chan EY MacLennan IC Sites of specific B cell activation in primary and secondary responses to T cell-dependent and T cell-independent antigens Eur J Immunol 1991 21 2951 2962 1748148
Martin F Oliver AM Kearney JF Marginal zone and B1 B cells unite in the early response against T-independent blood-borne particulate antigens Immunity 2001 14 617 629 11371363 10.1016/S1074-7613(01)00129-7
Song H Cerny J Functional heterogeneity of marginal zone B cells revealed by their ability to generate both early antibody-forming cells and germinal centers with hypermutation and memory in response to a T-dependent antigen J Exp Med 2003 198 1923 1935 14662910 10.1084/jem.20031498
Martin F Kearney JF Marginal-zone B cells Nat Rev Immunol 2002 2 323 335 12033738 10.1038/nri799
| 15836790 | PMC1087843 | CC BY | 2021-01-04 16:03:31 | no | BMC Immunol. 2005 Apr 18; 6:8 | utf-8 | BMC Immunol | 2,005 | 10.1186/1471-2172-6-8 | oa_comm |
==== Front
BMC Med GenetBMC Medical Genetics1471-2350BioMed Central London 1471-2350-6-151584273510.1186/1471-2350-6-15Technical AdvanceDetection of large deletions in the LDL receptor gene with quantitative PCR methods Damgaard Dorte [email protected] Peter H [email protected] Lillian G [email protected] Gitte G [email protected] Anette [email protected] Mogens L [email protected] Ole [email protected] Department of Medicine and Cardiology, Aarhus Sygehus, Aarhus University Hospital, Tage Hansens Gade 2, 8000 Aarhus C, Denmark2 Department of Clinical Biochemistry, Aarhus Sygehus, Aarhus University Hospital, Aarhus, Denmark3 Department of Clinical Genetics, Aarhus Sygehus, Aarhus University Hospital, Aarhus, Denmark2005 20 4 2005 6 15 15 21 1 2005 20 4 2005 Copyright © 2005 Damgaard 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
Familial Hypercholesterolemia (FH) is a common genetic disease and at the molecular level most often due to mutations in the LDL receptor gene. In genetically heterogeneous populations, major structural rearrangements account for about 5% of patients with LDL receptor gene mutations.
Methods
In this study we tested the ability of two different quantitative PCR methods, i.e. Real-Time PCR and Multiplex Ligation-Dependent Probe Amplification (MLPA), to detect deletions in the LDL receptor gene. We also reassessed the contribution of major structural rearrangements to the mutational spectrum of the LDL receptor gene in Denmark.
Results
With both methods it was possible to discriminate between one and two copies of the LDL receptor gene exon 5, but the MLPA method was cheaper, and it was far more accurate and precise than Real-Time PCR. In five of 318 patients with an FH phenotype, MLPA analysis revealed five different deletions in the LDL receptor gene.
Conclusion
The MLPA method was accurate, precise and at the same time effective in screening a large number of FH patients for large deletions in the LDL receptor gene.
==== Body
Background
Familial Hypercholesterolemia (FH) is clinically characterised by elevated concentrations of LDL-cholesterol in plasma, tendon xanthomas and an increased risk of developing coronary artery disease inherited in an autosomal dominant manner [1]. At the molecular level FH is most commonly due to mutations in the LDL receptor gene, in which more than 900 different mutations in the LDL receptor gene have been reported worldwide [2,3]. Approximately 90 of them are various major structural rearrangements such as deletions and insertions of sizes ranging from 37 bp to 25 kb [2,3]. Major structural rearrangements account for approximately 5% of identified mutations in the LDL receptor gene in genetically heterogeneous populations [1,4,5]. In Denmark four different deletions in the LDL receptor gene have been described [6-8]. They were found in four of 97 patients tested [8]. Southern blotting, followed by hybridisation with LDLR probes, has been used to screen for major structural rearrangements [6], but this method is labour-intensive and requires large amounts of DNA. Alternatively long-range PCR analysis has been used to amplify segments of the LDL receptor gene up to approximately 20 kb in lengths [9,10]. The major drawback of long-range PCR is that the deletion or duplication has to be within the borders of the primers of the amplified segment, because the allele containing the deletion or duplication will otherwise not be amplified, and the analysis result will falsely appear to be homozygous wild-type (normal). With quantitative PCR methods, it is possible to relate the number of allele copies to that of appropriate controls independently of the extent of the deletion /insertion. Only small amounts of DNA are required, and the methods are faster than Southern blotting analysis.
This study aimed at testing and comparing the abilities of two quantitative PCR methods, i.e. Real-Time PCR and Multiplex Ligation-dependent Probe Amplification (MLPA) for detection of deletions in the LDL receptor gene in a group of patients in whom deletions in the LDL receptor gene had been demonstrated by Southern blotting or long-range PCR analyses. If one of the methods proved valid and effective, the study also aimed at re-estimating the relative contribution of major structural rearrangements to the spectrum of LDL receptor gene mutations in FH patients in Denmark.
Results
For testing and comparing Real-Time PCR and MLPA, seventeen patients with a deletion of LDL receptor gene exon 5 were used as positive controls. The deletion of exon 5 was due either to a 9 kb deletion of exons 3–6 (14 patients from one family) or to a 1 kb deletion only of exon 5 (three patients from two families) described by Rüdiger et al and Jensen et al [6,8]. These mutations had been demonstrated earlier by either Southern blotting or long-range PCR analysis. Twelve individuals from the family with the 9 kb deletion that have been tested negative for the deletion were used as negative controls.
Real-Time PCR
Real-Time PCR is a quantitative PCR method, which quantifies the molecular concentration of DNA that can serve as template for amplification [11]. The amplification progress was measured by binding of SYBR Green to double stranded DNA. SYBR Green is an intercalating dye, the fluorescence of which is higher in the bound than in the free-state [12]. The fluorescence signal is measured real-time in the extension phase of the PCR reaction cycle, and the measurement, proportional to the amount of double stranded DNA, is plotted as an amplification curve against cycle number. A threshold value of fluorescence in the exponential part of the amplification curve is selected, and for each sample the number of cycles needed for the signal to reach the threshold is measured (threshold cycle (CT)).
Quantification of the molecular concentration of template DNA was performed with the standard curve method [13]. Data analysis was performed with the iCycler™ iQ Optical System Software Version 3.0a (BIO-RAD), and the results were exported to Excel sheets for storing and further processing. The copy number of exon 5 in the LDL receptor gene was divided by the copy number of the reference gene (albumin). The ratio is 2:2 (i.e. 1) if both alleles of both genes are present, and the ratio is 1:2 (i.e. 0.5) if one of the alleles of the target gene is absent. In contrast, the ratio is 3:2 (i.e. 1.5) if one of the alleles of the target gene has been duplicated.
Multiplex Ligation-dependent Probe Amplification (MLPA)
With the MLPA method developed by MRC-Holland [14], relative quantification of up to 40 different DNA sequences in one reaction is possible. Each MLPA probe consists of two fluorescently labelled oligonucleotides that can hybridise, adjacent to each other, to a target gene sequence. When hybridised, the two oligonucleotides are joined by a ligase, and the probe can then be amplified by PCR. To each probe is attached a set of universal primers, and one of the oligonucletides contains a stuffer sequence of variable length that enables us to separate the single fragments according to their length by gel electrophoresis [14,15]. The area under peak for each fragment was measured with the GeneScan Analysis Software Version 3.1.2 and Genotyper Software Version 2.5 (Applied Biosystems) and exported to Excel sheets for storing and for further processing. The peak area was normalised by dividing it by the combined area of all peaks in that lane. This normalised peak area was then divided by average normalised peak area from five normal control subjects. With this method the results are given as allele copy numbers as compared to normal controls, and a ratio of 1 is obtained if both alleles are present, a ratio of 0.5 if one allele is absent, and a ratio of 1.5 if one allele is duplicated.
A comparison of Real-Time PCR to MLPA for detection of known deletions of LDL receptor gene exon 5 is given in Table 1 and illustrated in Figure 1. With both Real-Time PCR and MLPA analysis, it was possible to distinguish between one and two copies of the LDL receptor gene exon 5, but the coefficients of variation were larger for the Real-Time PCR method than for the MLPA analysis. The results of the MLPA analysis of 16 out of 18 exons in 14 patients with a 9 kb deletion of exons 3–6 are shown in Figure 2 as wells as the results for 12 negative controls.
A cost-analysis and comparison of Real-Time PCR analysis and MLPA analysis are given in Table 2. The cost analysis included costs of reagents and laboratory technician time. When 10 samples were analysed together, the MLPA method was considerably cheaper as well as faster than the Real-Time PCR method.
To expand our assessment of the contribution of major structural rearrangements in the LDL receptor gene to the spectrum of mutations causing FH [8], we further studied, by MLPA analysis, 318 patients with an FH phenotype referred for molecular genetic analysis to Aarhus Sygehus, Aarhus University Hospital in the period January 1995 to June 2004, in whom no mutations in the LDL receptor gene had been detected by SSCP analysis and in whom the apoB R3500Q mutation also had been excluded. Two-hundred and seventy-six of these 318 patients have been described previously [16].
A deletion of the LDL receptor gene was found in five patients tested with the MLPA method. The five deletions were a 9.3 kb deletion of the promoter and exon 1 [16], a 1 kb deletion of exon 5 [6], a 3 kb deletion of exons 7–8 [16,17], a 9.5 kb deletion of exons 9–14 [16] and 5 kb deletion of exons 13–15 [17]. The graphical results of the MLPA analysis are given in Figure 3. The detection of the deletions was confirmed by long-range PCR analysis with amplification of segments that MLPA analysis had suggested might contain the deletion (data not shown).
Discussion
This study showed that the MLPA method unambiguously discriminates between one and two copies of LDL receptor gene exon 5. MPLA also enabled us to distinguish clearly between deleted and non-deleted exons in patients with a known 9 kb deletion of exons 3–6. Finally the MLPA method proved efficient in screening of large numbers of patients for major structural rearrangements of almost the whole LDL receptor gene in one PCR reaction. The protocol described by MRC Holland was followed, and no further optimisation was necessary.
It was also possible to distinguish between carriers and non-carriers of the exon 5 deletions with the Real-Time PCR method. The coefficients of variations (CV) for measurement of relative copy numbers with the Real-Time PCR method were larger than with the MLPA method, however. The CV's found in this study are comparable to other studies of measurements of relative allele copy numbers with Real-Time PCR and the use of SYBR green fluorescence in other genes [18,19], but higher than coefficients of variation reported with the use of TaqMan probes in Real-Time PCR analysis [20,21]. This difference could be due to multiplexing the PCR reactions when TaqMan probes are used such that both test gene and reference gene are amplified in the same reaction, which is not possible when SYBR Green fluorescence is used. It might have been possible to reduce the coefficients of variation of the Real-Time PCR analysis, with further optimisation or test of more primer sets. However, with the present data the intervals of mean ± 2SD's of relative allele copy numbers in carriers and non-carriers are overlapping, and in a considerably number of cases it was not possible to distinguish between one and two copies of exon 5 with the same degree of certainty as with the MLPA method. If screening of the entire LDL receptor gene was to be performed, a segment representative of each exon had to be amplified, and the need of optimisation would have been extensive. That would also have been the case if TaqMan probes were used and the PCR reactions were multiplexed.
Thus, in our hands, the MLPA method was cheaper, and it was more accurate and more precise than Real-Time PCR.
Previous work has not identified major structural rearrangements in the LDL receptor gene as a common cause of FH in Denmark [8]. It has therefore been necessary to employ a method for screening of the whole LDL receptor gene for pathogenic variations. In the period January 1995 to June 2004, a mutation in the LDL receptor gene was identified in 162 patients (data not shown) and large deletions accounted for 3.1% (5 of 162) of these LDL receptor gene mutation carriers, a finding similar to that obtained in other genetically heterogeneous populations [1,4,5]. In a study of Wang et al [22] eight different abnormal patterns of MLPA analysis were found in 12 of 21 patients with a clinical diagnosis of FH in whom sequencing of the LDL receptor gene had not revealed any mutations. In five patients with the same abnormal pattern, this was confirmed as a deletion with another method.
Conclusion
The MLPA method was accurate, precise and at the same time effective in screening a large number of patients for large deletions in the LDL receptor gene. Large deletions accounted for the 3.1 % of LDL receptor mutation carriers in the population studied.
Methods
DNA
Genomic DNA was extracted from EDTA stabilised blood with the PUREGENE Genomic DNA Purification Kit (Gentra Systems).
Real-Time PCR
A fragment of 116 bp located in exon 5 of the LDL receptor gene was amplified with the following primer set: forward exon 5 primer: 5'CCCTGCTTGTTTTTCTCTGG 3' and reverse exon 5 primer: 5'TGCAGTTTCCATCAGAGCAC 3'. For relative quantification of number of alleles in exon 5 of the LDL receptor gene, we used albumin as an internal reference gene with the primers described by Laurendeau et al [20]. PCR was carried out in reaction volumes of 25 μL with 12.5 μL of iQ™ SYBR Green Supermix (BIO-RAD), 300 nM of each primer and 4 ng of DNA. All samples were analysed in triplicate, and each run included separate standard curves for both primer pairs resulting from the amplification of serially diluted (50 ng, 10 ng, 2 ng, 0.4 ng and 0.08 ng) control DNA. Thermal cycling was performed on the iCycler™ iQ system (BIO-RAD) with a first denaturation step of 90 s at 95°C, followed by 40 cycles at 94°C for 10 s, 61°C for 20 s and 72°C for 20 seconds. PCR efficiencies in both reactions (exon 5 and albumin) were approximately 90%. Melting curve analysis was performed to exclude amplification of non-specific products.
Multiplex Ligation-dependent Probe Amplification (MLPA)
A kit for screening the LDL receptor gene for deletions or duplications was obtained from MRC-Holland (SALSA P062 LDLR exon deletion test kit). It contains probes for 16 out of 18 exons in the LDL receptor gene as well as two probes for genes located just upstream and downstream of the LDL receptor gene. It also contains 13 probes for other human genes located on different chromosomes as controls. MLPA was performed as described by Schouten et al [14] and the manufacturer. Samples consisted of approximately 100 ng of genomic DNA. The amplified fragments were run on an ABI PRISM 377 DNA Sequencer (Applied Biosystems).
Long-range PCR analysis
When the results of MLPA analysis suggested that a major structural rearrangement was present in the LDL receptor gene, the results were confirmed with long-range PCR analysis [10] using the Expand 20 kbPLUSPCR System (Roche). The sizes of deletions were estimated from the difference in size of the amplified allele with the deletion and the normal allele with long range PCR.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
DD designed the study, interpreted the data and drafted the manuscript. PHN conceived of the study and participated in the design of the study, the interpretation of data and helped to draft the manuscript. LGJ participated in the design of the study and helped to draft the manuscript. GGN and AS carried out the molecular genetic studies. MLL and OF evaluated the patients clinically, and OF 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
The work was supported by the Danish Heart Foundation (grants: 00-1-3-42-22804, 00-1-3-42-22804B, 01-1-3-13A-22888 and 03-1-3-39-22062). We are grateful for the expert technical assistance of Mrs Kirsten Kruse Olsen and Mrs Kirsten Hald from the Department of Clinical Biochemistry, Aarhus Sygehus.
Figures and Tables
Figure 1 Relative copy number of LDL receptor gene exon 5. Boxplots of relative copy number of LDL receptor gene exon 5 measured with Real-Time PCR Analysis and MLPA analysis showing median; box: 25th -75th percentile; bars: largest and smallest values within 1.5 box lengths; circles: outliers.
Figure 2 Mean relative copy numbers of exons in the LDL receptor gene. Mean relative allele copy number ± SD of LDL receptor gene exons 1–9, exons 11–12, exons 14–18, 5' (C3) and 3' (C12) flanking control fragments in 14 patients with a 9 kb deletion of exons 3–6 measured with MLPA analysis compared to mean relative copy number in 12 normal controls.
Figure 3 Graphical results of MLPA analysis. Peak profiles of the MLPA analysis in one normal individual and in the five patients with the deletions described in the results section.
Table 1 Relative copy number of LDL receptor gene exon 5 measured with Real-Time PCR and MLPA
Method Sample population N Mean SD CV (%)
Real-Time PCR Normal 12 0.84 0.17 20
Deletion 17 0.44 0.097 22
MLPA Normal 12 1.03 0.02 2
Deletion 17 0.54 0.05 10
Mean relative allele copy number in 12 normal individuals and in 17 individuals with deletion of exon 5. SD: standard deviation. CV (%): coefficient of variation in percent.
Table 2 Cost-analysis for detection of major structural rearrangements in the LDL receptor gene with quantitative PCR methods
Real-Time PCR MLPA
Price reagents (in euros) 1120 110
Time laboratory technician (hours) 56 14
Data are given for a batch of 10 samples tested in 16 exons in the LDL receptor gene with the Real-Time PCR method and with the MPLA method. Costs of data analysis and interpretation of results were the same for the two methods.
==== Refs
Goldstein JL Hobbs HH Brown MS Scriver CR, Beaudet A, Sly WS, Vale D Familial Hypercholesterolemia The metabolic and molecular basis of inherited disease 2001 New York: McGraw Hill
Heath KE Gahan M Whittall RA Humphries SE Low-density lipoprotein receptor gene (LDLR) world-wide website in familial hypercholesterolaemia: update, new features and mutation analysis Atherosclerosis 2001 154 243 246 11137106 10.1016/S0021-9150(00)00647-X
Villeger L Abifadel M Allard D Rabes JP Thiart R Kotze MJ Beroud C Junien C Boileau C Varret M The UMD-LDLR database: additions to the software and 490 new entries to the database Hum Mutat 2002 20 81 87 12124988 10.1002/humu.10102
Sun XM Webb JC Gudnason V Humphries S Seed M Thompson GR Knight BL Soutar AK Characterization of deletions in the LDL receptor gene in patients with familial hypercholesterolemia in the United Kingdom Arterioscler Thromb 1992 12 762 770 1319734
Fouchier SW Defesche JC Umans-Eckenhausen MW Kastelein JP The molecular basis of familial hypercholesterolemia in The Netherlands Hum Genet 2001 109 602 615 11810272 10.1007/s00439-001-0628-8
Rudiger NS Heinsvig EM Hansen FA Faergeman O Bolund L Gregersen N DNA deletions in the low density lipoprotein (LDL) receptor gene in Danish families with familial hypercholesterolemia Clin Genet 1991 39 451 462 1863993
Nissen H Hansen AB Guldberg P Hansen TS Petersen NE Horder M Evaluation of a clinically applicable mutation screening technique for genetic diagnosis of familial hypercholesterolemia and familial defective apolipoprotein B Clin Genet 1998 53 433 439 9712531
Jensen HK Jensen LG Meinertz H Hansen PS Gregersen N Faergeman O Spectrum of LDL receptor gene mutations in Denmark: implications for molecular diagnostic strategy in heterozygous familial hypercholesterolemia Atherosclerosis 1999 146 337 344 10532689 10.1016/S0021-9150(99)00158-6
Rodningen OK Leren TP Application of long polymerase chain reaction in the study of the LDL receptor gene Scand J Clin Lab Invest 1996 56 93 96 8850178
Kim SH Bae JH Chae JJ Kim UK Choe SJ Namkoong Y Kim HS Park YB Lee CC Long-distance PCR-based screening for large rearrangements of the LDL receptor gene in Korean patients with familial hypercholesterolemia Clin Chem 1999 45 1424 1430 10471644
Wilhelm J Pingoud A Real-time polymerase chain reaction Chembiochem 2003 4 1120 1128 14613102 10.1002/cbic.200300662
Wittwer CT Herrmann MG Moss AA Rasmussen RP Continuous fluorescence monitoring of rapid cycle DNA amplification Biotechniques 1997 22 130 138 8994660
Freeman WM Walker SJ Vrana KE Quantitative RT-PCR: pitfalls and potential Biotechniques 1999 26 112 115 9894600
Schouten JP McElgunn CJ Waaijer R Zwijnenburg D Diepvens F Pals G Relative quantification of 40 nucleic acid sequences by multiplex ligation-dependent probe amplification Nucleic Acids Res 2002 30 e57 12060695 10.1093/nar/gnf056
Gille JJ Hogervorst FB Pals G Wijnen JT van Schooten RJ Dommering CJ Meijer GA Craanen ME Nederlof PM de Jong D McElgunn CJ Schouten JP Menko FH Genomic deletions of MSH2 and MLH1 in colorectal cancer families detected by a novel mutation detection approach Br J Cancer 2002 87 892 897 12373605 10.1038/sj.bjc.6600565
Damgaard D Larsen ML Nissen PH Jensen JM Jensen HK Soerensen VR Jensen LG Faergeman O The relationship of molecular genetic to clinical diagnosis of Familial Hypercholesterolemia in a Danish population Atherosclerosis 2005 180 155 160 15823288 10.1016/j.atherosclerosis.2004.12.001
Hobbs HH Brown MS Goldstein JL Molecular genetics of the LDL receptor gene in familial hypercholesterolemia Hum Mutat 1992 1 445 466 1301956
De Preter K Speleman F Combaret V Lunec J Laureys G Eussen BH Francotte N Board J Pearson AD De Paepe A Van Roy N Vandesompele J Quantification of MYCN, DDX1, and NAG gene copy number in neuroblastoma using a real-time quantitative PCR assay Mod Pathol 2002 15 159 166 11850545 10.1038/modpathol.3880508
Joncourt F Neuhaus B Jostarndt-Foegen K Kleinle S Steiner B Gallati S Rapid identification of female carriers of DMD/BMD by quantitative real-time PCR Hum Mutat 2004 23 385 391 15024733 10.1002/humu.20007
Laurendeau I Bahuau M Vodovar N Larramendy C Olivi M Bieche I Vidaud M Vidaud D TaqMan PCR-based gene dosage assay for predictive testing in individuals from a cancer family with INK4 locus haploinsufficiency Clin Chem 1999 45 982 986 10388473
Wilke K Duman B Horst J Diagnosis of haploidy and triploidy based on measurement of gene copy number by real-time PCR Hum Mutat 2000 16 431 436 11058901 10.1002/1098-1004(200011)16:5<431::AID-HUMU8>3.0.CO;2-Z
Wang J Ban MR Hegele RA Multiplex ligation-dependent probe amplification of LDLR enhances molecular diagnosis of familial hypercholesterolemia J Lipid Res 2005 46 366 372 15576851 10.1194/jlr.D400030-JLR200
| 15842735 | PMC1087844 | CC BY | 2021-01-04 16:03:33 | no | BMC Med Genet. 2005 Apr 20; 6:15 | utf-8 | BMC Med Genet | 2,005 | 10.1186/1471-2350-6-15 | oa_comm |
==== Front
BMC MicrobiolBMC Microbiology1471-2180BioMed Central London 1471-2180-5-181584768810.1186/1471-2180-5-18Research ArticleCrosstalk regulation among group 2- Sigma factors in Synechocystis PCC6803 Lemeille Sylvain [email protected] Johannes [email protected] Amel [email protected] Laboratoire Adaptation et Pathogénie des Microorganismes, Université Joseph Fourier Bâtiment Jean Roget – Faculté Médecine-Pharmacie. Domaine de la Merci. 38700 La Tronche, France2 Laboratoire de chimie bactérienne, IBSM-CNRS, 31 chemin Joseph Aiguier. 13402 Marseille cedex 20, France2005 22 4 2005 5 18 18 22 12 2004 22 4 2005 Copyright © 2005 Lemeille 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 cyanobacterium Synechocystis PCC6803 contains one group 1 (sigA) and four group 2 (sigB, sigC, sigD and sigE) sigma factors. The activity of these multiple sigma factors determines the transcriptional program of this bacterium. We wanted to study the role of the group 2 sigma factors in Synechocystis. We have therefore constructed mutants of each of the group 2 sigma factors and investigated their crosstalk.
Results
We used quantitative RT-PCR analysis to measure the relative abundance of the sig mRNAs in the four sigma mutants. Our data indicate that a network of mutual transcriptional regulation links the expression of the sigma genes. Accordingly, an environmental stress acting on only one of the sigma factors will indirectly modify the expression of most of the other sigma factors. This was confirmed by the transcriptional analysis of the sig mRNAs as a function of nitrogen starvation.
Conclusion
Taken together, our observations suggest that the crosstalk regulation between all group 1 and group 2 genes could be important for the adaptation of the bacterium to different environmental and physiological conditions.
==== Body
Background
Bacterial sigma subunits of RNA polymerase are global regulators of gene expression, conferring specificity to the recognition of promoters by the core enzyme. Two broad families of sigma factors have been identified: the σ70 type, and the σ54 type factors [1]. The σ54 family regulates a variety of genes such as those involved in chemotaxis, synthesis of structural components of flagella and enzymes involved in the response to nitrogen starvation [2]. The σ70 family is subdivided into three groups [1]. Group 1 comprises the primary sigma factors that control the transcription of housekeeping genes, and these sigma factors are therefore essential for cell viability. Groups 2 and 3 include the so-called alternative sigma factors that coordinate the regulation of gene expression in bacteria on a global level. They direct the transcription of a specific genetic program that allows bacteria to cope with particular environmental changes and stress conditions. Group 2 sigma factors are similar in sequence to primary sigma factors and include proteins such as the stationary-phase-specific sigma factor, RpoS [3]. Group 3 sigma factors show less sequence similarity with those of group 1 and include proteins required for the heat shock response [4] and motility [5]. The inactivation of a gene encoding a particular groups 2 or group 3 sigma factor usually produces growth defects or other phenotypes under specific physiological or environmental conditions. An E. coli rpoS mutant, e.g., has a pleitropic phenotype: it shows a loss of viability in stationary phase and a decreased resistance to some stresses such as the osmotic stress [6]. In Synechocystis PCC6803, inactivation of the sigF gene, encoding a group 3 sigma factor, leads to the loss of motility and pilus formation [7]. In Synechococcus four mutants of rpoD genes show defects in the circadian expression of the psbAI gene, encoding the protein D1 of the photosystem II reaction center [8].
The unicellular cyanobacterium Synechocystis sp. strain PCC6803 possesses one group 1 sigma factor, sigA (slr0653), four group 3 sigma factors (sll 0687, sll0856, slr1545, slr1564) and four group 2 sigma factors, sigB to sigE (sll0306, sll0184, sll2012, sll1689) [9]. SigE is involved in the response to nitrogen stress [10] and a contribution of the SigB/SigD factors to the dark/light adaptation has been reported recently [11]. The synthesis of the other alternative sigma factors is also modulated in response to particular stresses [12-14].
In order to assess the role of alternative sigma factors in Synechocystis, we have chosen to study the group 2 sigma factors. We have analyzed the transcription of all members of this family of sig genes as well as the transcription of the group 1 sigma factor in Synechocystis PCC6803 wild type strain and mutants lacking the group 2 sigma genes. Based on our results we suggest that these sigma factors are linked by a network of mutual regulation that could allow them to act in concert in the global transcriptional control of this bacterium.
Results and discussion
Construction and growth of sigma mutants
We have constructed null mutant strains lacking either sigB, sigC, sigD, or sigE gene as described in Materials and Methods. These four mutants segregated completely (data not shown). We were not able to inactivate the sigA gene, suggesting that it is necessary for the viability of the organism. The growth of the sigB, sigC, sigD, or sigE mutants compared to the wild type strain was examined under normal growth conditions. The results of Figure 1 indicate that all four mutants have a similar growth rate as the wild type strain during the exponential phase, showing that these sigma factors are dispensable for growth under our culture conditions and confirming their classification as group 2 sigma factors.
Expression of the group 2 sigma genes during normal growth
Recent studies have shown that all four group 2 sigma genes are expressed in normally growing cells [12]. We wanted to assess if their transcription changes in response to a gradual physiological modification of the internal and external environment of the bacterium. Our analysis is based on quantitative RT-PCR. We quantified the amount of the sigB, sigC, sigD and sigE transcripts during three stages of growth: mid log, early stationary phase and late stationary phase. The level of the transcripts was normalized to the level of rpoA transcription as described in Material and Methods. The results presented in Figure 2 show that all four group 2 sigma factors are transcribed in all growth phases. However, the relative abundance of the different sigma factors changes all along the growth of the culture. The sigB transcript was maximum during exponential phase, and decreased as cells grew into stationary phase. sigC and sigD transcription decreased upon entry into stationary phase and increased as cells were adapted to this stage of growth, sigE had the weakest variation. Imamura et al. [12] have shown that the protein concentrations of three (SigC, SigD, SigE) sigma factors vary in a similar manner as in our transcriptional analysis, suggesting that, at least for these factors, RNA levels correlate well with protein amounts.
These expression profiles show that the transcript levels of at least three of these sigma factors depend on the cell density (A750 value) of the culture and suggest that they could be important for the global physiology of this bacterium at all stages of growth.
All sigma factors are expressed in many environmental conditions tested, such as iron and sulfur starvations, and heat and osmotic shocks (data not shown). Similar results were previously obtained in Synechococcus elongatus PCC7942 [15] where all sigma genes were found to be active under many growth conditions. Recently Imamura et al. [12] measured the concentration of all five sigma factors during normal growth of Synechocystis PCC6803. They found the amounts of these proteins to vary between 1 and 10 fmoles/μg of total protein. These data suggest that all five factors are important for the cellular physiology of Synechocystis PCC6803 under standard conditions. This conclusion is somewhat surprising since four of the five sig genes can be mutated, indicating that neither of them is essential for viability under these conditions. One possible way to reconcile these two divergent observations would be to suppose that the function of the different sigma factors is redundant. According to this model, most genes would be transcribed by more than one sigma factor. Indeed, the sigma genes themselves are transcribed from multiple promoters [12]. Furthermore, it is even possible that the different sigma factors recognize the same promoter. This hypothesis is supported by data obtained from in vitro transcription experiments in which all three different sigma factors (RpoD1, RpoD3, RpoD4) could initiate the transcription of the rrnA, cpcB1A1 and P1a promoters of Synechoccocus sp. strain PCC7942 [15]. This specificity crosstalk among sigma factors is also revealed by an in vivo analysis of psbAI promoter activities in S. elongatus where the principal sigma factor, as well as each group 2 sigma factor, all recognize the psbA1 promoter of this bacterium [16].
Transcription of sig genes in σ mutants
Since the sig genes are transcribed by RNA polymerase holoenzyme, they necessarily regulate each other's transcription. In other systems it is well documented that the transcription of alternative sigma factor genes is controlled by other σ factors [17-19]. In order to investigate this regulatory network, we measured the transcription of each of the five sig genes in all four σ mutants during exponential growth. Our method can detect transcripts of all sig genes in all of the mutants because the cDNA synthetized during the RT-PCR used primers that anneal upstream of the inactivating chloramphenicol cassette. The results shown in Figure 3 confirm the existence of complex regulatory connections between the different sigma factors and suggest a highly interconnected network: (i) mutation in the sigB gene leads to a 6-fold decrease of sigA, sigC and sigE genes (Figure 3), (ii) in the sigD mutant, transcription of the sigA, sigB, sigC and sigE genes decreased about 3 to 4-fold (Figure 3), (iii) sigE mutation leads to a strong decrease (about 20-fold) of the transcription of the sigA and sigB genes and to a 5-to 3-fold decrease of the expression of the sigC and sigD genes (Figure 3), (iii) mutation in the sigC gene does not negatively affect the transcription of any of the 4 sigma genes tested (Figure 3).
SigE seems to be a particularly important sigma factor because it controls the expression of three other sig genes. Mutation of the sigE gene had the strongest effects among all mutants inactivating sigma genes, affecting particularly the housekeeping gene sigA and sigB. The role of the housekeeping sigma factor, SigA, remains less well defined. Since inactivation of this gene is lethal, we will have to investigate its role using conditional mutants or biochemical methods.
By quantifying the sigma transcripts in different sigma mutants we have shown that the transcription of the sig genes is controlled by a network of mutual connections between the sigmas. Previous studies in related organisms had also shown a mutual transcriptional regulation of sigma factors. In Synechococcus PCC7942, the rpoD1 gene is transcribed by RpoD3 and RpoD4 factors [15] and SigC factor has a negative effect on SigB expression [11]. In Borrelia burgdorferi, for example, RpoN regulates the expression of rpoS [18].
Sigma factors transcription under nitrogen starvation
According to the network of mutual transcriptional regulation of the sigma factors, we speculated that an environmental stress acting on only one of the sigma factors will indirectly modify the expression of most of the other sigma factors. We tested this hypothesis by analyzing the transcription profiles of the group 1 sigma gene and the four group 2 sigma factors under nitrogen starvation. We measured the mRNA concentration of the group 1 sigma gene and the four group 2 genes by quantitative RT-PCR under nitrogen stravation. As shown in Figure 4a, sigE expression was induced, as expected, about 5-fold with respect to the reference gene rpoA. The transcription of the other four sigma genes was also induced (about 2- to 6-fold). Nitrogen starvation does not only lead to the overexpression of the sigE gene but rather provoked a readjustment of the relative abundance of the sigma factors. This global change of the expression of all sigma genes in response to this particular environmental condition agrees with our hypothesis that the cross-talk regulation among the sigmas could lead to the transmission of one particular signal to all of them. The transcription of a structural gene involved in this stress should be affected by more than one sigma factor. glnN, encoding the GSIII glutamine synthase, known to be highly expressed under conditions of nitrogen deficiency [10,17] was chosen as a target gene for the nitrogen starvation. We have analyzed its transcription in the wild-type and mutant strains after one week of nitrogen starvation. Induction of glnN transcription was observed in the wild type strain. This induction was abolished when SigB, SigC or SigE factors were inactivated (Figure 4b). These results clearly demonstrated that more than one sigma factor affects the transcription of the glnN gene.
Our analysis is based on measuring the first level of control of sig genes expression. In other systems, post-transcriptional modifications of sigma factors can occur and may not correlate with transcriptional profiles. The recent data of Imamura et al [12] suggest however that the intracellular concentration of the sigma factors correlates well with the transcriptional control of these sigma factors (with the possible exception of SigB). The same authors demonstrated that a negative effect of SigC on sigB transcription correlates with the reduction of SigB protein levels in this mutant [12]. These data suggest that transcriptional regulation of sig genes is not drastically modified by post-transcriptional control. This hypothesis is actually under investigation in our laboratory.
Conclusion
Synechocystis PCC6803 possesses multiple sigma factors and its transcriptional program is largely determined by the activity of these multiple sigma factors. Our study has explored the relationship among all members of the group 2 sigma factors and their connection with the housekeeping sigma factor. We have shown that the transcription of the sig genes is controlled by a network of mutual connections. The strongest effects are compiled in Figure 5, where the thickness of the arrows is proportional to the effect of a given mutation on the expression of the sigma genes. For example, mutating the sigE gene results in the reduced expression of four sigma genes, with a more pronounced effect on the sigA and sigB genes. A mutation in the sigB gene results in a small reduction of the transcription of the sigA, sigE and sigC genes. Expression of the group 1 sigma gene (sigA) is affected by mutations of three group 2 sigma genes. Our study has explored the relationship among all members of one family of sigma factors in eubacteria. We assume that their mutual connections are part of a more extended regulation network. In fact, it is tempting to speculate that all sigma factors in a cell control each other. A possible connection to group 3 sigma factors, as well as the relationships among sigma factors and sensor and regulatory proteins in the cell remain to be elucidated.
Methods
Culture and growth conditions
Synechocystis sp. strain PCC6803 was obtained from the Pasteur culture collection. Wild-type and mutant strains were grown at 30°C with continuous illumination at approximately 20 μE m-2 s-1, with 3% CO2 in air, in BG11 medium [20], buffered with 5 mM Hepes-KOH, pH8. Growth was monitored by measuring the optical density at 750 nm (A750).
For nitrogen starvation, BG11 medium lacking the nitrogen source (NaNO3) was buffered with 20 mM N-tris(hydroxymethyl)methyl-2-aminoethanosulfonic acid (TES) buffer, pH 7.5. Strains used in this condition were grown to an A750 = 1, transferred to the nitrogen-depleted medium and incubated for one week. All cyanobacterial strains were grown on BG11 plates containing 1.5% Difco Bacto Agar. When needed, chloramphenicol was added to a concentration of 10 μg/ml. Growth rates of mutants were compared to a Synechocystis strain carrying the same antibiotic resistance cassette inserted into an inessential gene, ureA. We call this strain the wild-type for our experiments.
DNA manipulation and RNA isolation
Molecular techniques were performed according to standard procedures [21]. Synechocystis genomic DNA was prepared according to the method of Tandeau de Marsac et al. [22]. RNA was extracted from pelleted cells, broken by freezing in liquid nitrogen, and using the RNeasy kit (Qiagen) according to the manufacturer's specifications. Chromosomal DNA was removed by treating RNA preparations with 1 μl of DNAse (at 2U/μl) (Ambion) for 1 hour at 37°C. The concentration of RNA was determined spectrophotometrically.
Gene inactivation
The sigB (sll0306), sigC (sll0184), sigD (sll2012) and sigE (sll1689) genes [9] were cloned using the TOPO-TA cloning kit (Invitrogene) and the following primers: sigB-1 and sigB-2, sigC-1 and sigC-2, sigD-1 and sigD-2, sigE-1 and sigE-2.
These genes were then subcloned into pBluescript SK- plasmid (Stratagene) between the ApaI and SpeI sites. A chloramphenicol cassette was inserted at the unique site BglII in sigC, SmaI in sigB, and BamHI both in sigE and sigD. The cassette was obtained from the pACYC184 plasmid [23] by PCR amplification. The primers used were cat-1 and cat-2, they add BamHI and SmaI restriction sites at each end of the amplified sequence.
The SK- vector derivatives containing each interrupted gene were used to transform Synechocystis. Chloramphenicol-resistant transformants that were also ampicillin sensitive were selected and subsequently screened for replacement of the wild-type gene allele with the corresponding mutant. Genomic DNA isolated from individual CmR transformants was verified by PCR.
Reverse transcription
For each reaction, 1 μl of antisens primer mix at 2.5 μM of each of the primers and 200 ng of total RNA were denaturated at 95°C and chilled quickly on ice. A mix consisting of 4 μl of 5x buffer (250 mM Tris-HCl pH 8.3, 375 mM KCl, 15 mM MgCl2, 50 mM DTT), 0.5 μl of RNase Inhibitor (40U/μl), 1 μl of 5 mM dNTP and 1 μl of MMLV reverse transcriptase enzyme (200U/μl) was added in a total volume of 20 μl, followed by one hour of incubation at 37°C. The composition of the primer mix used in each reverse transcription reaction was dependent on the genes analyzed in the quantitative PCR experiments (for example when rpoA and glnN genes were quantified, the reverse transcription reaction used rpoA-RT and glnN-RT primers). The primers used were: rpoA-RT; sigB, C-RT; sigA, D, E-RT; glnN-RT.
The sequences of all the primers used in this study are listed in table 1.
Real-time quantitative PCR
PCR conditions were identical for all reactions. The 25 μl-reaction mixture consisted of 1x master mix buffer (Eurogentec), 0.75 μl of SYBR Green I Dye (Eurogentec), and 1 μl of each primer (2 μM). 5 μl aliquots of the diluted reverse transcription reaction were used as template. PCR amplifications (2' at 50°- 10' at 95° – 40 X [15 '' at 95° – 1' at 60°]) were carried out in a Gen Amp 5700 sequence detection system (Applied Biosystems). The forward (F) and reverse (R) primers used in these PCR reactions were designed using the Primer Express software (Applied Biosystems). For each RT-PCR reaction, the efficiency of the DNAse treatment was verified by an identical parallel PCR reaction, but omitting reverse transcription. Only DNA-free RNAs were used in our experiments. All PCR primers anneal to their target at temperatures comprised between 58 and 60°C and amplify 200-pb fragments internal to the coding sequence of the relevant gene. The primers used in the quantitative-PCR experiments are listed in table 1.
Quantitative analysis of the sample
Real-time PCR quantitatively detects the concentration of double-stranded PCR products by monitoring the fluorescence of SYBR Green I Dye, which selectively binds to double-stranded DNA. All measurements were carried out in triplicates. The analysis method was based on the threshold cycle value (CT) that indicates the fractional cycle number at which the amount of amplified target reaches a fixed threshold. The standard deviations of our measurements are indicated by the error bars in the figures. Since the amount of rpoA transcript was practically constant in all the conditions tested in this study (table 2) we report the quantities of the other cDNAs with respect to the concentration of the rpoA transcript. We calculate the difference (ΔCT) between the mean CT of this sample and the mean CT value of the rpoA mRNA for that sample. The relative concentration of the mRNA species is then calculated as 2C-ΔCT.
Authors' contributions
S L carried out all of the RT-PCR experiments and assembled the figures.
J G participated in the analysis of the data and the writing of the manuscript.
A L conceived the study, constructed the mutants and helped in writing the manuscript.
Figures and Tables
Figure 1 Growth of the sig mutants (solid circle) compared to the wild type strain (open circle) under normal conditions. a: sigB mutant, b: sigC mutant, c: sigD mutant, d: sigE mutant. Wild type and mutant strains were grown in Bg11 medium. Growth was monitored by measuring the optical density at 750 nm (A750).
Figure 2 Trascription of sig genes during normal growth. Measurement of relative concentrations of sigB, sigC, sigD and sigE mRNAs in Synechocystis by quantitative RT-PCR during three stages of normal growth: mid log phase in black (A750 = 1); early stationary phase in white (A750 = 8); and late stationary phase in grey (A750 = 12.6). Each sample was measured in triplicate and the standard deviation is indicated by error bars. Values were normalized to the rpoA transcript.
Figure 3 Transcription of sig genes in sigma mutants. sigABCDE genes were quantified in the wild type and the four sigma mutants (sigB, sigC, sigD and sigE). The gene quantified is indicated below each set of measurements. Within each set, the phenotype is indicated below each bar of the graph. All measurements were carried out under identical culture conditions for the wt and mutants strains. Values were normalized to the rpoA transcript. The expression of each gene in the wt strain was set equal to 10 (in arbitrary units). The reported values represent the average of 6 measurements obtained from two separate RNA preparations.
Figure 4 a: Levels of sigma factor mRNAs during nitrogen starvation. Cells were grown under normal (white columns) or nitrogen starved (grey columns) conditions. sig mRNAs were quantified using quantitative RT-PCR. Each sample was measured in triplicate and the standard deviation is indicated by error bars. Values were normalized to the rpoA transcript. The expression of each gene in non starved cells was set equal to 10 (in arbitrary units). b: Effect of sigma mutations on the transcription of glnN gene during nitrogen starvation. mRNAs were quantified using quantitative RT-PCR. Each sample was measured in triplicate and the standard deviation is indicated by error bars. Values were normalized to the rpoA transcript. The expression of glnN gene in non starved cells (white columns) was set equal to 10 (in arbitrary units). The grey bars represented the expression of glnN gene in nitrogen starved cells.
Figure 5 Schematic network of transcriptional interactions between group 2 sigma genes transcription in Synechocystis: The thickness of the arrows is proportional to the effect of a given mutation on the transcription of the sigma gene to which the arrow points.
Table 1 List of primers used in this study. All sequences are written from 5' to 3'.
Primer Sequence
sigB-1 CGGAATTCTTGGGTATCTTTTTAGC
sigB-2 CCGAAGCTTGGGCAACTAACTGGC
sigC-1 CGGAATTCAAAGCCTGCCATCGGCC
sigC-2 CCGAAGCTTGTGGCCTAACCCAAATTTTC
sigD-1 CGGAATTCGAGTATGTGCTTCACAA
sigD-2 CCGAAGCTTCTTTTCTTTAGCTAGCT
sigE-1 CGGAATTCTTTGGAAAATCAATGACA
sigE-2 CCGAAGCTTGGTATCTATAACCAACC
cat-1 CGGGATCCCGCCCGGGAATTACGCCCCGCCC
cat-2 CGGGATCCCGCCCGGGCAGGAGCTAAGGAAGCTA
rpoA-RT TAACCTA
sigB, C-RT TAACCTT
sigA, D, E-RT TAACCCT
glnN-RT CCATCCGTC
sigA forward TGGAGTTGGAAACCGCC
sigA reverse GACTGCACCATTTTGTCTTTGG
sigB forward AGAAATGGCCCGCTATCCC
sigB reverse GCCCGCAGTTGATAAAGCC
sigC forward TGGAGTTGGAAACCGCC
sigC reverse GACTGCACCATTTTGTCTTTGG
sigD forward GATGGCCCTGCTGGAGC
sigD reverse TTGCGCTTCTGATATTTCTTGG
sigE forward CGGGCCGCAGAATCC
sigE reverse CCAACTCCGCCCATCG
rpoA forward GAGTTCGCCACTATTCTAGGCG
rpoA reverse TTAGGATCAATAACCTCCACCTC
glnN forward GATTTAACCAAGGACGCTGGC
glnN reverse CAAAACGGTTACCAGTGAAGGC
Table 2 rpoA Ct values for the different growth conditions tested in this study. The reported values represent the average of 6 measurements obtained from two separate RNA preparations.
Culture condition or genetic background rpoA Ct value per 20 ng total RNA
Normal 28.5 ± 1.03
Nitrogen starvation 29 ± 0.27
Exponential phase 29.44 ± 0.03
Early stationary phase 28.60 ± 0.52
Late stationary phase 28.15 ± 0.04
Wild type strain 30 ± 0.91
sigB mutant 30.29 ± 0.25
sigC mutant 29.66 ± 1.06
sigD mutant 29.76 ± 0.8
sigE mutant 29.85 ± 0.69
==== Refs
Lonetto M Gribskov M Gross CA The sigma 70 family: sequence conservation and evolutionary relationships J Bacteriol 1992 174 3843 3849 1597408
Kustu S Santero E Keener J Popham D Weiss D Expression of sigma 54 (ntrA)-dependent genes is probably united by a common mechanism Microbiol Rev 1989 53 367 376 2677638
Mulvey MR Loewen PC Nucleotide sequence of katF of Escherichia coli suggests KatF protein is a novel sigma transcription factor Nucleic Acids Res 1989 17 9979 9991 2690013
Hecker M Schumann W Volker U Heat-shock and general stress response in Bacillus subtilis Mol Microbiol 1996 19 417 428 8830234 10.1046/j.1365-2958.1996.396932.x
Arora SK Ritchings BW Almira EC Lory S Ramphal R A transcriptional activator, FleQ, regulates mucin adhesion and flagellar gene expression in Pseudomonas aeruginosa in a cascade manner J Bacteriol 1997 179 5574 5581 9287015
Hengge-Aronis R Signal transduction and regulatory mechanisms involved in control of the sigma(S) (RpoS) subunit of RNA polymerase Microbiol Mol Biol Rev 2002 66 373 95 table 12208995 10.1128/MMBR.66.3.373-395.2002
Bhaya D Watanabe N Ogawa T Grossman AR The role of an alternative sigma factor in motility and pilus formation in the cyanobacterium Synechocystis sp. strain PCC6803 Proc Natl Acad Sci USA 1999 96 3188 3193 10077659 10.1073/pnas.96.6.3188
Nair U Ditty JL Min H Golden SS Roles for sigma factors in global circadian regulation of the cyanobacterial genome J Bacteriol 2002 184 3530 3538 12057947 10.1128/JB.184.13.3530-3538.2002
Kaneko T Sato S Kotani H Tanaka A Asamizu E Nakamura Y Sequence analysis of the genome of the unicellular cyanobacterium Synechocystis sp. strain PCC6803. II. Sequence determination of the entire genome and assignment of potential protein-coding regions DNA Res 1996 3 109 136 8905231
Muro-Pastor AM Herrero A Flores E Nitrogen-regulated group 2 sigma factor from Synechocystis sp. strain PCC 6803 involved in survival under nitrogen stress J Bacteriol 2001 183 1090 1095 11208809 10.1128/JB.183.3.1090-1095.2001
Imamura S Asayama M Takahashi H Tanaka K Takahashi H Shirai M Antagonistic dark/light-induced SigB/SigD, group 2 sigma factors, expression through redox potential and their roles in cyanobacteria FEBS Lett 2003 554 357 362 14623094 10.1016/S0014-5793(03)01188-8
Imamura S Yoshihara S Nakano S Shiozaki N Yamada A Tanaka K Purification, characterization, and gene expression of all sigma factors of RNA polymerase in a cyanobacterium J Mol Biol 2003 325 857 872 12527296 10.1016/S0022-2836(02)01242-1
Tuominen I Tyystjarvi E Tyystjarvi T Expression of Primary Sigma Factor (PSF) and PSF-Like Sigma Factors in the Cyanobacterium Synechocystis sp. Strain PCC 6803 J Bacteriol 2003 185 1116 1119 12533490 10.1128/JB.185.3.1116-1119.2003
Nakamura Y Kaneko T Sato S Ikeuchi M Katoh H Sasamoto S Complete genome structure of the thermophilic cyanobacterium Thermosynechococcus elongatus BP-1 DNA Res 2002 9 123 130 12240834
Goto-Seki A Shirokane M Masuda S Tanaka K Takahashi H Specificity crosstalk among group 1 and group 2 sigma factors in the cyanobacterium Synechococcus sp. PCC7942: In vitro specificity and a phylogenetic analysis Mol Microbiol 1999 34 473 484 10564489 10.1046/j.1365-2958.1999.01608.x
Nair U Thomas C Golden SS Functional elements of the strong psbAI promoter of Synechococcus elongatus PCC 7942 J Bacteriol 2001 183 1740 1747 11160106 10.1128/JB.183.5.1740-1747.2001
Caslake LF Gruber TM Bryant DA Expression of two alternative sigma factors of Synechococcus sp. strain PCC 7002 is modulated by carbon and nitrogen stress Microbiology 1997 143 3807 3818 9421905
Hübner A Yang X Nolen DM Popova TG Cabello FC Norgard MV Expression of Borrelia burgdorferi OspC and DbpA is controlled by a RpoN-RpoS regulatory pathway Proc Natl Acad Sci USA 2001 98 12724 12729 11675503 10.1073/pnas.231442498
Kulkarni RD Golden SS Adaptation to high light intensity in Synechococcus sp. strain PCC 7942: regulation of three psbA genes and two forms of the D1 protein J Bacteriol 1994 176 959 965 8106338
Rippka R Deruelles J Waterbury JB Herman M Stanier RY Genetics assignments, strain histories and properties of pure cultures of cyanobacteria J Gen Microbiol 1979 111 1 61
Sambrook J Fritsch EF Maniatis T Molecular Cloning: a laboratory manuel 1989 Cold Spring Harbor Laboratory Press
Tandeau de Marsac N Borrias WE Kuhlemeier CJ Castets AM van Arkel GA van den Hondel CA A new approach for molecular cloning in cyanobacteria: cloning of an Anacystis nidulans met gene using a Tn901-induced mutant Gene 1982 20 111 119 6298064 10.1016/0378-1119(82)90092-0
Chang AC Cohen SN Construction and characterization of amplifiable multicopy DNA cloning vehicles derived from the P15A cryptic miniplasmid J Bacteriol 1978 134 1141 1156 149110
| 15847688 | PMC1087845 | CC BY | 2021-01-04 16:03:39 | no | BMC Microbiol. 2005 Apr 22; 5:18 | utf-8 | BMC Microbiol | 2,005 | 10.1186/1471-2180-5-18 | oa_comm |
==== Front
BMC Med Inform Decis MakBMC Medical Informatics and Decision Making1472-6947BioMed Central London 1472-6947-5-101584514710.1186/1472-6947-5-10Research ArticleDevelopment and preliminary evaluation of the VPS ReplaySuite: a virtual double-headed microscope for pathology Johnston Dan J [email protected] Sean P [email protected] Peter A [email protected]'Shea Daniel G [email protected] Medical Informatics Group, School of Biotechnology, Dublin City University, Dublin, Ireland2 The Conway Institute of Biomolecular and Biomedical Research, University College Dublin and The Pathology Department, Mater Misericordiae Hospital, Dublin, Ireland2005 21 4 2005 5 10 10 14 12 2004 21 4 2005 Copyright © 2005 Johnston et al; licensee BioMed Central Ltd.2005Johnston 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
Advances in computing and telecommunications have resulted in the availability of a range of online tools for use in pathology training and quality assurance. The majority focus on either enabling pathologists to examine and diagnose cases, or providing image archives that serve as reference material. Limited emphasis has been placed on analysing the diagnostic process used by pathologists to reach a diagnosis and using this as a resource for improving diagnostic performance.
Methods
The ReplaySuite is an online pathology software tool that presents archived virtual slide examinations to pathologists in an accessible video-like format, similar to observing examinations with a double-headed microscope. Delivered through a customised web browser, it utilises PHP (Hypertext PreProcessor) to interact with a remote database and retrieve data describing virtual slide examinations, performed using the Virtual Pathology Slide (VPS).
To demonstrate the technology and conduct a preliminary evaluation of pathologists opinions on its potential application in pathology training and quality assurance, 70 pathologists were invited to use the application to review their own and other pathologists examinations of 10 needle-core breast biopsies and complete an electronic survey. 9 pathologists participated, and all subsequently completed an exit survey.
Results
Of those who replayed an examination by another pathologist, 83.3% (5/6) agreed that replays provided an insight into the examining pathologists diagnosis and 33.3% (2/6) reconsidered their own diagnosis for at least one case. Of those who reconsidered their original diagnosis, all re-classified either concordant with group consensus or original glass slide diagnosis. 77.7% (7/9) of all participants, and all 3 participants who replayed more than 10 examinations stated the ReplaySuite to be of some or great benefit in pathology training and quality assurance.
Conclusion
Participants conclude the ReplaySuite to be of some or of great potential benefit to pathology training and quality assurance and consider the ReplaySuite to be beneficial in evaluating the diagnostic trace of an examination. The ReplaySuite removes temporal and spatial issues that surround the use of double-headed microscopes by allowing examinations to be reviewed at different times and in different locations to the original examination. While the evaluation set was limited and potentially subject to bias, the response of participants was favourable. Further work is planned to determine whether use of the ReplaySuite can result in improved diagnostic ability.
TelepathologyVirtual SlidePathologyEducationQuality AssuranceDouble-Headed Microscope
==== Body
Background
Diagnostic pathology involves the classification of disease, using tissues obtained during biopsies, operations or autopsies. Classification is based on a complex set of visual features identified with the aid of a microscope. To reach a confident diagnosis, pathologists are required to possess well-developed searching, perception and classification skills, acquired with the assistance of intensive one-on-one tutoring sessions with expert pathologists. The workloads of expert pathologists, however, often restrict the frequency of these interactions.
While telepathology has yet to be incorporated into most pathology labs in a clinical role [1-4], advances in computing and telecommunications have resulted in the availability of a range of quality online training tools. Designed to supplement rather than replace human tutoring, such tools enable training pathologists to gain a wider range of educational experiences. Studies have shown the use of online training tools can improve diagnostic performance, beyond what is found with human tutoring alone [5,6].
The majority of training tools comprise interactive tutorials that display a limited number of pre-selected images per case, often with accompanying notation [7-15]. However, as has been observed through the use of static telepathology systems for remote diagnosis, diagnostic accuracy is often dependent on appropriate field selection [16-21]. Novices often make errors when searching the slide [22], so the development of the skills required to locate relevant visual features are as important as those required for identifying them. Neither the searching nor identification skills utilised to reach a diagnosis are assessed by conventional quality assurance studies; diagnostic accuracy is considered the principal indicator of competence.
In contrast to static systems, virtual slides (static-dynamic hybrids) enable unrestricted examination of entire digitised tissue sections [5,23-27]. This is a more accurate representation of the microscopic environment used by pathologists for diagnosis, and when incorporated into training tools, can provide unrestricted but supportive examining environments. With the cost and time required to digitise entire slides decreasing [28], virtual slides are finding greater application in pathology education [5,22,26].
In order to elucidate the diagnostic process of examining pathologists, the VPS (Virtual Pathology Slide), a virtual slide viewer that records diagnostic behaviour, has been developed and validated [23]. The VPS tracks each pathologist's slide examination, recording the location of, and time spent viewing each field, and storing the data on a remote, relational database. This data is available post-examination for interrogation, potentially enabling the diagnostic technique of different pathologists to be analysed and studied. While the VPS tracks user interactions with the software, it does not utilise this data constructively to provide an intrinsic benefit to the end user, the pathologist. For this purpose, the ReplaySuite was developed.
The ReplaySuite is a web-based, user-friendly software tool that enables pathologists to replay virtual slide examinations, performed using the Virtual Pathology Slide (VPS) [23,29,30]. Unlike interactive tutorials and annotated virtual slides, pathologists using the ReplaySuite are able to observe the diagnostic trace of examiners, in a manner similar to the use of a double-headed microscope. This possesses significant potential for use in both pathology training, where trainee pathologists may learn from the diagnostic techniques of experts, and quality assurance, for the detection and elucidation of sources of error in participants diagnostic technique. This paper describes the development and preliminary evaluation of the ReplaySuite technology and discusses its potential applications.
Methods
VPS
The VPS is an interactive microscope emulator that presents a complete digitised tissue section via the Internet or CD ROM, allowing both stepwise increases of magnification (16× to 2000× on standard 17" monitor) and eight-way directional lateral-motion. During VPS examinations (Fig. 1), the locations of examined fields of view and the duration of time they were examined are recorded. While images are being retrieved from the remote server/CD ROM, examination data is transmitted back to the remote VPS server, where it is stored on a relational database management system (Oracle).
Figure 1 The VPS interface used by pathologist's when examining a slide.
Development of the ReplaySuite
The ReplaySuite is a web application that dynamically generates HTML (HyperText Markup Language) web pages to display images and text, using PHP, a server-side scripting language. It communicates with an Oracle database, on which all user tracking data is stored, via OCI (Oracle Call Interface) to retrieve and write data. JavaScript is used to deliver dynamic client side functionality within the graphical user interface (GUI). The ReplaySuite is deployed in a Microsoft Foundation Class (MFC) application written in Visual C++, which acts as a customised web browser. A complete description of the ReplaySuite specification is available in the ReplaySuite technical document [32].
Data describing an examination is retrieved from the remote server, via server side scripting, enabling communication between the application and the database. Structured Query Language (SQL) queries are embedded within PHP, interrogating the database and retrieving datasets. This data is used to determine which images to request from the server or CD-ROM, and how long to display these images. Server side scripting is also used to dynamically generate static HTML, which provides the GUI to display the images. During a replay, each image viewed during the examination is displayed in the order it was examined and for the duration it was examined. Diagnostic comments made during the examination are displayed adjacent to the appropriate field. Client side scripting (JavaScript) is used to dynamically alter the main field of view locator on the slide overview, indicate the number of seconds left until the next field of view is displayed and specify the upcoming action (increase/decrease magnification/lateral movement) performed to navigate to the next field.
Using the ReplaySuite
Users are required to login into the ReplaySuite, using a unique username and password. In order to maintain user anonymity, users are assigned unique VPS ID numbers. When examination or diagnostic data is displayed, these VPS ID numbers are used to identify the examining pathologist. Fig. 2 shows a flowchart of user navigation through available functionality when using the ReplaySuite. Users may navigate between lists of examinations, graphs of diagnostic concordance for examined cases and examination replays. During replays, users may return to the last viewed list of examinations, or their own examination list.
Figure 2 Flowchart of ReplaySuite operational functionality.
Examination lists
Upon logging in, the ReplaySuite displays an Examination list of all VPS examinations performed by that user. Users may also view, on demand, an Examination list of examinations for a specific case or sub-classify to view all examinations returning a specific diagnosis for a specific case. Examination lists display summarised information about each VPS examination, and for each examination displays, from left to right, the examining pathologists identification number (ID), the slide examined, the examiners diagnosis, group consensus diagnosis for the slide and the degree of group consensus (Fig 3). Highlighting an examination, within an Examination list, displays the slide overview and case history on the left of the interface. Examination lists are interactive and provide users with three options:
Figure 3 ReplaySuite list of examinations performed by User 6.
1. View an examination list of all examinations for a selected slide
2. View Diagnostic concordance graphs for a selected slide
3. Replay any users examination of a selected slide
Diagnostic concordance graphs
Diagnostic concordance graphs are graphical representations of the variation in classification concluded by all examiners of a slide. Only data relating to slides previously examined by the user may be accessed. Two graphs are displayed; the top graph showing the variation in classification, the bottom graph the variation in sub-classification for B5 (malignant) diagnosis (Fig. 4). The number of examiners that concluded each diagnosis and the percentage concordance is also displayed in each graph. Clicking on any bar on either graph displays the list of all examinations of the selected slide that concluded with the selected classification/sub-classification. The ReplaySuite generates graphs by setting the dimensions of each bar according to the percentage of participants it represents, and can represent alternate scoring regimens by making simple changes to the system.
Figure 4 ReplaySuite variation in classification graph for slide 1.
Replaying an examination
Clicking 'View' in the Replay column on the right of an Examination list replays an examination. During a replay, each field of view examined is displayed in chronological order, appearing on-screen for the duration it was originally examined. Unlike real-time observation of an examination, replays may be paused, fast-forwarded and rewound. A slide overview, located in the top left of the interface (Fig 5), relates the location of the magnified field of view to the tissue section, and notes made by the examining pathologist are displayed in the comment box. In addition, the summary report form submitted by the examining pathologist is displayed prior to a replay (Fig. 6), and a statistical analysis of the examination afterwards (Fig. 7). This statistical analysis displays the number and type of moves performed during the examination, the amount of time spent, number of fields viewed and area of tissue examined at each magnification and the specifications of the examining pathologists PC (screen resolution, colour depth and operating system).
Figure 5 Replaying an examination of slide 2 using the ReplaySuite.
Figure 6 Summary report form from an examination replay using the ReplaySuite.
Figure 7 Statistical breakdown of a VPS examination using the ReplaySuite.
Preliminary evaluation of the ReplaySuite
Study procedure
A preliminary study was conducted to demonstrate the capabilities of the ReplaySuite to pathologists, and to evaluate their opinions on its use and potential application in training and quality assurance. Participants were provided with open access to the 10 needlecore biopsies examined during the VPS validation study [23], which they were required to examine using the VPS. Once a participant had submitted a diagnosis for a case, that participant was permitted to review their own examination data and that of other pathologists, for that slide, using the ReplaySuite. Participants who had previously participated in the VPS validation study (by examining cases) were not required to re-examine cases, and could review any examination.
Using the ReplaySuite, participants were permitted to view group concordance graphs, replay their own and others examinations, view summary report forms and view statistical breakdowns of individual examinations. Use of the ReplaySuite was monitored and recorded to the database, allowing the identification of light/heavy users and highlighting the most frequently used functions. Table 2 provides a summary of all participant use of the ReplaySuite. Cross-referencing VPS use (who examined which slides) with ReplaySuite use (who replayed which examinations) enabled the identification of which participants examinations were replayed and for which slides.
Table 2 Participant use of VPS and different functionality of ReplaySuite
VPS ReplaySuite use
User Examinations performed Total examinations replayed Own examinations replayed Other participants examinations replayed Diagnostic concordance graphs viewed Examination lists viewed Total number of functions performed Survey Submitted
10 10 13 7 6 3 19 35 Y
74 10 12 4 8 11 35 58 Y
55 10 11 10 1 28 69 108 Y
7 10 3 2 1 4 12 19 Y
36 10 2 1 1 0 4 6 Y
39 10 2 2 0 13 9 24 Y
102 10 1 0 1 0 4 5 Y
57 10 0 0 0 9 4 13 Y
60 2 0 0 0 0 1 1 Y
Total 92 44 26 18 68 157 269 9
Slides
The ten needle core biopsies examined during the VPS validation study [23] were obtained by selecting the first breast biopsy generated each month for a ten month period from the Department of Pathology, Mater Misericordiae Hospital, Dublin, Ireland. Glass slide diagnosis (clinical diagnosis) was provided by a pathologist with a special interest in breast pathology using the Core Biopsy Reporting Guidelines for Non-operative Diagnostic Procedures and Reporting in Breast Cancer screening [31], as used by the British National Co-ordinating Committee for Breast Screening Pathology. Consultation was required to finalise glass slide diagnosis on some of the cases selected.
Participants
70 pathologists were invited to participate in the study. 38 Irish participants were invited with the assistance of Professor Peter Dervan, Mater Miscordiae Hospital, Dublin, and the remaining invitees comprising the 32 members of the European Working Group of Breast Screening Pathology (EWGBSP). EWGBSP members are recognised as expert breast pathologists in their native countries and all member states of the European Union (EU) are represented within the group. 17 of the 70 invited participants previously participated in the VPS validation study [23], 4 of which were members of the EWGBSP.
Exit survey
Subsequent to the study, participants completed an electronic questionnaire on their use and impressions of the software and its potential applications. Participants were asked to submit their level of agreement/disagreement with 19 statements using a Likert scale (5-point scale e.g. strongly disagree, disagree, neutral, agree, strongly agree), and answer Yes or No to a further 5 questions. A number of questions were inverted to mitigate against the well-known bias of positively phrased questions. The following are the pertinent questions asked:
Where applicable answer yes or no to the following
• Did replaying an examination by another pathologist with a diagnosis different to your own provide you with an insight into why that diagnosis was concluded?
• Did replaying examinations performed by other pathologists make you reconsider your diagnosis for any slides?
Please state how much you agree with the following statements
• The ReplaySuite is user-friendly
• The diagnostic pathway followed by the examining pathologist was apparent
• Being able to pause, fast-forward and rewind replays was useful
• The information panel was not useful
Rate the benefit of using the ReplaySuite for the following applications
• Training
• Quality Assurance
Participants were also given the opportunity to provide open-ended feedback:
• Add any further comments, or have any additional features you would like to see incorporated into the ReplaySuite.
• If yes (to the question "Did you encounter technical difficulties while using the ReplaySuite?"), please state the difficulty that occurred
Post study survey
Participants who used the ReplaySuite were resurveyed post-study in order to determine if they regularly participated in teaching. 3 questions were asked:
• Are you, or have you been involved in providing undergraduate medical training on a regular basis?
• Are you, or have you been involved in providing postgraduate pathology training on a regular basis?
• If you are not currently involved but have previously been in either activity, please state how long ago you were involved.
Results
Study participation
9 participants used the ReplaySuite to review examination data. All 9 completed the electronic questionnaire, and all but one had more than five years experience in pathology practice, the exception possessing three years experience. 4 of the 9 participants were members of the European Working Group of Breast Screening Pathology. Table 1 summarises study participation. All 9 participants completed the post-study survey.
Table 1 User participation in evaluation of ReplaySuite technology
Irish EWGBSP Total
Invited to participate 38 32 70
Used ReplaySuite 5 4 9
Completed electronic survey 5 4 9
Using the ReplaySuite
7 participants (77.7%) replayed at least one examination and of these, 3 participants replayed more than 10. All participants who replayed at least one examination agreed or strongly agreed that the ReplaySuite was user-friendly (Fig. 8) and 85.71% (6/7) of these agreed or strongly agreed that being able to pause, fast-forward and rewind an examination replay was useful. 6 of the 7 of participants (85.71%) who replayed an examination considered replays to be of some or of great importance as a ReplaySuite feature. 66.6% (6/9) of all participants viewed diagnostic concordance graphs for at least 2 slides and 2 participants (44.4%) viewed graphs for at least 7 slides.
Figure 8 Levels of agreement with the statement 'The ReplaySuite is User-Friendly'.
Diagnostic re-evaluation
All participants who replayed at least one examination agreed or strongly agreed that the diagnostic pathway of the examining pathologist was apparent during a replay. Of the 7 participants who replayed their own or others examinations, 6 (85.71%) replayed another pathologists examination. Of these 6 participants, 5 (83.3%) agreed that it provided an insight into the examining pathologists diagnosis. Of those who replayed another's examination, 2 (33.3%) reconsidered their original diagnosis, 1 re-diagnosing concordant with group consensus and 1 re-diagnosing concordant with original glass slide diagnosis.
After originally concluding an original diagnosis of B2 (Benign) for Slide 5, User 7 replayed his/her own examination, and then replayed an examination of Slide 5 by User 6. This examination (User 6) concluded a diagnosis of B5 (Malignant), concordant with group consensus.
User 10 concluded an original diagnosis for Slide 8 of B4 (Suspicion of malignancy), which concurred with group consensus but not glass slide diagnosis, B3 (Benign but of uncertain malignant potential). User 10 replayed their own examination, then an examination of the same slide that concluded the same diagnosis by User 5. Two examinations concluding a B3 diagnosis were then replayed (Users 87 and 55), concordant with glass slide diagnosis.
ReplaySuite applications
When asked to rate the potential benefit of the ReplaySuite in pathology training, 6 out of 7 of participants who had replayed at least one examination (85.71%) and all of those who had replayed more than 10 examinations rated it as of some or of great benefit (Fig. 9). Of the 9 that completed the post-study survey, 7 and 8 stated they were currently, or had been involved in providing undergraduate medical training and postgraduate pathology training on a regular basis respectively.
Figure 9 Levels of perceived potential benefit of the ReplaySuite in Pathology Training.
When asked to rate the potential benefit of the ReplaySuite in quality assurance, 5 out of 7 of participants who had replayed at least one examination (71.4%) and all participants who had replayed more than 10 examinations considered it of some or of great benefit (Fig. 10).
Figure 10 Levels of perceived potential benefit of the ReplaySuite in Pathology Quality Assurance.
Discussion
The VPS tracks users examinations of VPS virtual slides, however, in isolation this data does not directly benefit the end user. The ReplaySuite was developed to exploit the VPS's examination tracking capabilities and utilise the data in a manner that would have tangible benefits for pathologists. Using a combination of server-side (PHP) and client-side (JavaScript) open source scripting languages, it presents this data to users via an intuitive Graphical User Interface (GUI), which was considered easy to use by participants who replayed at least 1 examination (Fig. 8). Utilising the same technology as the VPS, the ReplaySuite does not require the installation of any additional software other than the customised browser, and Internet Explorer, which is the default web browser used by 95% of the web browser market [33].
Use of the VPS for diagnosis has already demonstrated "substantial" diagnostic agreement between users, 88.23% of examining pathologists obtaining a Kappa of between 0.97 and 0.65 [23]. In determining whether pathologists could benefit from using the ReplaySuite, the primary consideration was whether participants might reassess their diagnosis as a result of observing VPS examinations of the same slides by others. In the case of 2 users, observing another's examination caused them to reconsider their original diagnosis. In contrast to the re-diagnosis of Slide 8 by User 10: B4 (suspicion of malignancy) to B3 (Benign but of uncertain malignant potential), the difference between original and secondary diagnosis of Slide 5 by User 7: B2 (benign) to B5 (malignant) is a significant re-evaluation, one which would result in significantly different courses of treatment in a clinical environment. While the degree of discordance between original diagnosis and group consensus may, in part be attributable to poor screen resolution used during examination (640 × 480 pixels), it may also be related to the relative difficulty of the case; Slide 5 achieved the second lowest group consensus (47.1%). It cannot be suggested, based on these individual examples, that group concordant re-diagnosis subsequent to ReplaySuite use will be the rule rather than the exception. However, these examples are worth noting, as it highlights the fact that using the system can result in diagnostic re-evaluation.
A number of caveats should be considered when reviewing study data, the first being small sample size. There is a significant difference between the number of pathologists invited to participate and those who completed the study (9/70). While not unique in studies involving pathologists use of telepathology systems, the small sample size may be considered a potential source of error. Bamford et al (2003) have reported similar difficulties with low participation rates [34], however, other studies have also attempted to evaluate telepathology tools using small sample sizes [35]. Low participation rates may be due to a number of factors. Bamford et al (2003) cited technical difficulties and pathologists workloads as principal factors for low participation [34]. System speed was highlighted as an issue by a number of participants during the initial VPS evaluation study [23]. As both VPS and ReplaySuite systems utilise similar technology, speed may be a potential contributing factor to low participation rates for this follow-on study. User 36 illustrated this when asked to comment on any technical difficulties encountered during the ReplaySuite evaluation study, stating, "It took a great deal of time downloading image via dial up network connection". User 60 also commented to this effect regarding technical difficulties, commenting "very slow".
Secondly, it is not unreasonable to suggest that the positive evaluation of the ReplaySuite may in part be attributable to bias from participants (4/9) who also previously participated in the VPS validation study. Pathologists who participated in both VPS and ReplaySuite studies expressed greater satisfaction and confidence in the VPS during the VPS validation study than pathologists who participated only in the VPS study [23]. However during the ReplaySuite study, these participants who participated in both did not provide more favourable evaluation of the ReplaySuite than participants who only participated in the ReplaySuite study. In Figure 8, the participant who considered the ReplaySuite the most user-friendly did not participate in the VPS evaluation study. In Figure 9, all participants who considered the ReplaySuite of greatest benefit in training were participants who did not participate in the VPS evaluation study, and both participants and non-participants in the VPS evaluation study considered the ReplaySuite of great benefit in quality assurance studies (Figure 10).
It also goes without question that those who participate in studies of this nature are often early adapters and often possess a positive bias towards new technology. This is an issue when evaluating any new software, and in that context, bias is unavoidable. 20 of the 70 participants invited to participate had foreknowledge of the VPS, however none, had previously seen the ReplaySuite and, therefore, had no preconceived notions about the software.
The concept of a virtual double-headed microscope is not new [36], however previous references to such concepts utilised video-conferencing that was dependent on the presence of an expert. While both pathologists were not required to be present at the same location, they were required to be available at the same time. In contrast, archived VPS examinations are available at any time at any location, irrespective of expert availability. Unlike real-time observation of examinations, the ReplaySuite permits examinations to be paused, fast-forwarded and rewound, and images may be annotated, providing a description of visual features within the field of view.
The benefits of various online tools and their impact on diagnostic performance have already been highlighted, however the ReplaySuite possesses a number of practical advantages. Many online tools present the diagnostic opinion of one pathologist. In contrast, the ReplaySuite can replay examinations of the same slide by multiple experts, illustrating a number of different diagnostic pathways that corroborate the same diagnostic hypothesis. Alternatively, reviewing pathologists may observe examinations that disagreed with group consensus, in order to identify the possible sources of disagreement. This is of particular interest for disorders that suffer from a high degree of inter-observer variability. Additionally, interactive tutorials and annotated virtual slides require considerable time to create, however, individual authoring time with the VPS/ReplaySuite is around 6 minutes, per user, per slide.
Figures 9 &10 illustrate that the more participants used the system, the greater potential benefit they perceived it having in pathology training and quality assurance. All 3 participants (Users 10,55,74) who replayed more than ten examinations considered the ReplaySuite of some or of great potential benefit in pathology training and quality assurance. While small sample size precludes the significance of a relationship between heavy use and favourable perception, it is not unreasonable to suggest that participants who fully appreciate the capabilities of the ReplaySuite will possess a more considered opinion.
Conclusion
The objective of the study was to develop the ReplaySuite technology, and to assess the opinions of pathologists on its use and potential application. Participants concluded the ReplaySuite to be of some or of great potential benefit to pathology training and quality assurance and considered the ReplaySuite to be beneficial in evaluating the diagnostic trace of an examination. Future evaluation of the ReplaySuite in a larger quality assurance study and training environment is anticipated.
One-to-one human tutoring remains, and will remain for the foreseeable future, the pre-eminent means of nurturing pathologist's diagnostic skills. However, as factors such as expert availability and case diversity vary, from institution to institution, it is impossible to replicate the same training environment universally. Supplementary online training tools may well help to redress the balance, providing wide access to expert pathologists and a diverse range of cases from multiple institutions, enabling an educational diversity often unachievable in a single institution. The ReplaySuite is a unique tool that permits pathologists to learn from multiple experts in a manner not possible with current tools.
Competing interests
DS and SC possess share holdings in Slidepath Ltd, a company delivering informatics solutions for use in pathology quality assurance. DJ and PD have share options in SlidePath Ltd.
Authors' contributions
DS conceived the ReplaySuite and participated in the design of the study. DJ created the ReplaySuite, carried out the study and drafted the manuscript. SC created the VPS and participated in the design of the database. PD participated in the design of the ReplaySuite and provided a pathologist's insight.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The Authors would like to thank the Health Research Board, Ireland for their funding, and the pathologists who participated in this study.
==== Refs
Mairinger T Netzer TT Schoner W Gschwendtner A Pathologists attitudes to implementing telepathology J Telemed Telecare 1998 4 41 46 9640709
Della Mea V Cortolezzis D Beltrami CA The economics of telepathology – a case study J Telemed Telecare 2000 6 S168 169 10794009
Steinberg DM Ali SZ Application of virtual microscopy in clinical cytopathology Diagn Cytopathol 2001 25 389 396 11747236 10.1002/dc.10021
Demichelis F Barbareschi M Dalla Palma P Forti S The virtual case: a new method to completely digitize cytological and histological slides Virchows Arch 2002 441 159 164 12189506 10.1007/s00428-001-0561-1
Dee FR Lehman JM Consoer D Leaven T Cohen MB Implementation of virtual microscope slides in the annual pathobiology of cancer workshop laboratory Hum Pathol 2003 34 430 436 12792915 10.1016/S0046-8177(03)00185-0
Heidger PM JrDee F Consoer D Leaven T Duncan J Kreiter C Integrated approach to teaching and testing in histology with real and virtual imaging Anat Rec 2002 269 107 112 12001217 10.1002/ar.10078
College of Medicine, University of Illinois at Urbana-Champaign. Atlas of histology
The WHO/ISUP Consensus Classification of Urothelial (Transitional Cell) Neoplasms tutorial and proficiency test
Kronz JD Silberman MA Allsbrook WC JrBastacky SI Burks RT Cina SJ Mills SE Ross JS Sakr WA Tomaszewski JE True LD Ulbright TM Weinstein MW Yantiss RK Young RH Epstein JI Pathology Resident's Use of a web-based tutorial to improve Gleason grading of prostate carcinoma on needle biopsies Hum Pathol 2000 31 1044 50 11014569 10.1053/hupa.2000.16278
University of Kansas Medical Center, Basic Histopathology website
Loyola University, Chicago. Dermatology Medical Education Website
School of Anatomy and Human Biology – The University of Western Australia. Blue Histology
Ponton L Break IT – The Breast Cancer Pathology Information Kit Breast Cancer Res 1999 1 0003
The Johns Hopkins Medical Institutions, Department of Pathology, Division of Hematopathology. Flow cytometry tutorial
The virtual hospital atlas of liver pathology
Weinstein LJ Epstein JI Edlow D Westra WH Static image analysis of skin specimens: the application of telepathology to frozen section evaluation Hum Pathol 1997 28 30 35 9013828 10.1016/S0046-8177(97)90275-6
Halliday BE Bhattacharyya AK Graham AR Davis JR Leavitt SA Nagle RB McLaughlin WJ Rivas RA Martinez R Krupinski EA Weinstein RS Diagnostic accuracy of an international static-imaging telepathology consultation service Hum Pathol 1997 28 17 21 9013826 10.1016/S0046-8177(97)90273-2
Weinstein MH Epstein JI Telepathology diagnosis of prostate needle biopsies Hum Pathol 1997 28 22 29 9013827 10.1016/S0046-8177(97)90274-4
Raab SS Zaleski MS Thomas PA Niemann TH Isacson C Jensen CS Telecytology-diagnostic accuracy in cervical-vaginal smears Am J Clin Pathol 1996 105 599 603 8623769
Weinberg DS Allaert FA Dusserre P Drouot F Retailliau B Welch WR Longtine J Brodsky G Folkerth R Doolittle M Telepathology diagnosis by means of digital still images: an international validation study Hum Pathol 1996 27 111 118 8617451 10.1016/S0046-8177(96)90363-9
Walter-Reed Army Medical Center. Pathology and area laboratory sciences: pathology to go!
Crowley RS Naus GJ Stewart J Friedman CP Development of Visual Diagnostic Expertise in Pathology. An Information Processing Study J Am Med Inform Assoc 2003 10 39 51 12509356 10.1197/jamia.M1123
Costello SP Johnston DJ Dervan PA O' Shea DG Development and Evaluation of the Virtual Pathology Slide: A New Tool in Telepathology J Med Internet Res 2003 5 e11 12857667 10.2196/jmir.5.2.e11
Bacus Laboratories Inc. Microscope imaging and virtual microscopy products for pathology, research and medical education
Fairfield Imaging. Software for digital pathology
Virtual microscope pathology images for pregraduate students of pathology and dermatology: Melanocytic lesions
Steinberg DM Ali SZ Application of virtual microscopy in clinical cytopathology Diagn Cytopathol 2001 25 389 396 11747236 10.1002/dc.10021
Aperio Technologies. Virtual microscopy solutions for analysis and discovery
Costello SP Johnston DJ Dervan PA O'Shea DG The virtual pathology slide: a new internet telemicroscopy tool for tracing the process of microscopic diagnosis and evaluating pathologist behaviour [Abstract] Arch Pathol Lab Med 2002 126 781 802
Costello SP Johnston DJ Dervan PA O' Shea DG Evaluation of the virtual pathology slide: using breast needle core biopsy specimens [Abstract] Br J Cancer 2002 S1 34 12085157
Non-operative Diagnosis Subgroup of the National Coordinating Group for Breast Screening Pathology. Guidelines for non-operative diagnostic procedures and reporting in breast cancer screening
ReplaySuite Technical Document
Internet Browser market share
Bamford WM Rogers N Kassam M Rashbass J Furness PN The development and evaluation of the UK national telepathology network Histopathology 2003 42 110 9 12558742 10.1046/j.1365-2559.2003.01540.x
Lee ES Kim IS Choi JS Yeom BW Kim HK Han JH Lee MS Leong AS Accuracy and reproducibility of telecytology diagnosis of cervical smears. A tool for quality assurance programs Am J Clin Pathol 2003 119 356 60 12645336
Furness P Rashbass J The virtual double-headed microscope: telepathology for all? Histopathology 2000 36 182 183 10733320 10.1046/j.1365-2559.2000.0866a.x
| 15845147 | PMC1087846 | CC BY | 2021-01-04 23:52:12 | no | BMC Med Inform Decis Mak. 2005 Apr 21; 5:10 | utf-8 | BMC Med Inform Decis Mak | 2,005 | 10.1186/1472-6947-5-10 | oa_comm |
==== Front
BMC NeurosciBMC Neuroscience1471-2202BioMed Central London 1471-2202-6-251581999610.1186/1471-2202-6-25Research ArticleNuclear localization of Annexin A7 during murine brain development Rick Michaela [email protected] Garrido Soraya I [email protected] Claudia [email protected] Dietmar R [email protected] Angelika A [email protected] Christoph S [email protected] Center for Biochemistry, Institute of Biochemistry I, Medical Faculty and Center for Molecular Medicine Cologne, University of Cologne, Joseph-Stelzmann-Str. 52, 50931 Köln, Germany2 Institute of Neuropathology, University Hospital Bonn, Sigmund-Freud Str. 25, 53105 Bonn, Germany2005 10 4 2005 6 25 25 10 11 2004 10 4 2005 Copyright © 2005 Rick et al; licensee BioMed Central Ltd.2005Rick 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
Annexin A7 is a member of the annexin protein family, which is characterized by its ability to interact with phospholipids in the presence of Ca2+-ions and which is thought to function in Ca2+-homeostasis. Results from mutant mice showed altered Ca2+-wave propagation in astrocytes. As the appearance and distribution of Annexin A7 during brain development has not been investigated so far, we focused on the distribution of Annexin A7 protein during mouse embryogenesis in the developing central nervous system and in the adult mouse brain.
Results
Annexin A7 is expressed in cells of the developing brain where a change in its subcellular localization from cytoplasm to nucleus was observed. In the adult CNS, the subcellular distribution of Annexin A7 depends on the cell type. By immunohistochemistry analysis Annexin A7 was detected in the cytosol of undifferentiated cells at embryonic days E5–E8. At E11–E15 the protein is still present in the cytosol of cells predominantly located in the ventricular germinative zone surrounding the lateral ventricle. Later on, at embryonic day E16, Annexin A7 in cells of the intermediate and marginal zone of the neopallium translocates to the nucleus. Neuronal cells of all areas in the adult brain present Annexin A7 in the nucleus, whereas glial fibrillary acidic protein (GFAP)-positive astrocytes exhibit both, a cytoplasmic and nuclear staining. The presence of nuclear Annexin A7 was confirmed by extraction of the nucleoplasm from isolated nuclei obtained from neuronal and astroglial cell lines.
Conclusion
We have demonstrated a translocation of Annexin A7 to nuclei of cells in early murine brain development and the presence of Annexin A7 in nuclei of neuronal cells in the adult animal. The role of Annexin A7 in nuclei of differentiating and mature neuronal cells remains elusive.
==== Body
Background
Annexins form a family of structurally related proteins, which bind to negatively charged phospholipids in a Ca2+-dependent manner [1]. They are characterized by a bipartite structure with a conserved C-terminal core domain and a unique N-terminal domain that varies in length and amino acid composition. The C-terminal domain is formed by either a four- or eightfold repeat of approximately 70 amino acids, each repeat carrying a Ca2+-binding site, and is responsible for phospholipid binding. The N-terminal regions are thought to confer functional diversity to the annexin proteins [2]. The biochemical features in vitro were analyzed extensively, but the in vivo functions of annexins remain unclear.
Annexin A7, the first annexin to be described, was isolated as the agent that mediated aggregation of chromaffin granules and fusion of membranes and phospholipids in the presence of Ca2+-ions [3]. Expression studies demonstrated the distribution of Annexin A7 in a wide variety of tissues and cells mainly enriched in the cytosol in close association with membranous structures, but it was also described in the nucleus of adrenal chromaffin cells [4]. The presence of an alternatively spliced cassette exon gives rise to two Annexin A7 isoforms corresponding in molecular mass to 47 kDa and 51 kDa. The isoforms differ in their N-terminal domain and exhibit a tissue-specific expression pattern. The 47 kDa isoform is present in all tissues except for skeletal muscle, where the 51 kDa isoform is exclusively present. Heart muscle, brain tissue and red blood cells contain both isoforms [5-8]. Previous studies indicated that the subcellular localization of Annexin A7 changes during myoblast differentiation. In undifferentiated cells the protein is equally distributed between cytosol and membrane fractions while in differentiated cells it is exclusively present in the membrane fraction [7]. Reports by Clemen et al. [9] and Herr et al. [8,10] demonstrated roles for Annexin A7 in shape and osmotic resistance of red blood cells, platelet aggregation velocity, and in the velocity of spreading astrocytic Ca2+-waves. Annexin A7 is also involved in the maintenance of regular cardiac electrophysiology and Ca2+-homeostasis [Schrickel et al., submitted]. Detailed reports on appearance and distribution of Annexin A7 during brain development are not available. In the present study we focus on the distribution of Annexin A7 in the developing brain of mice embryos between E5 and E16, and in the adult mouse brain.
Results
Annexin A7 is expressed in the early mouse embryo
First we examined the expression of Annexin A7 in ES cells (Bruce4, established from C57BL/6J mice) and the early stages of mouse embryonic development at the mRNA level by northern blot analysis and at the protein level by Western blotting and immunohistochemistry, respectively. Northern blot analysis shows in ES-cells and at embryonic days E7, E11, E15, and E17 two mRNA species of 1.8 kb and 2.4 kb, which result from alternative splicing in the untranslated 3'end (Fig. 1A) [11]. We found similar Annexin A7 mRNA levels in the four embryonic stages. Reprobing with a β-actin probe verified equal loading; the appearance of a faster migrating mRNA species in addition to the 2.0 kb species is characteristic for β-actin [12]. On the protein level, ES cells express only the smaller Annexin A7 isoform of 47 kDa (Fig. 1B, neuro-2a cells and heart tissue are included for control).
Figure 1 Expression of Annexin A7. (A) Northern blot analysis of RNA from ES-cells and embryos E7, E11, E15, E17 probed with a full length annexin A7 cDNA. Transcripts of 1.8 and 2.4 kb are detected in all stages examined (upper panel). Reprobing with a β-actin probe verified equal loading (lower panel, 2.0 kb band). (B) Murine ES cells (ES) express only the small Annexin A7 isoform. Both isoforms are expressed in neuroblastoma neuro-2a cells (N2A), heart (HRT) and brain tissue (CNS) from adult mice. Proteins from neuro-2a and ES cells, heart and brain were resolved by SDS polyacrylamide gel electrophoresis (12% acrylamide) and transferred to nitrocellulose membranes. The Western blot was probed with mAb 203–217 followed by incubation with a peroxidase coupled secondary antibody. Detection was with enhanced chemiluminescence.
Immunohistochemistry confirmed the presence of the protein during early development. Annexin A7 immunoreactivity using mAb 203–217, which specifically recognizes both isoforms of Annexin A7, can be observed in the two cell types of the cylinder stage at E5 with a weaker Annexin A7 expression in the ectoderm (Fig. 2A–C). At E8 Annexin A7 can be detected in almost all tissues of the embryo (data not shown). A correlation between the presence of Annexin A7 and the origin of a cell type from one of the three germ layers is not apparent. We specifically noted the localization of the protein in the neuroepithelium of the neural fold and tube (Fig. 2D–F, proximal, E8; Fig. 2G–I, distal, E13). At the subcellular level, Annexin A7 was present in punctate structures mainly in the cytosol of all examined cells (Fig. 2E,F). This distribution was also observed during subsequent development.
Figure 2 Annexin A7 immunoreactivity in early mouse embryos. (A) Phase contrast, embryo E5: The egg cylinder consists of an inner cell mass (a) representing the ectoderm and an outer layer of endoderm cells (b). (B) Immunostaining of the paraffin section was performed using purified mAb 203–217 and Cy3-conjugated anti-mouse IgG. Annexin A7 is expressed in both cell types of the egg cylinder with a strong staining of the endoderm and a weaker staining of the ectoderm. The nuclei are devoid of immune reactions. (C) Negative control using the secondary Cy3-antibody only. (D-F) Annexin A7 expression in the proximal neural tube (D) and nearby neural fold (E,F), embryo E8, transverse section. Immunolabeling of Annexin A7 was performed with purified mAb 203–217 and visualization was with an Alexa Fluor 488-conjugated anti-mouse IgG. (D) An intense Annexin A7 immunostaining is detectable in the neuroepithelium of the neural tube (a, lumen of neural tube). (E,F) Higher magnifications of the neuroepithelium show that Annexin A7 is expressed in the cytosol. Arrowheads point to Annexin A7 immunoreactivity in the cytosol. (G) Phase contrast, embryo E13, caudal neural tube. (H) Immunostaining of the paraffin section was performed using purified mAb 203–217 and Alexa Fluor 488-conjugated anti-mouse IgG. (I) Negative control using the secondary Alexa Fluor 488-antibody only. Bar, 20 μm.
Annexin A7 changes its subcellular localization in the embryonic brain at E16
Next we analysed the cellular and subcellular distribution of Annexin A7 in the developing mouse brain (Fig. 3). At E13–E16 in transverse sections of the embryonic brain, most of the immunoreactive cells are present in the ventricular germinative zone surrounding the lateral ventricle (Fig. 3A, overview E16; Fig. 3H, overview E13). A closer inspection revealed that Annexin A7 is mainly localized in the cytosol at E13 (Fig. 3B,E). During further development the immature cells have rounded up two days later, but they retain Annexin A7 in the cytosol at E15 (Fig. 3C,F). At E16 we observed the first prominent nuclear staining of Annexin A7 (Fig. 3D,G) in cells of the intermediate zone (Fig. 3I, oval), which contains neurons radially migrating towards the growing neopallium cortex [13,14] and also glial cells forming the white matter of the adult cortex. Cells located marginally in the neopallial cortex also exhibit a nuclear stain. Labelling in the ventricular germinative zone is less prominent and the cytosol is no longer more strongly stained than the nucleus.
Figure 3 Subcellular localization of Annexin A7 in embryos E13, E15 and E16. Paraffin sections of embryonic brain were stained with purified mAb 203–217. Annexin A7 was visualized with Alexa Fluor 488-conjugated secondary antibody. (A) Overview, at E16 the immature GFAP-negative cells of the forming cerebral neocortex (b) surrounding the lateral ventricle (a) are strongly stained; square, a higher magnification of this area is given in (I). (B) Higher magnification of the earlier stage E13 (H) shows, that the cells are stained in the cytosol (arrowhead). (C) Two days later at E15 the cells have rounded up, and Annexin A7 stays in the cytosol. (D) At E16 the first nuclear staining becomes apparent (arrowhead) in cells of the intermediate zone located between ventricular germinative zone and marginal neopallial cortex as seen in (I), oval. (E-G) Confocal microscopy confirms the results in B-D. (H) Overview, at E13 the immature GFAP-negative cells are strongly stained; a, lateral ventricle. (I) Higher magnification of a cortical section of (A, square) demonstrating ventricular, intermediate, and marginal zones; oval, a higher magnification of this area is given in (D,G).
Mature neurons in adult mouse brain show an intense nuclear staining of Annexin A7
Both Annexin A7 isoforms are found in adult brain tissue (Fig. 1B). In general, we observed two characteristic staining patterns for Annexin A7, a prominent nuclear stain in neurons (determined by lack of GFAP-immunoreactivity, localization, morphology) and a cytoplasmic and nuclear stain in astrocytes (GFAP-positive) in all areas of the mature murine brain. In the neocortex (isocortex) Annexin A7 was strongly enriched in nuclei of neurons of all six cortical laminae (Fig. 4A, section derived from cortex temporalis). For control, in the adult brain type-1 astrocytes can be characterized by a positive GFAP-stain (Fig. 4B). Fig. 4F demonstrates the presence of nuclear Annexin A7 in a neocortical neuron of the external granular layer (layer II) and in an astrocytic cell of the neocortical molecular layer (layer I) which additionally exhibits an intense signal in the cytoplasm and in cellular branches. A further cell type of the isocortex showed a strong immunoreactivity in both, nucleus and cytosol, and was identified as pyramidal neuron based on morphology and absence of GFAP immunoreactivity. However, Annexin A7 apparently is more enriched in the nucleus (Fig. 4E).
Figure 4 Annexin A7 is present in neurons and astrocytes of the cortex temporalis and hippocampal formation of 10-weeks-old mice. (A) Low magnification of the cortex temporalis presents an Annexin A7 expression in cells of the pial border, in neurons of all six isocortical laminae, and a weak signal in the adjacent white matter. (B) Corresponding section stained with GFAP. (C) Staining in the Stratum pyramidalis (a) and in the dentate gyrus of the hippocampus; square, a higher magnification of acorresponding area is given in (G,H). An intense Annexin A7 immunostaining is detectable. (D) Corresponding section stained with secondary antibody only. (E) Presence of Annexin A7 in pyramidal neurons (lamina pyramidalis externa) of the isocortex temporalis. These neurons were identified based on their morphology, distribution and lack of GFAP staining. AnnexinA7 exhibits a punctate staining, which is pronounced in the nucleus (arrowhead). (F) Higher magnification of image (A) also shows Annexin A7 in nuclei of neurons (lamina granularis externa (corpuscularis), arrowhead) and in the cytoplasm and nuclei of astrocytes (lamina molecularis, arrow; GFAP-confirmed). (G) Higher magnification of the pyramidal neurons in the hippocampus confirms the presence of the Annexin A7 protein in the nucleus (arrowhead) of mature neurons. (H) To further confirm this, a similar section derived from an AnxA7-/- mouse was stained with the annexin specific antibody and lacked the nuclear signal. The residual stain of the tissue is unspecific, as it is also observed in controls of the AnxA7-/- brain omitting the primary antibody (data not shown). All paraffin sections were stained with mAb 203–217 (A, C, E, F, G) or anti-GFAP-antibody (B). The hippocampal control section (D) lacks the primary antibody.
In the hippocampal formation we detected prominent Annexin A7 immunostaining in the stratum pyramidalis and in the dentate gyrus and also a weak astrocytic staining (Fig. 4C). The astrocytes are mainly localized in the area between cells exhibiting nuclear Annexin A7 staining and show the protein also in the nucleus and the cytoplasm (data not shown). When we analyzed the pyramidal neurons in the hippocampus at a higher magnification we found that Annexin A7 again mainly localized to the nuclei (Fig. 4G). Polyclonal antibodies gave a similar staining pattern (data not shown). The intense nuclear staining was absent in controls either stained only with the secondary antibody (Fig. 4D) or in brain sections derived from AnxA7-/- mice stained for Annexin A7 (Fig. 4H). The residual staining seen in the AnxA7-/- brain is unspecific as it is also present in corresponding negative controls lacking the primary antibody. The absence of Annexin A7 protein in brain and other tissues of the AnxA7-/- mouse has been verified in [10]. In another actual study of the knock out mouse we detected a thickened basal membrane and a widened intercellular space. This may cause unspecific binding of the secondary antibody.
An equal staining of Annexin A7 was also found in the cerebellum (Fig. 5A). Nuclei of neurons in the stratum granulosum show the most prominent staining (Fig. 5C,D). In the stratum moleculare, which is poor in cell bodies of neurons, only a few dots appear which also correspond to nuclei of neurons. Astrocytes are also positive for Annexin A7 in the nuclei and cytoplasm (data not shown). The signal of the pial border and the prominent stain of the white matter tracts (laminae medullares) are due to Annexin A7 positive astrocytes (Fig. 5 A–C). Higher magnification of the boundary layer between stratum moleculare and granulosum (Fig. 5D) revealed an Annexin A7 signal in nuclei, but also in the cytoplasm and at the plasma membrane of Purkinje cells (Fig. 5E). The typical pair of dendrites points to the margin of the cerebellar cortex. The specificity of the Annexin A7 signal was further confirmed in similar brain sections of the AnxA7-/- mouse (Fig. 5F). Fig. 5G shows Annexin A7 in neurites (axons) connecting the laminae medullares with the Purkinje-cell layer. Most likely these are the axons of the Purkinje-cells. Apart from these efferent neurites the laminae medullares (lamina alba, white matter) contain afferent mossy and climbing fibers. Thus, the intense stain of the cerebellar white matter arises from Annexin A7 located in astrocytes and neurites.
Figure 5 Annexin A7 immunostaining in the cerebellum of adult mice. (A) Low magnification of the cerebellum presents an Annexin A7 expression mainly in cells of the stratum granulosum and laminae medullares. (B) Corresponding section stained for GFAP. (C) Higher magnification of folia of the cerebellum, where a polyclonal anti-AnnexinA7 antibody was used; square, a higher magnification of acorresponding area stained with mAb 203–217 is given in (D). Between stratum granulosum and stratum moleculare the layer of Purkinje-cells (stratum neuronorum piriformium (ganglionare)) can be observed. (D) Staining of the band of Purkinje-cells (arrows). The positive Annexin A7-stain in the stratum granulosum is due to staining of the nuclei of neurons. (E) In addition to an intense staining of the nucleus, the cell body is AnnexinA7-positive including both dendrites of the Purkinje-cell shown. (F) Corresponding section from an AnxA7-/- mouse. (G) AnnexinA7 staining of axons (arrowheads) running from the laminae medullares to the Purkinje-cell layer located in the round end of a convolution. Sections A, D, E, F, G were stained with mAb 203–217.
Expression of Annexin A7 in the adult human isocortex
In the human parietal neocortex of aged individuals without any neuropathological alterations, subpial astrocytes exhibited a staining of Annexin A7 in the cytoplasm. A nuclear presence of Annexin A7 was limited to single astrocytes (Fig. 6B). Pyramidal neurons, predominantly those of layer V exhibited Annexin A7 at the plasma membrane of their perikaryon as well as of the apical dendrite (Fig. 6A,C). The neurons lacked a signal for Annexin A7 in their nuclei. Apical dendrites within the molecular layer also indicated a positive staining for Annexin A7 (Fig. 6B). The staining pattern of Annexin A7 in the human autopsy brain did not change after pre-treatment with trypsin.
Figure 6 Annexin A7 immunostaining in human isocortex. (A) Pyramidal neurons (bold arrow) and apical dendrites (small arrows) were clearly labeled with the antibody against Annexin A7 in the human parietal cortex. (B) A dendritic staining was also seen in the molecular layer (small arrows). In addition the subpial astrocytes were weakly labeled (double headed arrows). The staining was cytoplasmic, only few astrocytes showed a nuclear staining as well (inset). (C) Enlargement of (A): Annexin A7 was seen at the cell membrane of the perikaryon (bold arrow) and the apical dendrite (small arrows). (D) A blank control lacking the primary antibody did not show a specific labeling, but autofluorescent lipofuscin was detectable in the neurons (arrowheads). Bar, (A) 70 μm, (B) 50 μm, (B-inset, C) 17 μm, (D) 20 μm.
Presence of Annexin A7 in nuclei from neuronal and astroglial cells
The nuclear localisation of Annexin A7 in mice observed in immature cells at E16 and differentiated adult neurons could be only detected as a strong signal when the brain sections were pre-treated with trypsin before antibody staining. This procedure is thought to allow the antibodies to access epitopes masked by the formaldehyde fixation. Methanol fixation of cultured cells led to inconsistent results. Antigen retrieval in formalin-fixed and paraffin-embedded tissue sections employs various heating or proteolytic pre-treatment methods [[15-17]; Ein Handbuch für die Histologie, dianova GmbH, Hamburg, Germany]. These methods can result in moderate or strong specific antibody staining, but the detectability of other antigens might be decreased. Moreover, the optimal pre-treatment has to be individualised for each antigen.
To verify the presence of endogenous Annexin A7 in nuclei, we used a biochemical approach and purified nuclei from neuro-2a, PC-12, and C6 cells and treated them with a hypotonic extraction buffer to obtain the nucleoplasm (Fig. 7A). The nucleoplasm and the remaining nuclei, from which the nucleoplasm has been extracted partially, were subjected to SDS-PAGE and Western blotting. Western blots probed with mAb 203–217 showed Annexin A7 in the nucleoplasm of neuronal (neuro-2a, PC-12) and astroglial (C6) cell lines (Fig. 7A), however, the large isoform could only be clearly extracted with nucleoplasm from C6 cells. For control, antibodies against LAP2α, Emerin and tubulin were used to verify that the nucleoplasm (marker:LAP2α, which is anon-membrane-bound isoform of LAP2) was successfully extracted and also not contaminated by nuclear membranes (marker:Emerin) or by cytoplasm (marker:tubulin). Likewise, in these neuronal and astroglial cell lines Annexin A7 had been observed in the nucleus by immunofluorescence (Fig. 7B and data not shown). We noticed no difference between PC-12 and neuro-2a cells, however, as for the brain sections, we found an increase of the Annexin A7 staining after pre-treatment with trypsin.
Figure 7 Nuclear localization of Annexin A7. (A) Extraction of Annexin A7 from nuclei of neuronal (PC-12, N2A) and an astroglial (C6) cell line using a hypotonic buffer. Samples of total cells (c), and corresponding amounts of nuclear membranes (m) and nucleoplasm (p) were subjected to SDS-PAGE and Western blotting. Annexin A7 (AnxA7) isoforms are extractable from the nucleus of the indicated cells (47 kDa and 51 kDa isoform, arrowhead). For control, immunoblotting of LAP2α (75 kDa; non-membrane-bound isoform), Emerin (EM, 34 kDa), and tubulin (TB, 55 kDa) which are specific for nucleoplasm (LAP2α), nuclear membranes (Emerin), and the cytoplasm (tubulin) are shown. Note that the nucleoplasm is only partially extractable. (B) Immunofluorescence images of the PC-12 cells used. Cells were fixed with paraformaldehyde, permeabilized with Triton X-100, pretreated with trypsin as indicated, and stained with Annexin A7 specific mAb 203–217, DAPI in blue, bar 10 μm.
Discussion
In the present study we explored the appearance of Annexin A7 during mouse development at the mRNA and protein level and focused on the central nervous system during embryogenesis. Northern blot and immunohistochemistry analysis show the expression of Annexin A7 in all embryonic tissues from day E5 on, the earliest day studied. Differentiating cells, like the neural ectoderm, which is the origin of the cells belonging to the nervous system, show Annexin A7 immunoreactivity mainly in the cytosol. Endodermal as well as mesodermal cells exhibit a similar subcellular localization of the protein.
In the developing brain we noted a striking change in the subcellular distribution of Annexin A7. Cells in the stratum germinativum of the neopallial cortex, which surrounds the lateral ventricle, show at E13 and E15 a staining for Annexin A7 mainly in the cytosol. But this staining largely disappears and at the following day E16 we detect Annexin A7 in the nucleus of cells in the intermediate and marginal zone of the neopallium. The different expression patterns of Annexin A7-positive cells from the ventricular germinative zone to the marginal zone of the later neocortex observed in this study are similar to the developmental patterns of Tenascin-C-positive astroglial precursors following the guidance of the radial glial cells [18]. Moreover, the patterns resemble that of migrating and differentiated neurons described by Berry et al. [13]. Neurons that are generated prenatally in the proliferative ventricular layer of the neopallial cortex subsequently migrate through the intermediate zone to form the different cortical cell layers in a declining inside-out gradient of cell maturation. These observations suggested that the subcellular localization of Annexin A7 depends on developmental stage and cell type.
In the adult brain we generally observed a nuclear localization for Annexin A7 in neurons whereas astrocytes exhibited both a cytosolic as well as a nuclear staining. However, pyramidal neurons of the isocortex and Purkinje-cells of the cerebellum exhibited a cytosolic stain of intermediate intensity including their neurites and dendrites. We have previously reported the expression of Annexin A7 in human temporal neocortex and hippocampus obtained from neurosurgery for therapy-refractory epilepsy and found the two Annexin A7 isoforms restricted to the cytoplasm and nuclei of astrocytes, whereas neurons lacked any signal [19]. The hippocampal area showed typical signs of Ammon's horn sclerosis, but the adjacent temporal neocortex tissue did not show any histopathological alterations. This data is in discrepancy with our actual observation in the murine brain. To determine if this is due to different methods in immunofluorescence staining or indeed different expression patterns in mouse and human, we repeated the immunofluorescence of human brain. This time we investigated sections from human parietal cortex of autopsy brain. The astroglial expression of Annexin A7 could be confirmed, although not all astrocytes exhibited the nuclear staining. Pyramidal neurons however indicated a distinct staining of Annexin A7 most prominent along their dendrites. The different results obtained for the neurons may be explained by the fact that we previously used frozen tissue sections or that they were derived from patients suffering from temporal lobe epilepsia. On the other hand the actual human brain tissue used was from aged patients and did not correspond to the age of the mice included in this study. In cultured cells after cell damage or apoptosis (unpublished observations) or in cells treated with a Ca2+-ionophore Annexin A7 translocated from the cytoplasm to cellular membranes [19]. We therefore favor the hypothesis that Annexin A7 in the sensitive neurons of the human autopsy brain may have similarly translocated to the cellular membrane. This property of translocation and membrane binding is common to all annexins and commercially available kits for apoptosis detection employ recombinant AnnexinA5. The presence of nuclear Annexin A7 in murine brain was confirmed by controls using sections from the AnxA7-/- mouse and by a biochemical extraction of the protein from the nucleus. Both Annexin A7 isoforms described in brain tissue seemingly are expressed by neurons and astrocytes, which was shown using total cell extracts of cultured neuo-2a, PC-12, and C6 cells. In addition to their expression in brain, both Annexin A7 isoforms have only been described in heart muscle and red blood cells [5-8].
Although the inactivation of the annexin A7 gene did not interfere with the viability and development of knock out mice [10], their generation allowed us to address the role of Annexin A7 in specific cell types [8,9]. Indeed, when we analyzed primary astrocytes from an AnxA7-/- mouse for Ca2+-dependent signaling processes, we found that they exhibited a significantly increased velocity of mechanically induced astrocytic Ca2+-waves as compared to wild type [9]. This led us to propose, that Annexin A7 can act as a Ca2+-buffer and is involved in Ca2+-homeostasis. In neurons Ca2+ ions play major roles in various physiological and pathophysiological processes [20-25]. One can speculate about an involvement of Annexin A7 in the regulation of these Ca2+-dependent processes, propositions that need further investigation. However, such roles are confirmed for heart function by studies of Schrickel et al. [submitted], who described an involvement of Annexin A7 in the maintenance of a regular cardiac electrophysiology and Ca2+-homeostasis.
An altered subcellular location during embryogenesis was also reported for Annexin A11. The developing gray matter of the rat embryonic spinal cord exhibited primarily nuclear localization of Annexin A11, while immunoreactivity was lost from the nuclei in the adult spinal cord [26]. In contrast, our studies show a relocation of Annexin A7 from the cytosol to the nucleus in cells of the embryonic neuronal tissue. The Annexin A7 distribution is determined by a variety of factors. An important one appears to be Ca2+, which promotes binding of Annexin A7 to membranes and also allows aggregation of annexins [27]. Binding partners of Annexin A7 such as sorcin might represent additional factors [28]. Although the members of the annexin family are generally found at the plasma membrane, in the cytoplasm or in association with the cytoskeleton, Annexins A1, A2, A4, A5 and A11 have been described to be localized at least partially in the nucleus [26,29,30]. Studies with human foreskin fibroblasts demonstrated that Annexin A1, A4 and A5 are all present in the nucleus at higher concentration than in the cytosol [31]. Raising intracellular Ca2+ led to relocation of these annexins to the nuclear membrane. An important role for annexins in mediating the Ca2+-signal within the nuclei of the fibroblasts was proposed. These results mirror studies with stably transfected C6 cells, in which high intracellular Ca2+-concentrations induced a marked redistribution of Annexin A7 from its localization in the nucleoplasm to the nuclear membrane [19].
None of the annexins contains a typical nuclear localization signal and their mechanism of nuclear import remains to be elucidated. For Annexin A11 it was shown that nuclear localization is mediated by its N-terminal region, which also contains a binding side for the S100 protein calcyclin [29]. More recently Tomas and Moss [32] showed, that Annexin A11 and S100A6 assemble at the nuclear envelope during nuclear breakdown. Their role in this process is not known. In general it seems that the nuclear localization of the annexins is actively regulated. For example the nucleocytoplasmic compartmentalization of Annexin A2 is controlled by sequestration of the AnxA2/p11 complex modulated by phosphorylation and by a nuclear export signal found in the AnxA2 3–12 region [33]. One function of Annexin A2 in the nucleus, that appears not to involve binding of p11, has been suggested by its purification as part of a primer recognition protein complex that enhances DNA polymeraseα-activity in vitro [34]. The annexins may participate in a nuclear response to initial cell stimulation or to a Ca2+-transient, presumably by regulating DNA replication. For Annexin A7 such a pathway is very speculative at the moment. Future studies are however directed by these findings and will concentrate on the identification of the nuclear localization signal of Annexin A7 as well as on the role of Annexin A7 in the nuclear compartment.
Conclusion
In this article we report the translocation of Annexin A7 to nuclei of differentiating cells in the developing murine brain. In the adult brain Annexin A7 generally was detected in nuclei of neurons. Astrocytes, cerebellar Purkinje-cells and neocortical pyramidal neurons exhibited both, Annexin A7 in the cytosol and in the nuclei. Thus, the subcellular localisation of Annexin A7 depends on the developmental stage and the cell type. A role of Annexin A7 as well as of other annexins in nuclei remains speculative. The results obtained by immunofluorescence were confirmed by extraction of Annexin A7 from nuclei of neuronal and glial cell lines.
Methods
Animals and tissue preparation
129SV strain mice were used. The day on which the vaginal plug was confirmed was defined as E0. All the animal experiments were performed in accordance with the Guide for the Care and Use of Laboratory Animals (NIH Publication No. 85-23). Pregnant females were sacrificed by cervical dislocation at various intervals between E5 and E17. Before E9, embryos were embedded within the maternal uterus. After E9, embryos were removed from their membranes and embedded separately. Experiments were performed with four to six animals at each developmental stage. The mature brain used was derived from mice of three to four month of age. The embryos and tissues were fixed in freshly prepared 4% paraformaldehyde solution for 2–12 h depending on the tissue thickness [35]. They were dehydrated with ethanol and embedded in paraffin according to standard procedures. Microtome sections were cut at 7–10 μm, placed on slides and stored until further processing.
Human brain tissue
Five human autopsy brains of male and female patients without any neuropathological changes (age 61–77 years; post mortem interval: 12–72 h) were fixed in an aqueous solution of formaldehyde for three weeks. Blocks of the parietal neocortex were embedded in paraffin and microtomed at 12 μm thickness. Sections were stained with aldehydefuchsin Darrow-red to confirm the absence of neuropathological alterations. All procedures were conducted in accordance with the Declaration of Helsinki and approved by the ethics committee of the University of Bonn Medical Center.
Antibodies
Antibodies used in this study were mAb 203–217 directed against the core domain of mouse Annexin A7 [6] (purified IgG, 1:50) specifically recognizing both Annexin A7 isoforms [[6,10], and data not shown], a polyclonal antibody against mouse Annexin A7 [10] (1:200), a polyclonal antiserum raised against the astrocyte-specific intermediate filament glial fibrillary acidic protein (GFAP; DAKO, 1:400), anti-β-tubulin (Sigma, 1:1000), anti-Emerin (novocastra, 1:2000), and polyclonal anti-LAP2α (kindly provided by Ronald Foisner, 1:5000).
Immunofluorescence microscopy
Immunohistochemical staining using mAb 203–217 as the primary antibody was with AlexaFluor488-conjugated goat anti-mouse IgG (Molecular Probes, Leiden, The Nether-lands) or Cy3-labeled goat anti-mouse IgG as the secondary antibodies. Polyclonal anti-Annexin A7 antibody and anti-GFAP-antibody were combined with Alexa Fluor 568-conjugated goat anti-rabbit IgG (Molecular Probes). The specific labelling of Annexin A7 detected with the anti-mouse-IgG secondary antibodies in mouse tissue sections was confirmed using primary antibodies, which were directly coupled to Alexa Fluor488.
The paraffin-embedded sections were first treated three times for 2 min in xylene to remove paraffin and then rehydrated step by step with descending concentrations of ethanol for antibody staining [35]. After three washings in PBS (137 mM NaCl, 2.7 mM KCl, 8.1 mM Na2HPO4, 1.5 mM KH2PO4, pH 7.4), the tissue sections were incubated in PBS containing 0.1% trypsin for 10 min at room temperature to free formaldehyde-crosslinked epitopes [15-17]. After additional three washes with PBS, non-specific binding sites were blocked by incubating the slides with PBG (0.5 % BSA, 0.045 % fish gelatine in PBS) in 10% normal goat serum (NGS) for 2 h. Incubation with the primary antibodies was performed at 4°C overnight diluted in PBG supplemented with 1 % NGS. After five washes in PBG for 5 min each, the slides were incubated for 1 h with the secondary antibody at room temperature, washed three times in PBG and three times in PBS, 5 min each, rinsed in water and embedded in gelvatol [36]. For control, sections were incubated only with the secondary antibody or sections obtained from an annexin A7 knockout mouse were used [10]. Samples were examined with a confocal laser scan microscope (Leica TCS SP; Leica, Bensheim, FRG).
Immunofluorescence microscopy of human brain
The paraffin-embedded human brain sections were deparaffinzed likewise. Immunostaining for Annexin A7 was performed using the monoclonal antibody mAb 203–217. Microwave pre-treatment was used for all sections [15]. Trypsin pre-treatment was used as indicated. Carbocyanine 2 and 3 labeled anti-mouse IgG secondary antibodies (Dianova, Hamburg, Germany) were used to detect the primary antibody and showed similar staining patterns. Blank controls were performed, too.
Cell culture and growth conditions
The cell lines used were cultured according to the instructions provided by ATCC: murine neuroblastoma neuro-2a (ATCC CCL-131), rat pheochromocytoma PC-12 (ATCC CRL-1721), and rat glioblastoma C6 (ATCC CCL-107). One day before fixation with 4%paraformaldehyde and permeabilization with 0.5% Triton X-100, cells were seeded on coverslips. Immunostaining was performed as described for the tissue sections with or without the trypsin treatment.
Extraction of nuclear Annexin A7
Cultured cells were trypsinized, counted and washed once in PBS. They were resuspended in buffer1 (0.32 M sucrose, 10 mM Tris, pH8.0, 3 mM CaCl2, 2 mM Mg-acetate, 0.1 mM EDTA, 0.5% NP40, 1 mM DTT, 0.5 mM PMSF; 1 × 107cells/400μl buffer) and sonified on ice (sonifier UP200S, dr.hielscher, 40 s, amplitude 50%, cycle 0.5). This step was repeated until no more intact cells were observed by light microscopy. The resulting homogenate was centrifuged at 500 g for 5 minutes, washed twice in 1 ml buffer 1 without NP40 and centrifuged again. The resulting pellet contains purified nuclei. For partial extraction of the nucleoplasm according to [37], nuclei were resuspended in buffer 2 (20 mM Hepes, pH 7.9, 1.5 mM MgCl2, 20 mM KCl, 0.2 mM EDTA, 25%v/v glycerine, 0.5 mM DTT, 0.5 mM PMSF; nuclei of 107cells/30 μl buffer) and incubated on ice for 30 minutes. An equal volume of buffer 3 was added in aliquots (20 mM Hepes, pH 7.9, 1.5 mM MgCl2, 800 mM KCl, 0.2 mM EDTA, 25%v/v glycerine, 0.5 mM DTT, 0.5 mM PMSF, 1% NP40, protease inhibitor cocktail (Sigma)) and the solution was incubated while shaking for further 30 minutes on ice. Then the solution was centrifuged for 15 minutes at 14.000 g, the resulting supernatant was used as nucleoplasm, and the pellet was washed once and adjusted to an equal volume with PBS. Nucleoplasm and nuclear pellet were subjected to SDS-PAGE and Western blotting.
Western blotting
Cells or cellular fractions were lysed in SDS sample buffer and homogenates were subjected to SDS-PAGE. Western blotting was carried out using the method of Towbin et al. [38]. Peroxidase-coupled secondary antibodies (Sigma) and the supersignal enhanced chemiluminescence system (Pierce, Rockford, IL, USA) were used.
Embryo Multiple Tissue Northern Blot
The presence of Annexin A7 mRNA in mouse ES cells and embryos E7, E11, E15 and E17 was analyzed by northern blot analysis. An Embryo Multiple Tissue Northern Blot membrane was obtained from Clontech Laboratories (Heidelberg, FRG). It contained approximately 2 μg of poly(A)+ RNA per lane from four different stages of mouse development. RNA from ES cells was isolated with the RNeasy kit (Qiagen). The membranes were hybridized with a random-primed [α-32P]-dATP-labelled full length annexin A7 cDNA probe and with a β-actin specific probe for control in ExpressHyb solution according to the manufacturer's directions.
Authors' contributions
MR, SIRG, and CSC planned and carried out the experiments, evaluated the results, and drafted the manuscript. CH provided polyclonal anti-Annexin A7 antibodies and brain sections of the AnxA7 knock out mouse, and contributed to the analysis of data. DT provided the sections the human autopsy brain, analyzed them by immunofluorescence, and contributed to the analysis of data. AAN conceived of the studies and participated in its design. All authors read and approved the manuscript.
Acknowledgements
We thank Maria Stumpf, Rolf Müller and Berthold Gassen for technical assistance, Drs. Andreas Hasse and Rolf Schröder for helpful discussion and Bettina Lauss for help with the manuscript. This work is supported by the Center of Molecular Medicine Cologne.
==== Refs
Liemann S Lewit-Bentley A Annexins: a novel family of calcium- and membrane-binding proteins in search of a function Structure 1995 3 233 237 7788288 10.1016/S0969-2126(01)00152-6
Raynal P Pollard HB Annexins: the problem of assessing the biological role for a gene family Biochem Biophys Acta 1994 1197 63 93 8155692
Creutz CE Pazoles CJ Pollard HB Identification and purification of an adrenal medullary protein (synexin) that causes calcium-dependent aggregation of isolated chromaffin granules J Biol Chem 1978 253 2858 2866 632306
Kuijpers GA Lee G Pollard HB Immunolocalization of synexin (annexin VII) in adrenal chromaffin granules and chromaffin cells: evidence for a dynamic role in the secretory process Cell Tissue Res 1992 269 323 330 1423500 10.1007/BF00319624
Magendzo K Shirvan A Cultraro C Srivastava M Pollard HB Burns AL Alternative splicing of human synexin mRNA in brain, cardiac, and skeletal muscle alters the unique N-terminal domain J Biol Chem 1991 266 3228 3232 1825209
Selbert S Fischer P Pongratz D Stewart M Noegel AA Expression and localization of annexin VII (synexin) in muscle cells J Cell Sci 1995 108 85 95 7738119
Clemen CS Hofmann A Zamparelli C Noegel AA Expression and localisation of annexin VII (synexin) isoforms in differentiating myoblasts J Muscle Res Cell Motil 1999 20 669 679 10672515 10.1023/A:1005524623337
Herr C Clemen CS Lehnert G Kutschkow R Picker SM Gathof BS Zamparelli C Schleicher M Noegel AA Function, expression and localization of annexin A7 in platelets and red blood cells: Insights derived from an annexin A7 mutant mouse BMC Biochemistry 2003 4 8 12925238 10.1186/1471-2091-4-8
Clemen CS Herr C Hövelmeyer N Noegel AA The lack of annexin A7 affects functions of primary astrocytes Exp Cell Res 2003 291 406 414 14644162 10.1016/j.yexcr.2003.07.012
Herr C Smyth N Ullrich S Yun F Sasse P Hescheler J Fleischmann B Lasek K Brixius K Schwinger R Fässler R Schröder R Noegel AA Loss of annexin A7 leads to alterations in frequency-induced shortening of isolated murine cardiomyocytes Mol Cell Biol 2001 21 4119 4128 11390641 10.1128/MCB.21.13.4119-4128.2001
Zhang-Keck ZY Burns AL Pollard HB Mouse synexin (annexin VII) polymorphisms and a phylogenetic comparison with other synexins Biochem J 1993 289 735 741 7916616
Olski TM Noegel AA Korenbaum E Parvin, a 42 kDa focal adhesion protein, related to the α-actinin superfamily J Cell Sci 2001 114 525 538 11171322
Berry M Gottlieb G Development of the cerebral neocortex of the rat Aspects of neurogenesis 1974 2 New York: Academic Press 7 67
Marotte LR Leamey CA Waite PME Timecourse of development of the Wallaby trigeminal pathway: III. Thalamocortical and corticothalamic projections J Comp Neurol 1997 387 194 214 9336223 10.1002/(SICI)1096-9861(19971020)387:2<194::AID-CNE3>3.0.CO;2-W
Hazelbag HM van den Broek LJ van Dorst EB Offerhaus GJ Fleuren GJ Hogendoorn PC Immunostaining of chain-specific keratins on formalin-fixed, paraffin-embedded tissues: a comparison of various antigen retrieval systems using microwave heating and proteolytic pre-treatments J Histochem Cytochem 1995 43 429 437 7534785
Kashima K Yokoyama S Daa T Nakayama I Nickerson PA Noguchi S Cytoplasmic biotin-like activity interferes with immunohistochemical analysis of thyroid lesions: a comparison of antigen retrieval methods Mod Pathol 1997 10 515 519 9160320
Kanai K Nunoya T Shibuya K Nakamura T Tajima M Variations in effectiveness of antigen retrieval pretreatments for diagnostic immunohistochemistry Res Vet Sci 1998 64 57 61 9557807 10.1016/S0034-5288(98)90116-3
Yuasa S Development of astrocytes in the mouse embryonic cerebrum tracked by tenascin-C gene expression Arch Histol Cytol 2001 64 119 126 11310500
Clemen CS Herr C Lie AA Noegel AA Schröder R Annexin VII: an astroglial protein exhibiting a Ca2+-dependent subcellular distribution NeuroReport 2001 12 1139 1144 11338180 10.1097/00001756-200105080-00018
Heizmann CW Braun K Changes in Ca(2+)-binding proteins in human neurodegenerative disorders Trends Neurosci 1992 15 259 264 1381122 10.1016/0166-2236(92)90067-I
Miller RJ Regulation of calcium homoeostasis in neurons: the role of calcium-binding proteins Biochem Soc Trans 1995 23 629 632 8566431
Goda Y Sudhof TC Calcium regulation of neurotransmitter release: reliably unreliable? Curr Opin Cell Biol 1997 9 513 518 9261057 10.1016/S0955-0674(97)80027-0
Verkhratsky AJ Petersen OH Neuronal calcium stores Cell Calcium 1998 24 333 343 10091003 10.1016/S0143-4160(98)90057-4
Garcia ML Strehler EE Plasma membrane calcium ATPases as critical regulators of calcium homeostasis during neuronal cell function Front Biosci 1999 4 D869 D882 10577388
Mattson MP Chan SL Neuronal and glial calcium signaling in Alzheimer's disease Cell Calcium 2003 34 385 397 12909083 10.1016/S0143-4160(03)00128-3
Mamiya N Iino S Mizutani A Kobayashi S Hidaka H Development-related and cell-type specific nuclear localization of annexin XI: immunolocalization analysis in rat tissues Biochem Biophys Res Commun 1994 202 403 409 8037740 10.1006/bbrc.1994.1942
Zaks WJ Creutz CE Ca(2+)-dependent annexin self-association on membrane surfaces Biochemistry 1991 30 9607 9615 1911746 10.1021/bi00104a007
Brownawell AM Creutz CE Calcium-dependent binding of sorcin to the N-terminal domain of synexin (annexin VII) J Biol Chem 1997 272 22182 22190 9268363 10.1074/jbc.272.35.22182
Mizutani A Watanabe N Kitao T Tokumitsu H Hidaka H The long amino-terminal tail domain of annexin XI is necessary for its nuclear localization Arch Biochem Biophys 1995 318 157 165 7726557 10.1006/abbi.1995.1216
Barwise JL Walker JH Subcellular localization of annexin V in human foreskin fibroblasts: nuclear localization depends on growth state FEBS Lett 1996 394 213 216 8843166 10.1016/0014-5793(96)00950-7
Barwise JL Walker JH Annexins II, IV, V and VI relocate in response to rises in intracellular calcium in human foreskin fibroblasts J Cell Sci 1996 109 247 255 8834809
Tomas A Moss SE Calcium- and cell cycle-dependent association of annexin 11 with the nuclear envelope J Biol Chem 2003 278 20210 20216 12601007 10.1074/jbc.M212669200
Eberhard DA Karns LR VandenBerg SR Creutz CE Control of the nuclear-cytoplasmic partitioning of annexin II by a nuclear export signal and by p11 binding J Cell Sci 2001 114 3155 3166 11590242
Jindal HK Chaney WG Anderson CW Davis RG Vishwanatha JK The protein-tyrosine kinase substrate, calpactin I heavy chain (p36), is part of the primer recognition protein complex that interacts with DNA polymerase alpha J Biol Chem 1991 266 5169 5176 1825830
Vannahme C Smyth N Miosge N Gösling S Frie C Paulsson M Maurer P Hartmann U Characterization of SMOC-1, a novel modular calcium-binding protein in basement membranes J Biol Chem 2002 277 37977 37986 12130637 10.1074/jbc.M203830200
Langanger G De Mey J Adam H The sub-cellular localization of annexin V in cultured chick-embryo fibroblasts Mikroskopie 1983 40 237 241 6361609
Adachi O Kawai T Takeda K Matsumoto M Tsutsui H Sakagami M Nakanishi K Akira S Targeted disruption of the MyD88 gene results in loss of IL-1- and IL-18-mediated function Immunity 1998 9 143 150 9697844 10.1016/S1074-7613(00)80596-8
Towbin H Staehelin T Gordon J Electrophoretic transfer of proteins from polyacrylamide gels to nitrocellulose sheets: procedure and some applications Proc Nat Acad Sci USA 1979 76 4350 4354 388439
| 15819996 | PMC1087847 | CC BY | 2021-01-04 16:39:09 | no | BMC Neurosci. 2005 Apr 10; 6:25 | utf-8 | BMC Neurosci | 2,005 | 10.1186/1471-2202-6-25 | oa_comm |
==== Front
BMC NeurosciBMC Neuroscience1471-2202BioMed Central London 1471-2202-6-291584769310.1186/1471-2202-6-29Research ArticleResponse of SI cortex to ipsilateral, contralateral and bilateral flutter stimulation in the cat Tommerdahl Mark [email protected] Stephen B [email protected] Joannellyn S [email protected] Oleg [email protected] Barry [email protected] Department of Biomedical Engineering, University of North Carolina, Chapel Hill, NC 27599, USA2 Department of Cellular and Molecular Physiology, University of North Carolina, Chapel Hill, NC 27599, USA2005 22 4 2005 6 29 29 25 2 2005 22 4 2005 Copyright © 2005 Tommerdahl 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 SII cortex is considered to be the first cortical stage of the pathway that integrates tactile information arising from both sides of the body, SI cortex is generally not considered as a region in which neuronal response is modulated by simultaneous stimulation of bilateral (and mirror-image) skin sites.
Results
Optical intrinsic signal imaging was used to evaluate the response of SI and SII in the same hemisphere to 25 Hz sinusoidal vertical skin displacement stimulation ("skin flutter") applied contralaterally, ipsilaterally, and bilaterally (simultaneously) to the central pads of the forepaws. A localized increase in absorbance in both SI and SII occurred in response to both contralateral and bilateral flutter stimulation. Ipsilateral flutter stimulation evoked a localized increase in absorbance in SII, but little or no change in SI absorbance. In the forepaw representational region of SI, however, bilateral stimulation of the central pads evoked a response substantially smaller (approximately 30–35% smaller) than the response to flutter stimulation of the contralateral central pad.
Conclusion
The finding that the response of SI cortex to bilateral central pad flutter stimulation is substantially smaller than the response evoked by a contralateral flutter stimulus, together with the recently published observation that a region located posteriorly in SII responds with a substantially larger response to a bilateral flutter stimulus than the response evoked from the contralateral central pad, lead us to propose that the SI activity evoked by contralateral skin stimulation is suppressed/inhibited (via corticocortical connections between SII and SI in the same hemisphere) by the activity a simultaneous ipsilateral skin stimulus evokes in posterior SII.
==== Body
Background
It is established that multiple fields/areas in each cerebral hemisphere are activated at short latency by stimuli that trigger spike discharge activity in skin mechanoreceptive afferents. Of these the most extensively studied is SI – a region which most investigators have come to regard as responsive solely (the major exception being the face region of SI) to tactile stimuli delivered to contralateral skin sites. In contrast, SII has long been known to be activated at short latency by mechanical stimulation of skin sites on both sides of the body midline. Although this differential responsivity of SI and SII to stimulation of contralateral vs. ipsilateral skin sites has been widely accepted for more than five decades (since the pioneering evoked potential mapping studies of C.N. Woolsey and colleagues; e.g. [1]), the contribution of SII activation to cortical tactile information processing remains unknown. Additionally, although SII is considered to be the first cortical stage of the pathway that integrates information arising from both sides of the body, SI is generally not considered as a cortical region in which ipsilateral inputs play a major role in bilateral integration of information across the body mid-line. It is known, however, that even the distal limb representational regions in SI receive axonal projections (via the corpus callosum) from SI neurons in the opposite hemisphere [2,3].
Recently, we reported the results from experiments in which we obtained simultaneous observations of the activity evoked in both SI and SII in the same hemisphere of cat cerebral cortex by a 25 Hz sinusoidal vertical skin displacement stimulus ("skin flutter") applied contralaterally, ipsilaterally, or bilaterally to the central pads of the forepaws [4]. Briefly summarized, a localized increase in absorbance in both SI and SII was evoked by contralateral and also by bilateral flutter stimulation. Ipsilateral flutter stimulation also evoked a localized increase in absorbance in SII, but only a weak or negligible increase in the forepaw region of SI. Interestingly, the region of SII that responded with an increase in absorbance to ipsilateral stimulation was 2–3 mm posterior to the region in which absorbance increased maximally in response to stimulation of the contralateral central pad. Furthermore, in the posterior SII region that yielded the largest response to ipsilateral stimulation of the central pad, the response to bilateral central pad stimulation approximated a linear summation of the SII responses to independent stimulation of the contralateral and ipsilateral central pads. Conversely, in the anterior region of SII (the region that exhibited the largest response to contralateral stimulation), the response to bilateral stimulation was consistently smaller than (by approximately 30–35%) the response evoked from the contralateral central pad.
This report addresses the response of SI cortex to the same modes of stimulation used in the above-described study that focused on SII (i.e., contralateral, ipsilateral, and bilateral stimulation of the central pads of the forepaws). The central finding is that flutter stimulation of the ipsilateral central pad exerts a suppressive/inhibitory influence on the SI response to a 25 Hz flutter stimulus to the contralateral central pad (forepaw). In addition, the temporal relationship between stimulus-evoked activity in selected locations in the responding regions of SI and SII is evaluated quantitatively (using the approach of correlation mapping), revealing a previously unrecognized, and presumably functionally important high degree of coordination between the activities evoked by forepaw stimulation in both SI and the recently identified [4] anterior vs. posterior components that comprise SII cortex in the same hemisphere.
Results
Figure 1 shows the optical (OIS) responses evoked in SI and SII in one exemplary subject by contralateral, ipsilateral, and bilateral central pad stimulation. Visual inspection of the images for the three stimulus conditions shows that: (1) the SI response to bilateral stimulation occupies the same region that responded to stimulation of the contralateral central pad; and (2) stimulation of the ipsilateral central pad evokes little or no absorbance change in SI. Results such as those shown in Figure 1 were consistent across all subjects studied (n = 5) and have been reported previously [4]. In all subjects, the area of SI that responded with an increase in absorbance was confined to areas 3b/1. The region of area 2 that corresponds to the central pad of the forepaw is normally buried in the anterior bank of the ansate sulcus and could not be imaged. Inadvertently, the position of the cortical recording chamber, and thus the region of pericoronal cortex that was imaged, varied from one subject to the next. In Figure 1, for example, a region anterior to SI (bottom left corner; identified cytoarchitectually as area 4γ) was included in the imaged field and interestingly, this region not only responded vigorously to contralateral stimulation, but the magnitude of the absorbance increase in this area was largest with bilateral stimulation. Unfortunately, area 4γ was not imaged in any of the other subjects.
The SI response to each of the different stimulus conditions can be better appreciated when the imaging data are displayed as a multi-dimensional surface plot. Figure 2 compares the stimulus evoked response of SI to contralateral vs. bilateral stimulation at 5 seconds after stimulus onset (stimulus duration always was 5 seconds). The top left panel indicates the orientation of the sampled region of interest (ROI) in SI. The surface plots show the absorbance value at each pixel location in the imaged field, with absorbance defined both by elevation along the z-axis and by pseudocolor (indicated by the color-coded scale bar at right). Consistent with the difference images shown in Figure 1, the response evoked by ipsilateral stimulation is near-background (or nonexistent), whereas the responses to contralateral and bilateral stimulation are localized and clearly evident. Close visual inspection of these surface plots also reveals that the magnitude of the SI response evoked by contralateral stimulation is greater than the response evoked by bilateral stimulation.
Figure 3 enables one to directly compare the temporal sequence of the SI response to contralateral, ipsilateral, and bilateral central pad flutter. Each point in the plots in Figure 3 shows the average absorbance values within the ROI centered over SI at a particular time (frame) after stimulus onset. For both of the subjects that provided the data shown in Figure 3, the temporal sequences (one for each mode of stimulation) demonstrate that: (1) ipsilateral stimulation evokes little or no change in SI absorbance, (2) the magnitude of the absorbance change evoked in SI by bilateral stimulation is smaller than that evoked in the same SI region by contralateral stimulation, and (3) the discrepancy between the responses to bilateral vs. contralateral stimulation appears to increase with increasing time after onset of stimulation.
Scatter plots were used to directly compare the response of SI to the different conditions of contralateral and bilateral stimulation. In each plot in Figure 4 the absorbance value obtained at each pixel under these two stimulus conditions is plotted against each other – i.e., the x-axis is the absorbance value evoked by the contralateral stimulus and the y-axis is the absorbance value evoked by the bilateral stimulus. The plots reveal that there is a difference in the population response of SI neurons to the contralateral vs. bilateral stimulus conditions. Additionally, the plots suggest that there is a time dependency in the development of the response to bilateral stimulation – that is, there is little difference between the responses to the two conditions at 1 sec after stimulus onset (at t = 1 sec the pixels are located symmetrically about the 45° line), yet at a later time (at t = 5 in each subject many of the pixels are shifted to a position below the reference line). Points located below (or to the right of) the reference line (slope = 1) represent pixels, or spatial locations in SI, where the response to contralateral stimulation is greater than the response to bilateral stimulation. It should be emphasized that this type of graphic does not reflect spatial differences in the responses to the different stimulus conditions, but rather, emphasizes whether or not different members of a set respond differently to the different stimulus conditions. Accordingly, the plots in Figure 4 indicate that with increasing time after onset of bilateral stimulation of the central pads (i.e., between 1 and 5 seconds after stimulus onset), absorbance at particular locations within the responding region of SI pixels reduces relative to the value achieved during contralateral stimulation. It also should be noted that the plots in Figure 4 show that not all spatial locations (pixels) in the responding region of SI exhibit this reduction in absorbance – that is, each plot indicates that at a minority of locations within the responding SI region bilateral stimulation evokes the same or even a greater response than the response evoked by contralateral stimulation.
The across-subject consistency of the results was evaluated by determining the average absorbance values evoked by the 3 different stimulus conditions (with a 5 sec stimulus duration) for the 5 subjects at 5 sec after stimulus onset (Figure 5). Clearly, the contralateral stimulus condition evoked the largest average (across-subject) response, and therefore the values of the average (across-subject) absorbance increase obtained under the ipsilateral and the bilateral stimulus conditions were normalized to this absorbance value (thus standard error for the contralateral stimulus condition = 0). While the average response to ipsilateral stimulation was very weak or negligible, the average response evoked by bilateral stimulation was approximately 35% less than that evoked by contralateral stimulation. In each of the 5 subjects the response to bilateral stimulation was less than the response to contralateral stimulation. Statistical hypothesis testing was performed in order to determine if the average (across-subject) response evoked by bilateral stimulation was significantly less than the average response to contralateral stimulation (i.e., if the bilateral/contralateral response ratio was <1 at 5 sec after stimulus onset). Analysis of variance showed that the average bilateral/contralateral response ratio was between 0.51 and 0.92 in SI (p < 0.001; Ho was that the ratio = 1).
Comparisons between SI and SII responses
Our previous study [4] characterized the SII response to the same contralateral, ipsilateral and bilateral stimulus conditions that we used to characterize the response of SI cortex in this report. The SII responses evoked by these different stimulus conditions led to the identification of two regions in SII arranged along the anterior and posterior axis of the anterior ectosylvian gyrus. In that study, identical contralateral vs. bilateral stimulus conditions at homotopic skin sites demonstrated that the response evoked by a bilateral stimulus was 35% smaller than that evoked by a contralateral stimulus in the anterior region of SII [4]. This correspondence between the effects of bilateral stimulation on SI and the posterior region of SII in the same hemisphere led us to evaluate the extent to which the changes in the activity of posterior SII and SI observed under this condition are inter-related. To investigate the relationship between the spatiotemporal patterns of absorbance between the different cortical regions – SI, anterior SII, and posterior SII – correlation analysis was performed on the data. Figure 6 shows the results obtained when this approach is used to evaluate the relationship between the responses to contralateral and ipsilateral stimulus conditions for an exemplary subject. For both correlation maps, a correlation value is obtained for each location (pixel) in the responding SI region by cross-correlating the time course of the absorbance changes at that location with a "standard" time course. In the case of Figure 6, the "standard" time course that was correlated with each pixel was obtained from SI activity evoked by contralateral stimulation. The red line in the graph inset of Figure 6 is the standard time course obtained from the average response evoked within the indicated SI boxel. The blue plot in the graph inset is simply the negative of the red and indicates the time course of pixels that are negatively correlated with the standard time course. In the contralateral stimulus condition, note that the activity in each of the 3 zones within the responding region (indicated by boxels at top left of Figure 6) is positively correlated to the time course of activity evoked in SI by a contralateral stimulus. For the ipsilateral stimulus condition, however, only the activity in the region of SII is positively correlated with the activity in the responding region of SI, whereas the activity in SI and the anterior SII regions is negatively correlated. The negative correlation indicates that the time course of stimulus-evoked activity in SI and anterior SII is trending in a direction opposite to that observed under the contralateral stimulus condition, whereas in posterior SII the activity modifies in a manner similar to that observed in the responding region of SI. The correlation maps obtained from each of the other experimental subjects indicated similar trends.
A second method of correlation mapping was used to directly compare the responses evoked by contralateral vs. bilateral stimulation (a "cross-modal" correlation map). In this method, rather than correlating each pixel with a standard time course observed at one particular cortical region for the same stimulus condition, performs a cross correlates the time courses of absorbance values obtained at the same spatial location under different stimulus conditions. Thus, the correlation coefficient generated at each pixel location represents how similar (indicated by a positive correlation) or how different (indicated by a negative correlation) are the temporal responses of a cortical region recorded under 2 different stimulus conditions. The map in Figure 7 is indicative of the type result obtained by performing a "cross-modal" correlation (correlation of responses evoked by 2 different modes of stimulation) in all the subjects in which SI and SII was simultaneously imaged. Note that this type of correlation map shows that the responding region of SI and anterior SII (as defined by the boxels in the figure) have high values of correlation, whereas posterior SII shows only weak, or in some local regions, negative correlation. It should be pointed out that correlation values are an indication of how similar are the trends – in particular, the slopes of the time courses – of the activity of a specific region under the different modes of stimulation. Thus, although the activity levels of SI and anterior SII both undergo a reduction in magnitude (relative to the activity detected during contralateral stimulation) under the bilateral stimulus condition, the slopes of the time course are very similar – accordingly, the activity of the 2 regions is positively correlated under this condition. In posterior SII, on the other hand, response magnitude increases under the bilateral stimulus condition (relative to that obtained under the contralateral stimulus condition) – thus with bilateral stimulation the activity of this region is negatively correlated with the activity in SI and in anterior SII.
Discussion
The findings of this study demonstrate that bilateral flutter stimulation of the central pads of the forepaws evokes an SI response significantly smaller than the response evoked when the stimulus is applied contralaterally. At the locus of the maximal OIS response evoked in the SI region by a contralateral stimulus, bilateral stimulation evoked a response that was, on average, 35% smaller than that evoked by the contralateral stimulus. Concurrently, at the locus of the maximal OIS response evoked by contralateral stimulation in anterior SII (reported in a previous study; [4]), bilateral stimulation evoked a response that was, on average, 35% lower than the activity evoked by a contralateral stimulus. The same conditions that led to the decrease in activity in SI and anterior SII also led to an increase in activity in posterior SII [4].
Although the neural mechanisms responsible for the above-described effect remains uncertain, some workers have reported that callosally-transmitted inputs exert modulatory, but inconsistent effects on SI neurons. Schnitzler et al [5], using MEG in humans, showed that tactile stimulation of one hand enhanced the response of ipsilateral primary somatosensory cortex (SI) to median nerve stimulation. Conversely, Korvenoja et al [6] reported that activation of SI (measured using MEG in humans) by electrical stimulation of the contralateral median nerve was suppressed during movement of the fingers of the ipsilateral hand. In addition, Hoechstetter et al. [7] described "interactions" in SII cortex (defined as a response that was less than a summation of the responses to independent ipsilateral and contralateral stimulation) during bilateral stimulation, but reported that the response in SI evoked by a contralateral stimulus was not altered by the addition of an ipsilateral stimulus. Shimojo et al [8] also found no difference in sources localized to SI evoked by unilateral versus bilateral stimulation. Most recently, Staines, et al, [9] reported data (fMRI) showing that SI activity evoked by passive bilateral stimulation is weaker than the SI activity evoked by passive unilateral stimulation. Other findings in the same report showed that the modulation of SI activity, particularly in the case of bilateral stimulation, was task-dependent. Others have shown that the activity evoked in SI by a unilateral stimulus can be modulated in a context-dependent manner, both in humans (e.g., [10,11]) and in other primates (e.g., [12-15]).
Although an influence mediated via direct callosal projections from ipsilateral SI to contralateral SI [2] should not be ruled out, the data presented in this study lead us to propose that SII is the major source of the modulated SI response to contralateral skin stimulation that is observed under a variety of stimulus conditions [16-19]. Whereas the available literature is contradictory [20], it is clear that SII in cats receives its principal input from the thalamus [21-25]. While some investigators have proposed that monkey SII occupies a higher position in the somatosensory information processing hierarchy than does SII in cat [20], others have warned that "these are basically descriptive schemes of connections that do not illuminate what features of somesthesis are selectively projected" and that "they also tend to ignore potential interdependence between SI and SII" [26]. The OIS observations in the present study obtained by simultaneously imaging the contralateral SI and SII in cats during ipsilateral stimulation of the central pad revealed that the time course of the optical response (increasing absorbance) evoked in posterior SII is negatively correlated with the time course of the optical signal in SI and anterior SII. Perhaps more significantly, the results obtained by correlation of the responses evoked under contralateral and bilateral stimulus conditions revealed a similar relationship between the cortical activity evoked in SI, anterior SII and posterior SII – in this case, the time course of the activities of SI and anterior SII are positively correlated under the bilateral stimulus condition, whereas under the same condition those activities are negatively or only weakly correlated with the activity in posterior SII. A plausible (but not necessarily the only) explanation for these outcomes is that the activity evoked in posterior SII exerts an inhibitory influence on SI via the extensive corticocortical connections known [27,28] to link topographically corresponding regions in SI and SII.
An extensive literature addresses the neuroanatomical routes by which information about the status of skin mechanoreceptors accesses SI and SII (see [27,28] for review). While the data presented in this paper neither extend or modify what is known about the routes by which information reaches SI and SII, they (along with the evidence presented in our previous paper – [4]) address the issue of whether, and to what extent, the responses of SI, anterior SII, and posterior SII in the same hemisphere are independent. The data presented here, in conjunction with evidence published previously [4], shows that nonnoxious mechanical skin stimulation evokes afferent activity that is conveyed via central somatosensory pathways to SII in both the contralateral and the ipsilateral hemispheres, and raises the possibility that unlike the activity evoked by contralateral flutter, the SII activity generated by ipsilateral skin stimulation may depress (perhaps via inhibitory processes mediated by corticocortical connections) the response of SI to ongoing skin stimulation. Previously, we postulated that SII activity levels modulated the SI response when the SII activity was evoked by vibratory stimuli [18]. That study showed that contralateral vibration increased SII activity levels which were correlated with decreases in SI vibration-evoked activity. The findings make it seem likely that, at least in cat, the degree to which SII activation modifies the SI response to skin stimulation depends on the attributes of the peripheral stimulus – that is, that SI activity is modified by SII activity in a stimulus-dependent manner.
Conclusion
The responses evoked by contralateral, ipsilateral and bilateral flutter stimulation of the central pad of the cat forepaw indicate that the response of SI evoked by contralateral stimulation is reduced in the presence of an ipsilateral stimulus. Correlation analysis of the responses evoked by ipsilateral skin flutter stimulation showed that activity in posterior SII (the region of SII that maximally responds to an ipsilateral stimulus) is negatively correlated with the activity in SI. Additionally, correlation mapping of the results obtained from the contralateral vs. bilateral stimulus conditions demonstrated that the time course of the activities evoked in SI and anterior SII by these conditions are similar. The data lead to the proposal that increasing levels of activity in posterior SII evoked by an ipsilateral skin stimulus suppress/inhibit the responses normally evoked in both anterior SII and SI by contralateral stimulation.
Methods
Subjects & preparation
Adult cats (males and females; n = 5) were subjects. All surgical procedures were carried out under deep general anesthesia (1 – 4% halothane in a 50/50 mixture of oxygen and nitrous oxide). After induction of general anesthesia the trachea was intubated with a soft tube and a polyethylene cannula was inserted in the femoral vein to allow administration of drugs and fluids (5% dextrose and 0.9% NaCl). For each subject, a 1.5 cm diameter opening was made in the skull overlying somatosensory cortex, a chamber was mounted to the skull over the opening with dental acrylic, and the dura overlying anterior parietal cortex was incised and removed. Following the completion of the surgical procedures all wound margins were infiltrated with long-lasting local anesthetic, the skin and muscle incisions were closed with sutures, and each surgical site outside the recording chamber was covered with a bandage held in place by adhesive tape.
Subjects were immobilized with Norcuron and ventilated with a gas mixture (a 50/50 mix of oxygen and nitrous oxide; supplemented with 0.1 – 1.0% halothane when necessary) delivered via a positive pressure respirator 1–3 hours prior to the data acquisition phase of the OIS imaging experiments. Respirator rate and volume were adjusted to maintain end-tidal CO2 between 3.0 – 4.0%; EEG and autonomic signs (slow wave content; heart rate, etc.) were monitored and titrated (by adjustments in the anesthetic gas mixture) to maintain levels consistent with light general anesthesia. Rectal temperature was maintained (using a heating pad) at 37.5°C.
Euthanasia was achieved by intravenous injection of pentobarbital (45 mg/kg) and by intracardial perfusion with saline followed by fixative (10% formalin). Following perfusion fiducial marks were placed to guide removal, blocking, and subsequent histological sectioning of the cortical region studied. All procedures were reviewed and approved in advance by an institutional committee and are in full compliance with current NIH policy on animal welfare.
Stimuli and stimulus protocols
Results were obtained during stimulation of the contralateral central pad of the forepaw and/or the ipsilateral central pad of the forepaw. The stimuli always consisted of sinusoidal vertical skin displacements (25 Hz, 400 microns, stimulus duration 5 sec, inter-stimulus interval 60 sec) and were applied using a pair of servocontrolled transducers (Cantek Enterprises, Canonsburg, PA) that is capable of delivering sinusoidal stimuli in the range of 1–250 Hz at amplitudes in the range of 0–1000 microns. The stimuli were delivered independently to the ipsilateral and contralateral skin sites, and also were applied simultaneously to both sites (bilateral stimulation). The stimulus probes were positioned 500 microns beyond the point at which skin contact was detected (via force transducer on the Cantek). The bilateral stimulus protocols reported in this paper were synchronized to start and stop at the same time. The contralateral, ipsilateral and bilateral stimuli were interleaved on a trial-by-trial basis. This approach was used to control for temporal changes in cortical "state" unrelated to stimulus conditions which, if unrecognized, might obscure or modify any differences between the optical responses evoked by the contralateral, ipsilateral and bilateral stimulus conditions.
OIS imaging
Near-infrared (IR; 833 nm) OIS imaging was carried out using an oil-filled chamber capped with an optical window [29] Images of the exposed cortical surface were acquired 200 msec before stimulus onset ("reference" or "prestimulus" images) and continuously thereafter ("poststimulus" images; at a resolution of one image every 0.5 to 1.5 sec) for 15–20 sec following stimulus onset. Exposure time was 200 msec. Absorbance images were generated by subtracting each prestimulus (reference) image from its corresponding poststimulus image and subsequently dividing by the reference image. Averaged absorbance images typically show regions of both increased absorption of IR light and decreased absorption of light (to a depth of approximately 1400 microns) which have been shown to be accompanied by increases and decreases in neuronal activation, respectively [30-35].
Correlation analysis
Correlation maps were constructed for comparison of spatio-temporal characteristics of the OIS response. One aspect of this method of analysis has been previously described in detail [18,29]. Briefly, the correlation maps in Figure 6 were constructed by choosing a reference region within the imaged field and computing the intensity correlation rij between the absorbance value of each pixel (i, j) and the average absorbance value within the reference region over the time from stimulus onset to stimulus offset. The region selected as the reference was defined by a boxel (π mm2 area) centered on the region of interest (ROI). Each pixel (i, j) on the correlation map is represented by a coefficient of determination r2ij (-1 <r2 < 1; - 1 indicates negative correlation; + 1 indicates positive. The statistical significance of each of the correlations was tested with the standard t-test. A second type of correlation map, or cross-correlation map, was generated in a similar manner (such as that computed in Figure 7). In this method, rather than correlating each pixel with the time course observed at one particular cortical region for the same stimulus condition, cross correlation was performed between the time courses of absorbance values obtained at the same spatial location (i, j) for different stimulus conditions. Thus, each coefficient of determination r2ij represents how similar (positive correlation) or how different (negative correlation) the response in a region is to a change in stimulus condition.
Histological procedures/identification of cytoarchitectural boundaries
At the conclusion of the experiment, the imaged cortical region was removed immediately following intracardial perfusion with saline and fixative. The region then was blocked, postfixed, cryoprotected, frozen, sectioned serially at 30 μm, and the sections stained with cresyl fast violet. The boundaries between adjacent cytoarchitectonic areas were identified by scanning individual sagittal sections separated by no more than 300 mm and were plotted at high resolution using a microscope with a drawing tube attachment. The resulting plots then were used to reconstruct a two-dimensional surface map of the cytoarchitectonic boundaries within the region studied with optical and neurophysiological recording methods. The locations of microelectrode tracks and electrolytic lesions evident in the histological sections were projected radially to the pial surface and transferred to the map of cytoarchitectonic boundaries reconstructed from the same sections. As the final step, the cytoarchitectonic boundaries (along with the locations of microelectrode tracks and lesions whenever present) identified in each brain were mapped onto the images of the stimulus-evoked intrinsic signal obtained from the same subject, using fiducial points (made by postmortem applications of india ink or needle stabs) as well as morphological landmarks (e.g., blood vessels and sulci evident both in the optical images and in histological sections). Locations of cytoarchitectonic boundaries were identified using established criteria [36-38].
Authors' contributions
BW and OF participated in the design of the experiments, the data collection, and drafting of the manuscript. SS and JC made significant contributions to the data collection and the analysis of the data. MT played a major role in all aspects of the development of the manuscript.
Acknowledgements
This work was supported, in part, by NIH NS050587 (M. Tommerdahl, P.I.) and NIH NS35222 (B. Whitsel, P.I.).
Figures and Tables
Figure 1 Cat SI and SII optical responses to 25 Hz vibrotactile stimulation of the forepaws. A. View of the cortical surface, showing the vascular pattern and coronal (COR), ansate (ANS), and suprasylvian (SS) sulci. Exposed portions of SI and SII are indicated. Below: Averaged absorbance images for responses evoked by (B) contralateral, (C) ipsilateral and (D) bilateral stimuli. Individual absorbance images were generated by subtracting each prestimulus (reference) image from its corresponding poststimulus image and subsequently dividing by the reference image. Averages are generated by summating the data obtained at a particular frame across trials (in this case, at 5 sec after stimulus onset). Stimulus sites are indicated by figurines. Scale bar is 2 mm. Orientation of images indicated by P (posterior), A (anterior), M (medial) and L (lateral) axes. Components of this figure have been previously reported [4].
Figure 2 Surface plots of absorbance evoked in SI by contralateral, ipsilateral, and bilateral stimulation. Data displayed is a subset of the data displayed in Figure 1. Region of interest is indicated by the dashed box shown in the OIS image (top left). Orientation of the selected region is indicated by X (medial-lateral) and Y (anterior-posterior) labels on the dashed box and axes (in ipsilateral map). Cortical space along X and Y axes is measured in mm. Z axis is absorbance which is represented by both the height of the graphic as well as the color indicated by the color bar to the right. Surface plots represent absorbance values within the ROI at 5 seconds after stimulus onset and demonstrate that 1)response in SI to ipsilateral stimulation is small or nonexistent and, 2)magnitude of evoked absorbance in response to contralateral stimulation is slightly larger than the response evoked by bilateral stimulation.
Figure 3 Graphs obtained from OIS data in SI cortical regions evoked by flutter stimulus on the central pad of 2 subjects. Top: Dashed box on OIS images indicates dimensions and orientation of region of interest (ROI) in SI used for analysis. Orientation and scales are identical in both subjects. Stimulus duration was 5 seconds; stimulus onset was at time 0 s. Bottom: Absorbance values within the ROI were averaged and plotted as a function of time for each stimulus condition. In both subjects contralateral stimulation evokes the largest change in absorbance, while ipsilateral stimulation evokes only a weak change in absorbance. Bilateral stimulation evokes an absorbance change that is less than that evoked by a contralateral stimulus.
Figure 4 Scatter plots of contralateral vs. bilateral response of 2 subjects. For each plot, values of individual pixels are plotted as a function of the absorbance measured at that pixel evoked by the contralateral stimulus (horizontal axis) vs. the response measured at the same locus (or pixel) evoked by the bilateral stimulus (vertical axis). Top panel: cortical images taken with green filter, superimposed axes indicate anatomical orientation. Pixels superimposed on cortical image were selected using a threshold criteria (highest 5%) in response to both stimulus conditions and are represented by blue dots. Reference line, plotted at a slope of 1, indicates where pixels with equal values for both conditions lie.
Figure 5 Comparison of averaged normalized absorbance values (n = 5) evoked under 3 different stimulus conditions. Values were normalized to those obtained from the contralateral stimulus condition (thus, standard error bar for the contralateral condition is zero) and were obtained at 5 sec after stimulus onset of a 5 sec flutter stimulus. The contralateral stimulus evoked the greatest change in absorbance, while the ipsilateral stimulus evoked the weakest. The bilateral stimulus evoked a cortical response that was less than the response evoked by the contralateral stimulus. Analysis of variance showed that, at a 99% confidence interval, the ratio of the bilateral response evoked to the contralateral response evoked in SI was between 0.51 and 0.92.
Figure 6 Correlation maps for contralateral and ipsilateral stimulus conditions in Subject #2. Top Left: Orientation of the maps. SI boxel indicates region of interest (ROI) used for obtaining averaged absorbance values. Bottom Left: Red line is graph of averaged time course of SI activity (from the ROI) used for generating correlation maps (from 1 to 8 seconds after stimulus onset). Blue line indicates the approximated time course of SI activity that would be negatively correlated. A correlation of +1 would correspond to the signal shown in dark red while a correlation of -1 would correspond to the signal in dark blue. Top Right: The correlation map generated from the contralateral response data shows positively correlated activity in SI as well as both regions of SII. Bottom Right: The correlation map generated from the ipsilateral response shows a positive correlation in only the posterior region of SII and a negative correlation in both the anterior region of SII and SI. Color bars indicate correlation coefficient values for each map.
Figure 7 Cross-modal correlation map of Subject #2. Panel at top-left indicates orientation of the correlation map at right. The color bar shows correlation coefficient values used in the map. Correlation coefficients are determined by cross-correlation of time courses evoked at each spatial location to two different stimulus conditions. The map demonstrates that the responses to contralateral and bilateral stimulation are similar in SI and anterior SII, but that the different stimulus conditions evoke a weak or negative correlation in the posterior region of SII. Correlations were performed on data between 1 and 8 sec after stimulus onset.
==== Refs
Woolsey CN Fairman D Contralateral, ipsilateral and bilateral representation of cutaneous receptors in somatic areas I and II of cerebral cortex of pig, sheep and other animals Surgery 1946 19 684 702
Caminiti R Innocenti G Manzoni T The anatomical substrate of callosal messages from SI and SII in the cat Exp Brain Res 1979 35 295 314 86455 10.1007/BF00236617
Innocenti GM Manzoni T Spidalieri Peripheral and transcallosal reactivity of neurons within SI and SII cortical areas. Segmental Divisions Arch Ital Biol 1972 110 415 443
Tommerdahl M Simons SB Chiu JS Tannan V Favorov OV Whitsel BL Response of SII cortex to ipsilateral, contralateral and bilateral flutter stimulation in the cat BMC Neuroscience 2005 6 11 15710047 10.1186/1471-2202-6-11
Schnitzler A Salmelin R Salenius S Jousmaki V Hari R Tactile information from the human hand reaches the ipsilateral primary somatosensory cortex Neuroscience Letters 1995 200 25 28 8584258 10.1016/0304-3940(95)12065-C
Korvenoja A Wikstrom H Huttunen J Virtanan J Laine P Aronen HJ Seppalainen AM Ilmoniemi RJ Activation of ipsilateral primary sensorimotor cortex by median nerve stimulation Neuroreport 1995 6 2589 2593 8741769
Hoechstetter K Meinck H Henningsen P Scherg M Rupp A Psychogenic sensory loss: magnetic source imaging reveals normal tactile evoked activity of the human primary and secondary somatosensory cortex Neurosci Letters 323 137 140 2002, Apr 26
Shimojo M Kakigi R Hoshiyama M Koyama S Kitamura Y Watanabe S Intracerebral interactions caused by bilateral median nerve stimulation in man: a magnetoencephalographic study Neuroscience Research 1996 24 175 181 8929924 10.1016/0168-0102(95)00994-9
Staines WR Graham SJ Black SE McIlroy WE Task-relevant modulation of contralateral and ipsilateral primary somatosensory cortex and the role of a prefrontal-cortical sensory gating system NeuroImage 2002 15 190 199 11771988 10.1006/nimg.2001.0953
Knecht S Kunesch E Buchner H Freund H-J Facilitation of somatosensory evoked potentials by exploratory finger movements Exp Brain Res 1993 95 330 338 8224057 10.1007/BF00229790
Burton H Abend NS MacLeod A-MK Sinclair RJ Snyder AZ Raichle ME Tactile attention tasks enhance activation in somatosensory regions of parietal cortex: A positron emission tomography study Cerebral Cortex 1999 9 662 674 10554989 10.1093/cercor/9.7.662
Ro JY Debowy D Ghosh S Gardner EP Depression of neuronal firing rates in somatosensory and posterior parietal cortex during object acquisition in a prehension task Exp Brain Res 2000 135 1 11 11104122 10.1007/s002210000496
Gardner EP Ro JY Debowy D Ghosh S Facilitation of neuronal activity in somatosensory and posterior parietal cortex during prehension Exp Brain Res 1999 127 329 354 10480270 10.1007/s002210050803
Chapman CE Active versus passive touch: Factors influencing the transmission of somatosensory signals to primary somatosensory cortex Can J Physiol Pharmacol 1994 72 558 570 7954086
Meftah el-M Shenasa J Chapman CE Effects of a cross-modal manipulation of attention on somatosensory cortical neuronal responses to tactile stimuli in the monkey Journal of Neurophysiology 2002 88 3133 3149 12466436
Turman AB Morley JW Zhang HQ Rowe MJ Parallel processing of tactile information in cat cerebral cortex: effect of reversible inactivation of SII on SI responses Journal of Neurophysiology 1995 73 1063 1075 7608755
Rowe MJ Turman AB Murray GM Zhang HQ Parallel organization of somatosensory cortical areas I and II for tactile processing Clin Exp Pharmacol Physiol 1996 23 931 938 Review 8911737
Tommerdahl M Whitsel B Favorov O Metz C BL O'Quinn Responses of contralateral SI and SII in cat to same site cutaneous flutter versus vibration Journal of Neurophysiology 1999 82 1982 1992 10515988
Zhang HQ Murray GM Coleman GT Turman AB Zhang SP Rowe MJ Functional characteristics of the parallel SI- and SII-projecting neurons of the thalamic ventral posterior nucleus in the marmoset Journal of Neurophysiology 2001 85 1805 1822 11352998
Pons TP Garraghty PE Friedman DP Mishkin M Physiological evidence for serial processing in somatosensory cortex Science 1987 237 417 420 3603028
Alloway KD Sinclair RJ Burton H Responses of neurons in somatosensory cortical area II of cats to high-frequency vibratory stimuli during iontophoresis of a GABA antagonist and glutamate Somatosensory Motor Research 1988 6 109 140 2907395
Bennett R Ferrington D Rowe M Tactile neuron classes within second somatosensory area (SII) of cat cerebral cortex Journal of Neurophysiology 1980 43 292 309 7381522
Ferrington DG Rowe M Differential contributions to coding of cutaneous vibratory information by cortical somatosensory area I and II J Neurophysiol 1980 43 310 331 7381523
Fisher GR Freeman B Rowe MJ Organization of parallel projections from Pacinian afferent fibers to somatosensory cortical areas I and II in the cat Journal of Neurophysiology 1983 49 75 97 6827305
Herron P Dykes R The ventroposterior inferior nucleus in the thalamus of cats: a relay nucleus in the Pacinian pathway to somatosensory cortex Journal of Neurophysiology 1986 56 1475 1497 3806180
Burton H Sinclair RJ Second somatosensory cortical area in macaque monkeys: 2. Neuronal responses to punctate vibrotactile stimulation of glabrous skin on the hand Brain Research 1991 538 127 135 2018924 10.1016/0006-8993(91)90386-A
Burton H Fabri M Ipsilateral intracortical connections of physiologically defined cutaneous representations in areas 3b and 1 of macaque monkeys: projections in the vicinity of the central sulcus Journal of Computational Neurology 1995 355 508 538 10.1002/cne.903550404
Manzoni T Barbaresi P Bellardinelli E Caminiti R Callosal projections from the two body midlines Exp Brain Res 1980 39 1 9 7379877 10.1007/BF00237063
Tommerdahl M Delemos KA Favorov OV Metz CB Whitsel BL Response of anterior parietal cortex to different modes of same-site skin stimulation Journal of Neurophysiology 1998 80 3272 3283 9862921
Grinvald A Real-time optical mapping of neuronal activity: from single growth cones to the intact mammalian brain Annual Review of Neuroscience 1985 8 263 305 3885828 10.1146/annurev.ne.08.030185.001403
Grinvald A Bonhoeffer T Malonek D Shoham D Bartfeld E Arierli A Hildesheim R Ratzlaff E Squire L, Weinberger N, Lynch G, McGaugh J Optical imaging of architecture and function in the living brain Memory Organization and Locus of Change 1991 N.Y.: Oxford Univ Press 49 85
Grinvald A Lieke E Frostig R Hildesheim R Cortical point-spread function and long-range lateral interactions revealed by real-time optical imaging of macaque monkey primary visual cortex Journal of Neuroscience 1994 14 2545 2568 8182427
Tommerdahl M Whitsel B Franzen O, Johansson R, Terenius L Optical imaging of intrinsic signals in somatosensory cortex Somesthesis and the Neurobiology of Somatosensory Cortex 1996 Basel: Birkhauser Verlag AB 369 384
Tommerdahl M Delemos KA Whitsel BL Favorov OV Metz CB Response of anterior parietal cortex to cutaneous flutter versus vibration Journal of Neurophysiology 1999 82 16 33 10400931
Ebner T Chen G Use of voltage-sensitive dyes and optical recordings in the central nervous system Progress in Neurobiology 1995 46 463 506 8532849 10.1016/0301-0082(95)00010-S
Hassler R Muhs-Clement K Architektonischer Aufbau des sensomotorischen und parietalen Cortex der Katze Journal fur Hirnforschung 1964 6 377 420
McKenna T Whitsel B Dreyer D Metz C Organization of cat anterior parietal cortex: Relations among cytoarchitecture, single neuron functional properties and interhemispheric connectivity Journal of Neurophysiology 1981 45 667 697 7229676
Burton H Mitchell G Brent D Second somatic sensory area in the cerebral cortex of cats: somatotopic organization and cytoarchitecture Journal of Comparative Neurology 1982 210 109 135 7130474 10.1002/cne.902100203
| 15847693 | PMC1087848 | CC BY | 2021-01-04 16:03:47 | no | BMC Neurosci. 2005 Apr 22; 6:29 | utf-8 | BMC Neurosci | 2,005 | 10.1186/1471-2202-6-29 | oa_comm |
==== Front
BMC OphthalmolBMC Ophthalmology1471-2415BioMed Central London 1471-2415-5-61581118810.1186/1471-2415-5-6Research ArticleEffects of glaucoma drugs on ocular hemodynamics in normal tension glaucoma: a randomized trial comparing bimatoprost and latanoprost with dorzolamide [ISRCTN18873428] Zeitz Oliver [email protected] Eike T [email protected] Juliane [email protected] Anne [email protected] Lars [email protected] Peter [email protected] Gisbert [email protected] Maren [email protected] Universitätsklinikum Hamburg-Eppendorf, Klinik und Poliklinik für Augenheilkunde, Martinistr. 52, D- 20246 Hamburg, Germany2005 5 4 2005 5 6 6 24 10 2004 5 4 2005 Copyright © 2005 Zeitz 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
Reduced choroidal perfusion is hypothesized to play a role in the pathogenesis of normal tension glaucoma. Thus the impact of antiglaucomatous eye drops on ocular perfusion has been the focus of recent research and the subject of intensive investigations. The present study investigates whether topically applied latanoprost or bimatoprost influence ocular perfusion in patients with normal tension glaucoma and compares these effects with that changes detected after the treatment with dorzolamide.
Methods
Ocular hemodynamics were assessed by color Doppler imaging (CDI) shortly before and after a one-month treatment with either latanoprost, bimatoprost or dorzolamide. Primary end-points of the study were peak systolic and end-diastolic blood flow velocities in the short posterior ciliary artery (SPCA) under the new therapy. Intraocular pressure (IOP) and additional perfusion parameters in the SPCA and other retrobulbar vessels were tracked as observational parameters. n = 42 patients with normal tension glaucoma were enrolled in the study.
Results
Systolic and diastolic blood flow velocities in the SPCA showed no significant alteration after the treatment with latanoprost or bimatoprost. Dorzolamide lead to increase of peak systolic velocity. IOP was reduced by all three agents in a range reported in the literature.
Conclusion
Topically applied latanoprost and bimatoprost act in a hemodynamically neutral manner and have the capability to lower IOP even in patients with normal tension glaucoma and low initial IOP level. Dorzolamide accelerates blood flow in systole. None of the tested compounds has a negative impact on hemodynamics in the short posterior ciliary arteries.
==== Body
Background
Traditionally atrophy of the optic nerve head in glaucoma patients has been related to a chronically increased intraocular pressure (IOP).[1] In the majority of glaucoma patients increased IOP levels can be measured. At the same time there is a significant number of patients who develop the major features of glaucoma, meaning progressive visual field loss and atrophy of the optic nerve head, although the IOP is in the normal range below 21 mmHg in these individuals. [2-6] This variant of primary open angle glaucoma has been termed as normal tension glaucoma and the glaucomatous changes have been – at least in part – attributed to reduced perfusion of the optic nerve head by different authors. [6-8] Thus patients with normal tension glaucoma constitute a particularly interesting group for investigating pharmacological effects on ocular perfusion.
Up to now therapy of normal tension glaucoma does not differ substantially from the therapy of classical primary open angle glaucoma with increased IOP level: in both cases the focus is set on lowering IOP.[9,10] There is a tendency for defining lower target IOPs individually for each patient and optimizing ocular hemodynamics is still rated as complementary. Recently, several systemic approaches have been evaluated, e.g. oral administration of magnesium or the calcium antagonist nimodipine.[11,12] In addition, the hemodynamic effects of locally applied antiglaucomatous drugs are subject to intensive investigations within the last years: [13-16] Dorzolamide, an inhibitor of the carbonic anhydrase has shown the most promising effects on ocular hemodynamics. [17-19] In contrast, brinzolamide did not to alter ocular perfusion.[20] ß-receptor antagonists influence parameters of ocular hemodynamics differently depending on the compound used: betaxolol and carteolol lower vascular resistance while timolol seems to lead to an increased vascular tone.[15] The α2-receptor agonist brimonidine did not affect ocular perfusion in primary open angle glaucoma patients.[21] Derivatives of prostaglandins have become more and more popular for the treatment of glaucoma patients since they are very effective in lowering IOP and have to be applied only once a day.[22] Up to now there are predominantly four substances in clinical use: latanoprost, bimatoprost, travoprost and unoprostone. Only few studies have addressed the influence of prostaglandin analogues on ocular perfusion so far.
Prostaglandins constitute a large group of compounds produced endogenously with various and partially contrary effects on vascular tone.[23,24] Certain prostaglandins, such as PGI2 and PGE2, are potent vasodilators; others like PGF2α constricts arteries. [25-27] Investigations of the vasoactive effects of prostaglandins are complicated by the species dependency of their effects on vascular tone.[28] Latanoprost derives from PGF2α and has been reported to improve the perfusion of the optic nerve head[13], although in vitro studies indicate that latanoprost in contrast to its chemical origin PGF2α exerts no or minimal effects on vascular tone in the primate eye.[28] Bimatoprost has been demonstrated to have a vasoconstricting effect on isolated porcine ciliary arteries in vitro.[29] It is unknown whether this effect is of relevance in vivo and in humans.
The aim of the present study is to compare the hemodynamic effects of bimatoprost and latanoprost in patients with normal tension glaucoma. The effects were compared with the hemodynamic properties of dorzolamide, which has repeatedly been shown to improve ocular blood flow and can therefore be used as a reference compound for the evaluation of hemodynamic effects of antiglaucomatous eye drops. [17-19] It is hypothesized that these compounds lead to changes in systolic and diastolic blood flow velocities within the short posterior cilary artery. This vessel constitutes the main blood supply to the optic nerve head and alterations in its perfusion might play an important role in the pathogenesis and progression of normal tension glaucoma.
Methods
The study was performed in accordance to institutional, national, and international guidelines and was approved by the local ethics committee. The study was designed as an interventional, randomized, prospective, institutional, single-blinded, controlled, clinical trial.
Color Doppler imaging
Color Doppler imaging (CDI) was performed with a Sonoline Elegra Advanced System (Siemens, Erlangen, Germany) using a phased array transducer type 7.5L40 (Siemens, Erlangen, Germany) as described previously.[30] Ultrasound frequency was 6.5 MHz in the pulsed Doppler mode. The transducer was carefully set on the closed eye lid without exerting pressure on the bulb. Acoustic coupling between transducer and skin was optimized by a carbomeric gel (Vidisic®, Dr. Mann Pharma, Germany). After identification of the optic nerve as a landmark the following vessels were investigated in the color Doppler mode: central retinal artery (CRA), short and long posterior ciliary arteries (SPCA and LPCA) and ophthalmic artery (OA). Blood flow in the CRA was measured along its course through the optic nerve. Velocities in the SPCA and LPCA were recorded before entering the sclera. Flow velocity in the OA was measured close to its crossing of the optic nerve. The angle between transducer and orientation of the vessel was corrected. Gain and threshold were adjusted individually for each examination until noise disappeared and then kept constant during the entire examination. Pulse repetition frequency (PRF) was minimized to avoid aliasing (typically 5208 Hz for the OA, and 2500 Hz for the SPCA and LPCA in the Doppler mode). Sample size for all vessels was set constantly to 1.5 mm.
The changes of blood flow velocities over the course of a cardiac cycle were recorded continuously. Peak systolic velocity (PSV) and end diastolic velocity (EDV) can be determined directly and pulsatility index (PI) and resistive index (RI) are calculated automatically by the CDI software.
Patients
Consecutive patients with progressive normal tension glaucoma were randomized either to the latanoprost or to the bimatoprost group. Progression was defined as progressive excavation of the optic disc (funduscopic and controlled by the Heidelberg retina tomography – HRT) and/or progressive visual field loss in the Humphrey perimeter over the last 6–12 months. Glaucoma progression was assessed by the same glaucoma specialist (M.K.) in all patients. The patients underwent CDI measurements of ocular perfusion of the right eye by CDI shortly before and 3 to 5 weeks after initiation a local therapy with either latanoprost or bimatoprost. Both eye drops were applied once a day between 6 p.m. and 8 p.m. Normal tension glaucoma patients with unchanged excavation of the optic disc in funduscopy and HRT as well as stable visual field over the last 6–12 months served as negative controls. The glaucoma parameters in the control group were assessed by the same glaucoma specialist (M.K.) like in the therapy group. In the control group two subsequent CDI examinations were performed with an interval of 3 to 5 weeks. An additional group of patients with progressive normal tension glaucoma received dorzolamide eye drops three times a day, which has been reported to enhance blood flow in retrobulbar vessels. [17-19] This group was recruited to demonstrate that the CDI-methodology used is suitable to detect changes in retrobulbar blood flow. Regarding the assessment of progression the same criteria were applied as for the other study groups. CDI-examinations in the dorzolamide group were carried out shortly before and 3 to 5 weeks after initiation of the dorzolamide therapy.
All measurements in both therapy and control groups were carried out at 4 p.m. ± 2 h as a standardized time point to ensure best possible comparability. This specific time point was chosen due to organizational reasons. Prior to CDI intraocular pressure was measured. All examinations were performed with the patient sitting in an upright position and all the patients rested for several minutes before the examination started. CDI measurements for this study were performed by two experienced investigators (O.Z. and E.T.M.). To exclude investigator dependent effects, both CDI measurements in one patient were performed by the same physician. The study group membership and the individual therapy of the patients were masked for the CDI-investigator.
Patients with an intraocular pressure higher than 21 mmHg prior to therapy as well as patients with cardiovascular diseases were not included in the study. Systemic medication of the patients did not change between their first and second presentation. Patients who had ophthalmic surgery to the right eye were excluded from the study.
Bimatoprost
Lumigan® eye drops from Pharm Allergan (Ettlingen, Germany) were used in the bimatoprost treatment group. Lumigan® contains 0.3 mg/ml bimatoprost. The solvent contained benz alkonium chloride, sodium chloride, di sodium hydrogene phosphate 7H2O, citric acid, hydrochloric acid or sodium hydroxide (for pH adjustment), and water.
Latanoprost
Latanoprost (50 μg/ml) was obtained from Pharmacia Pfizer (Karlsruhe, Germany) as Xalatan®. Further ingredients are benz alkonium chloride, sodium chloride, sodium dihydrogene phosphate 1H2O, sodium monohydrogene phosphate, and water.
Dorzolamide
Dorzolamide was obtained in form of Trusopt® from MSD-Chibret, Munich, Germany. Trusopt® containes 22.26 mg dorzolamide hydrochloride per 1 ml. Further ingredients of Trusopt® are hydroxyethylcellulose, D-mannitole, sodium citrate, sodium hydroxide, benzalkonium chloride and water as solvent.
Statistics
Statistical analysis of the data was done with SPSS 10.0. All data are given as mean ± standard error of means (SEM). Student's t-test for paired data was used. It was hypothesized that the tested compounds would influence peak systolic and end-diastolic blood flow velocities (PSV and EDV) in the short posterior ciliary artery (SPCA). Thus two statistical tests were performed in each of the four study groups resulting in eight parallel statistical tests in total. Performing multiple statistical tests in a study necessitates the correction of the P-value to reach a significance level of 0.05. Therefore the Bonferroni-adjustment was applied and P < 0.006 is considered to be significant (double sided). All other parameters except EDV and PSV in SPCA were evaluated in a descriptive manner. Thus no P-values will be presented for these observational parameters.
In an a-priori-power-analysis the sample size was calculated. H1 was defined by an increase of flow velocity by 50% and a standard deviation of 35%. This definition was applied because preliminary data indicated a change in that range caused by dorzolamide. To reach a statistic power of 0.80 or more at least n = 9 individuals for PSV and n = 8 individuals for EDV per treatment group are required. For power analysis, the tool G*POWER V. 2.0 of F. Faul and E. Erdfelder [31] was applied.
Results
11 patients were treated with bimatoprost, 10 patients with latanoprost. Initially 9 patients were enrolled in the negative control group, but only 8 could be evaluated. One patient had to be excluded, because he did not return to the second examination due to personal reasons. 12 additional patients received dorzolamide. All data shown was obtained from measurements of the right eye of the patients.
Intraocular pressure (IOP)
Intraocular pressure was reduced from 14.6 ± 1.7 to 11.4 ± 0.8 mmHg under therapy with bimatoprost; latanoprost led to a decrease from 14.8 ± 2.2 to 12.3 ± 1.5 mmHg. Dorzolamide decreased the IOP in the dorzolamide group from 16.8 ± 1.6 to 14.8 ± 0.9 mmHg. In the negative control group intraocular pressure was 13.2 ± 1.7 mmHg at first presentation and 12.5 ± 1.9 mmHg at second presentation 3 to 5 weeks later (fig. 1).
Color Doppler imaging (CDI)
Neither bimatoprost nor latanoprost showed a statistically significant influence on peak systolic or enddiastolic velocities in the short posterior ciliary artery (fig. 2 and 3). All other parameters assessed by CDI measurements in the central retinal artery, short and long posterior ciliary arteries and ophthalmic artery appeared to be stable. Table 1 gives a detailed overview of the quantitative results obtained by CDI before and after treatment with latanoprost and table 2 does this for bimatoprost. In the negative control group all parameters were stable over time (table 3). Dorzolamide lead to a significant acceleration of systolic blood flow in the short posterior ciliary artery (table 4).
Discussion
Based on the presented results bimatoprost or latanoprost and have no significant influence on hemodynamics in the short posterior ciliary artery has to be rejected, although a tendency toward improved peak systolic and end diastolic blood flow velocities was detected in both treatment groups compared with the negative control group.
To our knowledge the present study is the first to directly compare hemodynamic effects of bimatoprost and latanoprost in patients with normal tension glaucoma. An altered blood flow velocity due to a vasoconstricting activity of bimatoprost reported from Allemann and colleagues[29] is not detectable in humans by CDI measurements. These differences might be explained by the in vitro system used by Allemann, who worked with isolated ciliary arteries from pigs. Apart from potential differences between species, the balance between neuronal, endothelial and myogenic influences on vascular tone is disturbed in such an in vitro setting. In vitro studies for latanoprost revealed minimal vasoactivity.[28] The chemical origin of latanoprost is the prostaglandin PGF2α, which increases arterial tone. [25-27] Bimatoprost is a derivative of the prostamide F2α, which is enzymatically formed from PGF2α and should act similar like PGF2α, but has a lower receptor affinity.[32] How vasoconstriction influences ocular perfusion depends on its site of action: Constriction of pre-capillary sphincters will cause a deterioration of blood flow. On the other hand, moderate constriction of the entire arterial system of an organ will increase perfusion pressure. From this theoretic view prostaglandins could influence ocular blood flow in positive and negative direction or both effects neutralize each other.
For latanoprost a handful of studies dealing with its effects on ocular perfusion have been published. Some of these studies state positive effects.[13,33,34] All these studies have in common that acute and short-term effects from less than an hour up to one week after application of the drug were observed. In accordance with the present results, studies with an extended follow up did not find any mid- or long-term effect of latanoprost on ocular hemodynamics.[13,35,36] A possible explanation for the difference between long- and short-term effects might be the acute reduction of intraocular pressure, which will result in an increased ocular perfusion pressure and leads to an improvement of ocular hemodynamics. For comparison with the presented results one has to point out that the previous reports were carried out with patients suffering from primary open angle glaucoma with initially increased IOP. Thus it is difficult to differentiate whether the observed improvement of ocular perfusion can be attributed to a primary effect of the active agent on the ocular vasculature or is secondary to IOP-reduction. In the setting of the present study the absolute IOP reduction in relation to ocular perfusion pressure is low and does affect ocular perfusion pressure only by approximately 5%. Retrobulbar blood flow velocities are directly proportional to ocular perfusion pressure. Subsequently, the effect of such a change in intraocular pressure on retrobulbar blood flow velocities is low, particularly compared to the changes seen after the administration of dorzolamide. Nevertheless, the tendency of an increase of retrobulbar blood flow velocity is visible in all treated groups.
Patients with normal tension glaucoma are therefore a particularly interesting group for the clinical investigation of hemodynamic effects of antiglaucomatous eye drops since results are less skewed by changes in IOP. In addition, a reduction in ocular perfusion might play a major role in the pathogenesis of normal tension glaucoma.[37] Reversal of this condition could be of great clinical benefit for this patient group and deterioration of ocular perfusion is expected to accelerate progression of glaucomatous optic neuropathy. Therefore normal tension glaucoma requires a therapy which ideally improves hemodynamics of the eye, but at least does not alter it in a negative way.
Although not primary end-point of the present study, the results indicate that latanoprost and bimatoprost have the capability to reduce IOP in patients with normal tension glaucoma, despite the fact that initial IOP levels before the treatment are low. Relative IOP-reduction in the present study is smaller with previous reports on treatment of patients with primary open angle glaucoma.[22,38,39] This is most likely due to the low initial intraocular pressure in the patient population with normal tension glaucoma. Studies on the IOP-lowering effect of prostaglandins, predominantly latanoprost, in normal tension glaucoma find an IOP-reduction in a similar magnitude. [40-42] This can be taken as an indicator for good compliance of the patients and it can be concluded that both tested compounds have the capability to reliably decrease IOP even in the lower range.
Some ophthalmologists doubt the reliability of CDI, but the reliability of the CDI method has repeatedly been shown in vitro[4] and in vivo[43]. Experiments performed by our group are in accordance with these findings.[44] The positive control with dorzolamide in the present study underlines the applicability of CDI for assessing pharmacological effects on ocular perfusion. To resolve the potential problem of investigator dependency both CDI-examinations for each patient were performed by the same investigator. In addition, the CDI procedure was performed in a highly standardized manner in our laboratory minimizing investigator dependency of the results.
The precise and clinically relevant evaluation of ocular perfusion still poses a great challenge. Quantitative determination of total ocular blood flow with a single method is not possible, but combination of different methods is proposed to give a better approximation.[45] This fact limits the explanatory power of this and other studies addressing ocular perfusion in ophthalmological diseases. Aim of the present study was to focus on glaucoma patients and it is hypothesized that glaucoma is associated with a localized disturbance of ocular hemodynamics at the optic disc. The optic disc is supplied with blood by the posterior ciliary arteries. The most reliable method for evaluation of hemodynamics in the short posterior ciliary artery is color Doppler imaging. Blood flow velocity in the short posterior ciliary arteries is influenced on the one hand by the vascular tone of the vessel itself but also by the resistance of the dependent downstream vasculature. Thus blood flow velocities in the short posterior ciliary artery reflect also the hemodynamics at the level of the optic disc. The influence of the tested compounds on perfusion of the entire eye cannot be answered by the present study.
Recently it has been shown that the correlation between measurements of ocular hemodynamics with different methodologies is limited.[30] Thus differences in outcome between studies investigating the effects of the same compound on ocular perfusion might at least partially be explained by the variability between the methodologies. Color Doppler imaging (CDI) measures blood flow velocities. Measurement of ocular blood flow by the method of Langham (LOBF) is based on assessing pulse synchronal oscillations of the IOP and calculate ocular perfusion from its amplitude by mathematical algorithms.[46] From the parameter determined by CDI only the pulsatility index correlates with the perfusion index determined by LOBF.[30] Therefore the comparison of the results of the present study e.g. with the work by Georgopoulus et al.[33] is limited.
Conclusion
In summary no significant influence of latanoprost and bimatoprost on ocular hemodynamics was detected in the present study. In accordance with previous reports dorzolamide led to an increase of blood flow velocity in systole. A suitable compound for the treatment of normal tension glaucoma should at least act hemodynamically neutral. This minimum requirement is fulfilled by all three tested compounds. Only dorzolamide was capable to improve perfusion in the posterior ciliary arteries. When comparing both prostaglandin-like eye drops, they showed similar effects on ocular hemodynamics and IOP. Subsequently latanoprost and bimatoprost can be considered as equivalent in the treatment of normal tension glaucoma. However, in order to comprehensively treat the pathogenic disturbances leading or at least contributing to normal tension glaucoma there is still a great demand for a pharmacological compound which significantly enhances perfusion of the optic nerve head.
Competing interests
M.K. received a grant from Pharm-Allergan (manufacturer of bimatoprost) in 2003. This grant is not related to the present study. Neither the present study nor the present manuscript is financed by this grant. There are no further commercial relationships for any author.
Authors' contributions
OZ designed and co-ordinated the study. ETM and OZ performed the CDI measurements. JR and AW documented the data and did under supervision of OZ the statistical analysis. LW and PG were additionally involved in editing the manuscript and interpreting the data. PG contributed knowledge about basic science aspects of prostaglandin action on vasculature. GR and MK supported the other authors in interpreting the results. MK as glaucoma specialist was additionally responsible for patient recruitment and randomization.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The authors would like to gratefully acknowledge Birgit Bischoff and Fatima Icagic for their excellent technical support in the study.
Figures and Tables
Figure 1 Intraocular pressure changes. Intraocular pressure (IOP) at first and second examination. IOP was an observational parameter and was not subjected to statistical testing.
Figure 2 Peak systolic velocity. Peak systolic velocity (PSV) at first and second examination. In tendency, both compounds led to a slight increase of PSV, which was statistically not significant. Dorzolamide led to a statistical increase of PSV.
Figure 3 End diastolic velocity. End-diastolic velocity (EDV) at first and second examination. In tendency, both compounds led to a slight increase of EDV, which was statistically not significant.
Table 1 Influence of bimatoprost on retrobulbar hemodynamics. Influence of bimatoprost on ocular perfusion. Overview of the results from CDI measurements before and one month after regular application of bimatoprost: Increase of blood flow velocities were statistically not significant (n = 11). Bold printed measures were subjected to statistical testing; all other parameters were observational only.
PSV (cms-1) EDV (cms-1) RI PI
OA before bimatoprost 31.6 ± 3.5 3.5 ± 0.9 0.89 ± 0.02 2.76 ± 0.20
after bimatoprost 38.5 ± 4.8 4.2 ± 0.7 0.87 ± 0.03 2.53 ± 0.21
CRA before bimatoprost 7.9 ± 0.9 1.4 ± 0.2 0.79 ± 0.04 2.22 ± 0.31
after bimatoprost 8.1 ± 0.5 1.7 ± 0.1 0.79 ± 0.02 1.90 ± 0.10
SPCA before bimatoprost 8.0 ± 1.1 1.9 ± 0.1 0.71 ± 0.03 1.61 ± 0.15
after bimatoprost 10.1 ± 1.2 2.8 ± 0.4 0.75 ± 0.01 1.66 ± 0.05
LPCA before bimatoprost 14.8 ± 3.0 3.0 ± 0.7 0.78 ± 0.03 1.90 ± 0.17
after bimatoprost 16.8 ± 2.2 3.0 ± 0.3 0.80 ± 0.02 1.90 ± 0.12
Abbreviations: OA – Ophthalmic artery, CRA – central retinal artery, SPCA – short posterior ciliary artery, LPCA – long posterior ciliary artery, PSV – peak systolic velocity, EDV – enddiastolic velocity, RI – resistive index, PI – pulsatility index.
Table 2 Influence of latanoprost on retrobulbar hemodynamics. Influence of latanoprost on ocular perfusion. Overview of the results from CDI measurements before and one month after regular application of latanoprost: Increase of blood flow velocities were statistically not significant (n = 10). Bold printed measures were subjected to statistical testing; all other parameters were observational only.
PSV (cms-1) EDV (cms-1) RI PI
OA before latanoprost 31.7 ± 3.1 4.6 ± 0.9 0.87 ± 0.02 2.59 ± 0.15
after latanoprost 39.0 ± 10.1 6.2 ± 1.2 0.83 ± 0.01 2.14 ± 0.09
CRA before latanoprost 7.6 ± 0.6 1.5 ± 0.1 0.78 ± 0.01 1.96 ± 0.10
after latanoprost 8.4 ± 0.9 1.7 ± 0.1 0.79 ± 0.02 1.95 ± 0.13
SPCA before latanoprost 8.6 ± 1.8 1.8 ± 0.2 0.74 ± 0.03 1.75 ± 0.13
after latanoprost 9.5 ± 1.1 2.5 ± 0.4 0.73 ± 0.03 1.60 ± 0.11
LPCA before latanoprost 13.1 ± 1.5 2.5 ± 0.3 0.81 ± 0.02 1.96 ± 0.14
after latanoprost 13.0 ± 1.9 2.5 ± 0.4 0.78 ± 0.04 1.90 ± 0.21
Abbreviations: OA – Ophthalmic artery, CRA – central retinal artery, SPCA – short posterior ciliary artery, LPCA – long posterior ciliary artery, PSV – peak systolic velocity, EDV – enddiastolic velocity, RI – resistive index, PI – pulsatility index.
Table 3 Influence of dorzolamide on retrobulbar hemodynamics. Ocular perfusion in the dorzolamide group. CDI measurements obtained from the patient group treated with dorzolamide (n = 12). The time period between between first and second measurement was 3–5 weeks. Data are given as mean ± standard deviation. Bold printed measures were subjected to statistical testing; all other parameters were observational only.
PSV (cms-1) EDV (cms-1) RI PI
OA before dorzolamide 29.6 ± 1.2 3.3 ± 0.4 0.89 ± 0.02 2.82 ± 0.25
after dorzolamide 37.2 ± 3.5 4.4 ± 0.7 0.88 ± 0.02 2.64 ± 0.20
CRA before dorzolamide 7.3 ± 0.8 1.4 ± 0.1 0.80 ± 0.02 1.98 ± 0.18
after dorzolamide 8.5 ± 0.8 1.8 ± 0.2 0.78 ± 0.02 1.97 ± 0.24
SPCA before dorzolamide 8.5 ± 0.8 1.9 ± 0.2 0.77 ± 0.02 1.79 ± 0.11
after dorzolamide 12.8 ± 1.4 * 2.2 ± 0.2 0.80 ± 0.03 1.91 ± 0.13
LPCA before dorzolamide 10.9 ± 0.8 2.0 ± 0.3 0.73 ± 0.07 1.80 ± 0.22
after dorzolamide 16.9 ± 1.7 2.9 ± 0.3 0.83 ± 0.01 1.95 ± 0.12
Asterisk (*) denotes statistic significance (P < 0.006). Abbreviations: OA – Ophthalmic artery, CRA – central retinal artery, SPCA – short posterior ciliary artery, LPCA – long posterior ciliary artery, PSV – peak systolic velocity, EDV – enddiastolic velocity, RI – resistive index, PI – pulsatility index.
Table 4 CDI measurements in the control group. Ocular perfusion in the control group. CDI measurements obtained from the control group (n = 8). The time period between between first and second measurement was 3–5 weeks. Local and systemic medication was not changed between the two examinations of the patient. Data are given as mean ± standard deviation. Bold printed measures were subjected to statistical testing; all other parameters were observational only.
PSV (cms-1) EDV (cms-1) RI PI
OA first examination 29.3 ± 2.6 3.7 ± 0.6 0.87 ± 0.02 2.57 ± 0.29
second examination 30.6 ± 3.2 5.4 ± 1.0 0.83 ± 0.02 2.08 ± 0.20
CRA first examination 8.2 ± 1.5 1.8 ± 0.4 0.76 ± 0.03 1.93 ± 0.19
second examination 7.5 ± 0.9 1.4 ± 0.1 0.80 ± 0.03 2.14 ± 0.21
SPCA first examination 10.8 ± 1.6 2.6 ± 0.5 0.75 ± 0.02 1.76 ± 0.17
second examination 10.1 ± 2.4 2.5 ± 0.6 0.75 ± 0.02 1.55 ± 0.09
LPCA first examination 18.0 ± 3.8 3.1 ± 0.8 0.82 ± 0.04 2.01 ± 0.21
second examination 18.0 ± 4.0 3.6 ± 0.9 0.77 ± 0.01 1.69 ± 0.12
Abbreviations: OA – Ophthalmic artery, CRA – central retinal artery, SPCA – short posterior ciliary artery, LPCA – long posterior ciliary artery, PSV – peak systolic velocity, EDV – enddiastolic velocity, RI – resistive index, PI – pulsatility index.
==== Refs
Goldmann H Open-angle glaucoma Br J Ophthalmol 1972 56 242 248 5032760
Sommer A Tielsch JM Katz J Quigley HA Gottsch JD Javitt J Singh K Relationship between intraocular pressure and primary open angle glaucoma among white and black Americans. The Baltimore Eye Survey Arch Ophthalmol 1991 109 1090 1095 1867550
Pillunat LE Stodtmeister R Marquardt R Mattern A Ocular perfusion pressures in different types of glaucoma Int Ophthalmol 1989 13 37 42 2744953
Klingmuller V Schmidt KG von Ruckmann A Koch B Stein A [Doppler sonography of the short posterior ciliary artery in patients with primary open angle glaucoma] Ultraschall Med 2000 21 32 37 10746282 10.1055/s-2000-8928
Maumenee AE Causes of optic nerve damage in glaucoma. Robert N. Shaffer lecture Ophthalmology 1983 90 741 752 6413917
Schmidt KG Ruckmann AV Mittag TW Hessemer V Pillunat LE Reduced ocular pulse amplitude in low tension glaucoma is independent of vasospasm Eye 1997 11 485 488 9425412
Schumann J Orgul S Gugleta K Dubler B Flammer J Interocular difference in progression of glaucoma correlates with interocular differences in retrobulbar circulation Am J Ophthalmol 2000 129 728 733 10926980 10.1016/S0002-9394(99)00481-X
Boehm AG Pillunat LE Koeller U Katz B Schicketanz C Klemm M Richard G Regional distribution of optic nerve head blood flow Graefes Arch Clin Exp Ophthalmol 1999 237 484 488 10379609 10.1007/s004170050266
Collaborative normal tension glaucoma study group The effectiveness of intraocular pressure reduction in the treatment of normal-tension glaucoma. Collaborative Normal-Tension Glaucoma Study Group Am J Ophthalmol 1998 126 498 505 9780094 10.1016/S0002-9394(98)00272-4
Collaborative normal tension glaucoma study group Comparison of glaucomatous progression between untreated patients with normal-tension glaucoma and patients with therapeutically reduced intraocular pressures. Collaborative Normal-Tension Glaucoma Study Group Am J Ophthalmol 1998 126 487 497 9780093 10.1016/S0002-9394(98)00223-2
Piltz JR Bose S Lanchoney D The effect of nimodipine, a centrally active calcium antagonist, on visual function and mascular blood flow in patients with normal-tension glaucoma and control subjects J Glaucoma 1998 7 336 342 9786563
Gaspar AZ Gasser P Flammer J The influence of magnesium on visual field and peripheral vasospasm in glaucoma Ophthalmologica 1995 209 11 13 7715920
Tamaki Y Nagahara M Araie M Tomita K Sandoh S Tomidokoro A Topical latanoprost and optic nerve head and retinal circulation in humans J Ocul Pharmacol Ther 2001 17 403 411 11765145 10.1089/108076801753266785
Tamaki Y Araie M Tomita K Nagahara M Sandoh S Tomidokoro A Effect of topical unoprostone on circulation of human optic nerve head and retina J Ocul Pharmacol Ther 2001 17 517 527 11777175 10.1089/10807680152729202
Altan-Yaycioglu R Turker G Akdol S Acunas G Izgi B The effects of beta-blockers on ocular blood flow in patients with primary open angle glaucoma: a color Doppler imaging study Eur J Ophthalmol 2001 11 37 46 11284483
Schmidt KG von Ruckmann A Becker R Pillunat LE [Ocular pulse amplitude, intraocular pressure and beta blocker/carbonic anhydrase inhibition in combined therapy of primary open-angle glaucoma] Klin Monatsbl Augenheilkd 1999 215 361 366 10637801
Schmidt KG von Ruckmann A Pillunat LE Topical carbonic anhydrase inhibition increases ocular pulse amplitude in high tension primary open angle glaucoma Br J Ophthalmol 1998 82 758 762 9924367
Schmidt KG Dick B von Ruckmann A Pillunat LE [Ocular pulse amplitude and local carbonic anhydrase inhibition] Ophthalmologe 1997 94 659 664 9410235 10.1007/s003470050179
Pillunat LE Bohm AG Koller AU Schmidt KG Klemm M Richard G Effect of topical dorzolamide on optic nerve head blood flow Graefes Arch Clin Exp Ophthalmol 1999 237 495 500 10379611 10.1007/s004170050268
Klemm M Zeitz O Reuss J Matthiessen ET Richard G [Therapy of normal tension glaucoma: effect of brinzolamide on ocular haemodynamics] Klin Monatsbl Augenheilkd 2003 220 330 333 12766822 10.1055/s-2003-39432
Schmidt KG Klingmuller V Gouveia SM Osborne NN Pillunat LE Short posterior ciliary artery, central retinal artery, and choroidal hemodynamics in brimonidine-treated primary open-angle glaucoma patients Am J Ophthalmol 2003 136 1038 1048 14644214 10.1016/S0002-9394(03)00631-7
Parrish RK Palmberg P Sheu WP A comparison of latanoprost, bimatoprost, and travoprost in patients with elevated intraocular pressure: a 12-week, randomized, masked-evaluator multicenter study Am J Ophthalmol 2003 135 688 703 12719078 10.1016/S0002-9394(03)00098-9
Terragno NA Prostaglandins, antidiuretic hormone and renin angiotensin system Hypertension 1981 3 II 65-70 6795118
Bassenge E Endothelial function in different organs Prog Cardiovasc Dis 1996 39 209 228 8970574
Kaley G Hintze TH Panzenbeck M Messina EJ Role of prostaglandins in microcirculatory function Adv Prostaglandin Thromboxane Leukot Res 1985 13 27 35 3159211
Nakayama K Osol G Halpern W Differential contractile responses of pressurized porcine coronary resistance-sized and conductance coronary arteries to acetylcholine, histamine and prostaglandin F2 alpha Blood Vessels 1989 26 235 245 2620104
Maigaard S Forman A Andersson KE Different responses to prostaglandin F2 alpha and E2 in human extra- and intramyometrial arteries Prostaglandins 1985 30 599 607 3866262 10.1016/0090-6980(85)90024-3
Stjernschantz J Selen G Astin M Resul B Microvascular effects of selective prostaglandin analogues in the eye with special reference to latanoprost and glaucoma treatment Prog Retin Eye Res 2000 19 459 496 10785618 10.1016/S1350-9462(00)00003-3
Allemann R Flammer J Haefliger IO Vasoactive properties of bimatoprost in isolated porcine ciliary arteries Klin Monatsbl Augenheilkd 2003 220 161 164 12664371 10.1055/s-2003-38167
Zeitz O Matthiessen E Richard G Klemm M Estimation of choroid perfusion by colour Doppler imaging vs. other methods Ultrasound Med Biol 2002 28 1023 12217438 10.1016/S0301-5629(02)00555-0
Faul F Erdfelder E G*Power
Matias I Chen J Petrocellis LD Bisogno T Ligresti A Fezza F Krauss AH Shi L Protzman CE Li C Liang Y Nieves AL Kedzie KM Burk RM Di Marzo V Woodward DF Prostaglandin-ethanolamides (prostamides): in vitro pharmacology and metabolism J Pharmacol Exp Ther 2004 309 745 757 14757851 10.1124/jpet.103.061705
Georgopoulos GT Diestelhorst M Fisher R Ruokonen P Krieglstein GK The short-term effect of latanoprost on intraocular pressure and pulsatile ocular blood flow Acta Ophthalmol Scand 2002 80 54 58 11906305 10.1034/j.1600-0420.2002.800111.x
Geyer O Man O Weintraub M Silver DM Acute effect of latanoprost on pulsatile ocular blood flow in normal eyes Am J Ophthalmol 2001 131 198 202 11228295 10.1016/S0002-9394(00)00797-2
Arend O Harris A Wolter P Remky A Evaluation of retinal haemodynamics and retinal function after application of dorzolamide, timolol and latanoprost in newly diagnosed open-angle glaucoma patients Acta Ophthalmol Scand 2003 81 474 479 14510794 10.1034/j.1600-0420.2003.00122.x
Harris A Migliardi R Rechtman E Cole CN Yee AB Garzozi HJ Comparative analysis of the effects of dorzolamide and latanoprost on ocular hemodynamics in normal tension glaucoma patients Eur J Ophthalmol 2003 13 24 31 12635671
Flammer J Orgul S Costa VP Orzalesi N Krieglstein GK Serra LM Renard JP Stefansson E The impact of ocular blood flow in glaucoma Prog Retin Eye Res 2002 21 359 393 12150988 10.1016/S1350-9462(02)00008-3
Cantor LB Bimatoprost: a member of a new class of agents, the prostamides, for glaucoma management Expert Opin Investig Drugs 2001 10 721 731 11281821
Gandolfi SA Cimino L Effect of bimatoprost on patients with primary open-angle glaucoma or ocular hypertension who are nonresponders to latanoprost Ophthalmology 2003 110 609 614 12623831 10.1016/S0161-6420(02)01891-2
Hoyng PF Kitazawa Y Medical treatment of normal tension glaucoma Surv Ophthalmol 2002 47 Suppl 1 S116 24 12204707 10.1016/S0039-6257(02)00322-3
Liu CJ Ko YC Cheng CY Chiu AW Chou JC Hsu WM Liu JH Changes in intraocular pressure and ocular perfusion pressure after latanoprost 0.005% or brimonidine tartrate 0.2% in normal-tension glaucoma patients Ophthalmology 2002 109 2241 2247 12466165 10.1016/S0161-6420(02)01247-2
McKibbin M Menage MJ The effect of once-daily latanoprost on intraocular pressure and pulsatile ocular blood flow in normal tension glaucoma Eye 1999 13 ( Pt 1) 31 34 10396380
Baxter GM Williamson TH Color Doppler imaging of the eye: normal ranges, reproducibility, and observer variation J Ultrasound Med 1995 14 91 96 8568968
Matthiessen ET Zeitz O Richard G Klemm M Reproducibility of blood flow velocity measurements using colour decoded Doppler imaging Eye 2004 18 400 405 15069438 10.1038/sj.eye.6700651
Rechtman E Harris A Kumar R Cantor LB Ventrapragada S Desai M Friedman S Kagemann L Garzozi HJ An update on retinal circulation assessment technologies Curr Eye Res 2003 27 329 343 14704917 10.1076/ceyr.27.6.329.18193
Langham ME To'Mey KF A clinical procedure for the measurements of the ocular pulse-pressure relationship and the ophthalmic arterial pressure Exp Eye Res 1978 27 17 25 668825 10.1016/0014-4835(78)90049-0
| 15811188 | PMC1087849 | CC BY | 2021-01-04 16:03:49 | no | BMC Ophthalmol. 2005 Apr 5; 5:6 | utf-8 | BMC Ophthalmol | 2,005 | 10.1186/1471-2415-5-6 | oa_comm |
==== Front
BMC Public HealthBMC Public Health1471-2458BioMed Central London 1471-2458-5-291580198410.1186/1471-2458-5-29CommentaryPublic health education: A report from Mosul and a plan for change Hughes Gail D [email protected] Ally [email protected] Kathie Stromile [email protected] Department of Preventive Medicine-Epidemiology, University of Mississippi Medical Center, 2500 North State Street, Jackson, MS 39216, USA2 Mississippi Consortium for International Development (MCID), 1225 Robinson Street, Jackson, MS 39203, USA3 Department of International Programs, Mississippi Valley State University, 14000 Highway 82 W., #5098 Itta Bena, MS 38941-1400, USA2005 31 3 2005 5 29 29 3 9 2004 31 3 2005 Copyright © 2005 Hughes et al; licensee BioMed Central Ltd.2005Hughes 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
Today Iraq suffers from severe shortages of food, medicine, clean water and adequate sanitation. Malnutrition and communicable diseases are major factors in the rising morbidity and mortality rates. However, supplies and equipment are insufficient or outmoded, and public health training is outdated. The Universities have been unable to help because under-funding and isolation from their professional colleagues has limited their effectiveness.
Methods
To revitalize public health education, we describe a partnership between a US education consortium and the University of Mosul that will be carried out in the next several years. The plan is based on "three R's": Recovery from the past damage due to war and neglect; Retooling of key public health faculty to remedy the years of isolation and restriction of activity; and Reestablishment of the University as a resource for the its constituents, for the community and for other educational institutions. In all these activities, Iraqi minorities, especially women, will participate and contribute.
Conclusion
The work to repair the public health educational infrastructure has just begun. When completed, it will represent a small but necessary step in restoring normalcy to the people of Mosul, and of Iraq.
==== Body
Background
The current situation in Iraq
Today Iraq is a country plagued by poverty, deprivation, and disease. The war with Iran in the 1980's, the Persian Gulf War in 1991, a crippling economic blockade with lasting economic sanctions, and finally the recent war to oust the Hussein regime from power have caused severe shortages of food, medicine, clean water and adequate sanitation. The current state of the health system has been described as "a shambles" (Aziz, 2003).
The problems are both acute and chronic. Malnutrition, high rates of communicable diseases, and increased mortality have been documented by Iraqi scholars and international researchers. Children, who comprise nearly half the Iraqi population (UNICEF, 2003. At a glance: Iraq) are most vulnerable and have been most studied. Jawadi found that 38 percent of 1881 children ranging in age from newborn to five who attended a Mosul healthcare center were chronically malnourished and 30 percent suffered wasting attributable to "the increase of water-borne diseases like severe gastroenteritis together with the shortage of basic food" (Al-Jawadi, 1996). A later investigation reported increased mortality as the result (Al-Jawadi, 1999). UNICEF studies confirm these findings in the country as a whole (UNICEF, 2003. The Situation of Children) and chart the rise in mortality risk for small children over the last 14 years (Ali MM 2003). Beginning in 1974, mortality rates for infants and for children under five in Iraq declined overall until 1990, but then essentially doubled and remained high. Now one in every eight children will die before age five, making Iraq's rate the 36th highest of about 200 countries reporting to UNICEF. (UNICEF, 2003. The state of the world's children pp 102-5)
The shortage of food in Mosul reported by Jawadi is experienced throughout Iraq, contributing to the risk of vulnerable populations. In May 2003, UNICEF reported that acute malnutrition among children had almost doubled since before the war, and stood at 7.7 percent. Part of the cause was the collapse of the Primary Health Care Centres, which identify and treat malnutrition (UNICEF, 2003. Restore public health system for malnourished children). Of the country's 1200 Centres, some 20 percent have been destroyed or looted, while half the balance needs extensive repairs. Almost none have a consistent water supply, and about a third have no physician on staff (Volker 2004).
Poor sanitation, polluted water and decline in health services are among the reasons for this disastrous situation, (Al Nouri 2003) exacerbated by the recent military hostilities. Prior to the first Gulf war, almost all the urban population of Iraq had piped, filtered and disinfected water (Martone 2003). Destruction of the electric and the supply infrastructure since then, coupled with neglect, has resulted in severe water pollution. Concurrently, communicable disease rates have risen to near-epidemic proportions, with 70 percent of children in Baghdad suffering from diarrhea at the end of the present war (Martone 2003). Shortages of certain laboratory agents and reagents that could not be imported during the sanctions are another major factor in the increased morbidity and mortality from meningitis, influenza, pneumonia, tuberculosis, and other infectious diseases. Even now, laboratory and medical supplies are extremely difficult to get, and the staff of one hospital in Mosul fought off looters of these items with their bare hands (Dyer, 2003). The dearth of equipment and the outdated training of public health professionals contribute to the difficulty of making an accurate diagnosis. For example, one study found that the antiquated methods and technologies of diagnosis were responsible for a 35 percent misdiagnosis rate in cases of typhoid fever (Al-Dabbagh, 1999).
Iraq's health system was once among the most advanced in the Middle East (Aziz 2003). There are plans for rebuilding Iraqi science (Stone 2004), and public health training has been listed as a priority in this effort by the Iraqi Ministry of Health and others (Fleck 2003). Because progress depends on professionals who can lead the way forward by training others, developing a cadre of skilled workers is essential for the task of rebuilding health care. However, virtually no training of health professionals has been available for 15 years, medical libraries are out of date, Internet access is minimal, and educational facilities have been destroyed. (Amin 2003; Furber 2004). Thus the training must begin at the top, with those who will become trainers themselves.
The need for improved professional training runs through all proposals for Iraqi reconstruction. In particular, public health training is a priority because many of Iraq's health problems can be addressed with public health solutions. Here we describe a plan at the University of Mosul that is now being implemented, with the goal of reestablishing a major Iraqi university as a source of expertise, education and service in public health.
Methods
A plan to reestablish the University of Mosul: The three R's
The Mississippi Consortium for International Development (MCID) is a collaborative endeavor of four historically black institutions of higher learning in Mississippi (Alcorn State University, Jackson State University, Mississippi Valley State University, and Tougaloo College). The Consortium provides international human resource development training and technical assistance and currently maintains formal partnerships with more than 50 universities worldwide. In July 2003, MCID established an International Partnership between the Consortium members, the University of Mississippi Medical Center and the University of Mosul to plan for the revitalization of this formerly eminent university, and a plan was mapped out for the revitalization in the next six months. Public health and sanitation were chosen as the focus of the effort because they are essential areas in Iraq's recovery. The plan, known as the HEAD (Higher Education and Development) Project, is now in its first year of operation.
Three phases of the plan, the "Three R's", describe the thrust of the HEAD Project:
• Recovery from damage at the University of Mosul due to political party policies, looting, and allowing a return to normal functioning.
• Retooling of key faculties, updating knowledge to allow their reentry into professional activities worldwide.
• Reestablishment of the institutional viability of the University, as well as its relationships with community organizations and other institutions of higher education.
The Partners in the HEAD Project support full participation of Iraq's minorities, including women, in the implementation of the three R's. This policy of inclusion gives the University a leadership role in the development of the new Iraq.
Recovery of the facility
The most immediate need of the University of Mosul is replacing laboratory, classroom and office equipment, and textbooks destroyed or stolen during the April 9–10 2003 looting rampage. The University has initiated repairs to physical facilities, but decisions facing the University go far beyond mere replacement of a basic stock of equipment and supplies. More than two decades of under-funding and nearly complete isolation from the outside world since 1991 have rendered the Mosul faculty unable to specify their needs. When asked, professors readily admit that they are not aware of the advances made in their academic disciplines or of the options available to them. They suggest that the best way to solve the dilemma is to include equipment identification as a part of the Partnership faculty exchange program (described below). To rush these decisions could result in purchasing out of date, inappropriate, or unneeded equipment.
The HEAD Project proposes to fund limited equipment purchases throughout the University's colleges rather than exclusively for the Colleges of Medicine, Nursing, and Engineering, where public health training is usually carried out. Project staff believes that the viability of individual colleges will depend upon the institution as a whole being functional. University officials, with MCID, are determining which purchases should be made immediately to assure that essential services are available to the administration, faculty and students to conduct classes and support sensitive research projects.
In the past, the University of Mosul has benefited from donations of equipment, textbooks and journals from U.S. companies and private individuals. MCID proposes to conduct a campaign during the initial year to elicit additional support for the HEAD Project from a select number of private entities at the beginning of the project. The HEAD Project will target at least 15 U.S. and foreign businesses operating in Iraq with established corporate social responsibility funding programs. The next step is an intensive round of meetings with the Baghdad-based representatives of these companies to identify activities within the HEAD Project that match their corporate interests or the priorities of their social responsibility programs. Past experience suggest that the matching process does not take long but the decision can take several months. Obtaining social responsibility funds invariably involves approval by the corporate headquarters, even though the recommendation of their in-country personnel is typically followed. In the interim, and in addition to these public-private partnerships described above, additional donors of smaller gifts such as back issues of technical journals and excess laboratory equipment will be sought.
Retooling the faculty
There is broad consensus among the more than 2,000 University of Mosul faculty members that the intellectual isolation (some would argue 25 years) has left them woefully out of touch with their professions. They argue convincingly that to deliver a quality education to their students and to participate in the debates that define the outer reaches of their disciplines, a program must be undertaken to rapidly bring them up to date. A critical first step is improved access to information via the Internet and subscriptions to academic journals. Beyond that, however, direct contact with colleagues in the Western world is essential, through refresher courses, faculty exchange programs, sabbaticals, and participation in international conferences. The majority of exchange programs will bring faculty to the United States, but visitations by MCID faculty to assist in designing the retooling agenda are also planned if security permits.
Retooling cannot be completed quickly, for the teaching pool must be maintained at an adequate level during the academic year. Therefore, we propose to spread this component over a three-year period. The cadence will be negotiated by a coordinating committee that includes representatives of the Colleges of Engineering, Nursing and Medicine, where most of the training will take place, with the President of the University serving as the final arbiter of the distribution of the training programs among the colleges.
The first year of work requires achievable objectives while simultaneously focusing on long term capacity building. Several key efforts are now ongoing or planned for the near future. These are set out below.
Faculty refresher courses address the request of University of Mosul professors for two-week reviews of key public health areas. Courses for senior staff include clinical epidemiology, primary health care, demography, and maternal and child health. Recommended refresher courses for junior staff are biostatistics, social and behavioral sciences, and occupational health. The courses are based on the model, Train the Trainer, whereby the newly trained faculty at the University of Mosul will replicate the courses for students and offer them to other public health professionals and medical practitioners.
Faculty Exchange Programs allow faculty to learn first-hand of innovative and state-of-the-art technology and trends in their respective professions by visiting MCID institutions. The major exchanges may be for any period from two weeks to six months. Faculty are surveyed on their research interests and needed areas of growth; they are then matched to an MCID institution and faculty member to assist them in professional development. Collaborative research, curriculum development, and participation in lectures, grand rounds, and national and international conferences are all encouraged. In addition, a Research Mentoring Program, which will foster collaboration between faculty at Mosul University and MCID faculty/research investigators, is planned in order to provide increased one-to-one faculty interaction.
An MCID faculty residency will target the development of a specific project. For example, Mosul faculty specializing in environmental health might take residency at Jackson State University and participate in work addressing a specific environmental health project being studied there; similarly, an MCID faculty member would assume residency at the University of Mosul to serve as technical advisor in the development of targeted activities. The faculty residencies are one to four months in length.
Conferences and workshops are essential components of faculty retooling. Funds are available to support participation in national and international activities in areas of public health, sanitation and nursing. Examples are:
• Identification of best practices and applications in specific areas of public health, sanitation and nursing given the needs of northern Iraq;
• Participation in skills-building workshops in targeted areas of professional growth in public health, sanitation and nursing;
• Exchange of research findings and technical information between the University of Mosul and other higher educational institutions in Iraq and the Middle East, and/or Iraqi governmental structures in public health, sanitation and nursing.
• Representation of the University of Mosul and/or Iraq at national and international conferences in public health, sanitation and nursing.
All of the aforementioned programs have an emphasis on minorities, especially women. Furthermore, allowances will be made to ensure active participation of minority faculty. In addition, research and course development on health issues specific to minorities will be strongly encouraged through each initiative.
Reestablishment of institutional viability
No institution, especially an educational institution, can thrive in isolation. Frustrated by years of financial neglect, over-crowding, and deprived of contact with the outside world, the faculty and staff of the University of Mosul have remained at their institution only out of sheer dedication. Over the past 12 years, many of their colleagues have left for more receptive environments where they could teach and conduct research in tranquility. For those who remained at the University, life has been difficult. They endured policies intended to exclude the interests of ethnic and religious minorities; were frustrated in their attempts to communicate with their professional colleagues by all means including the postal service and the Internet; and their every move was closely scrutinized to detect evidence of disloyalty. One of the first steps to restore the University of Mosul's viability must be reconnecting faculty and students with peers and colleagues inside and outside Iraq and reestablishing the University's outreach to community-based organizations. The HEAD Project is accomplishing this via several avenues.
Expansion of internet access to the administration, faculty and students
Internet access expansion can be organized in a variety of ways, from the creation of a centralized facility to one that is totally decentralized, with a mixture of the two also possible. Engineers prefer a college-based system that would be integrated into the specialized computer system in place in their College. Because of the need to accomodate sensors and other devices specifically designed for engineering functions, a centralized system would not serve their needs. A plan for upgrading Internet access can only be constructed after a user needs survey, development of alternate plans, and a study of their budgetary implications.
Information technology is necessary for national and international communications, in addition to staying current on public health and medical advances. While computer equipment, servers and Internet connectivity are being restored, a University of Mosul/MCID website is being developed which will allow for national and international faculty exchanges, the E-book program (see below), and other informational services. As time progresses, the web site will be expanded to link all participating institutions. The importance of an adequate Internet network and universal access cannot be overestimated in a country where postal and telephone services are primitive, at best.
The electronic book program is an exciting option in the search for access to books, now extremely limited in Iraq. Electronic books represent a vast social opportunity to increase access to information, books, and education, especially for disadvantaged communities. For example, with collections of cultural texts captured and offered in an easily accessible format, the histories of Iraq's diverse communities can be preserved and shared, fostering awareness and understanding of the distinct heritage of each ethnic and religious group. The economics of electronic collection delivery are powerful; the incremental cost to deliver another book as an electronic text file is extraordinarily low. A repository of scanned and publisher-supplied electronic books, as well as thousands of public domain and freely available books, is being established on a server at the university, and these ebooks will be made available to students, faculty and professionals in all of Iraq. Electronic access will be free at the university, and books printed on demand will be available at prices closely linked to local production costs.
Community outreach programs help the university reach out to the community it serves. The HEAD Project proposes to finance a variety of outreach programs including for support of work of civil society organizations, those working in public health and human rights, faculty research programs and student thesis research, and a mini-grants program to encourage academic inquiry dedicated to addressing the specific needs of disenfranchised minority groups, including Iraqi women.
Mini-grants funded by the HEAD Project will provide funds for several short-term projects. Using a review process similar to that of national and international agencies, proposals will be peer reviewed and granted expeditiously to ensure timely completion. Funded projects may include collaborations with other colleges on the University of Mosul campus, other Iraq universities, and the MCID affiliates. They may supplement activities in other HEAD Project programs. For student theses (master's program), emphasis will be placed on those research projects that promote and focus on outreach in their communities.
Training for Research Awareness in Nursing (TRAIN) is a program designed to increase the pool of nurse researchers by providing theoretical and experiential learning activities related to the research process, as well as nursing input into the health of the community and further post-graduate preparation. Academic development will be achieved in a nine-week program based on a successful US model by 1) providing intensive theoretical and practical experience with research related to racial/ethnic and gender-specific (female) health issues and concerns; 2) acquainting the student with materials and techniques necessary to facilitate entry into graduate programs such as nursing and public health, and 3) facilitating networking with faculty and students of MCID. This allows for inter-university relations between the School of Nursing and College of Medicine, for example, and it can be a pilot program for other Iraqi institutions as well as collaborations within institutional departments.
The community sustainable health outreach program (CSHOP) is proposed for the University of Mosul's Department of Community Medicine. The program, with key faculty from the Colleges of Engineering, Nursing and Medicine, will have as its goal the development of sustainable health outreach programs in sanitation. These include ensuring proper water safety (teaching techniques of hand washing, boiling water, for example), preventive health practices (proper nutrition, breastfeeding, immunizations) and educational awareness (media campaigns, religious center outreach, informational health materials). The faculty team will consist of representatives from each college with technical expertise in selected areas of public health and sanitation.
Existing NGO's will be enlisted as an immediate resource and mechanism for the University of Mosul's CSHOP projects. In addition, key informants and expert opinions will be acquired from identified "gate keepers" in the community, with the goal of developing new NGOs to provide outreach and services to various targeted communities (women, ethnic/racial minorities). The CSHOP will conduct needs assessments to determine current community health concerns and provide technical assistance/consulting, professional development workshops, health classes, health screenings and preventive services, and limited funding for outreach activities. The CSHOP advisory board will include representatives from each NGO and designated community gatekeeper and will meet on a regular basis so that communication between the University and community remains open and fluid as community health needs are met. The CSHOP personnel will consist of both graduate and professional students (medical, nursing) supervised by faculty mentors. This service will staff the CSHOP as well as serve as a training experience for students during their academic pursuits.
Expansion of inter- and intra-university relations
Finally, the HEAD Project Partners intend to support cross-university programs in public health, nursing and sanitation between the University of Mosul and other Iraqi institutions of higher education in other regions. The strengthening of inter-university ties will help reduce the isolation each university has suffered over the past decade. At the same time, MCID institutions in the US will work toward the establishment of joint degree programs with the University of Mosul in public health, nursing, and sanitation, which could include joint research, student study abroad, and long distance learning. Planning for these programs is underway in this initial year of the HEAD Project. Finally, Jackson State University intends to propose a Masters of Public Health joint degree program with the University of Mosul, once faculty assessments have been initiated and interest in the program fully determined.
Conclusion
The decline in public health services in Iraq has contributed to a rise in Iraqi morbidity and mortality. Severe acute and chronic problems need to be addressed. The universities have been unable to help in training public health personnel. Now, with the change of regime in Iraq, an important opportunity exists to chart a new course. We propose an International Partnership between a US educational consortium and the University of Mosul in northern Iraq consisting of three essential parts: recovery from past damage with return to normal university function, retooling key faculties after years of isolation, and reestablishment of the institutional viability of the University of Mosul. This resurgence of public health education is a small step, but a necessary one in restoring normalcy to the people of Mosul, and of Iraq.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
GDH and AM were involved in the design and concept of project, and were awarded the funding for the research. All authors were involved in the acquisition of results and drafting of 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
The authors would like to thank Dr. Ruth E. Little of Epidemiology Services International, Bern'Nadette Knight, Sandra Evans, LaToya G. Williams and LaToshia Bonds Robinson for assistance with manuscript preparation. The Higher Education And Development (HEAD) Project is funded by United States Agency for International Development (USAID), Cooperative Agreement No.: RAN-A-00-03-00100-00.
==== Refs
Al-Dabbagh SA The Validity of Widal Test In Diagnosing Typhoid Fever Annals of the College of Medicine, Mosul 1999 25 68 71
Al-Jawadi A Under fives nutrition five years after the blockade Annals of the College of Medicine, Mosul 1996 22 36 45
Al-Jawadi AA Hussien NY High Risk Indicators Among Under Fives in Mosul, Iraq Journal of the Arab Board of Medical Specializations 1999 1 75 79
Al-Nouri L Al-Rahim Q The effect of sanctions on children of Iraq Archives of Diseases of Children 2003 329 819
Amin NM Khoshnaw MQ Medical education and training in Iraq Lancet 2003 362 1326 14579817 10.1016/S0140-6736(03)14580-1
Aziz C Struggling to rebuild Iraq's health-care system: War sanctions and mismanagement have left the health system in shambles Lancet 2003 362 1288 9 14577435 10.1016/S0140-6736(03)14618-1
Blacker J Jones G Ali M Annual morality rates and excell deaths of children under five in Iraq, 1991–98 Population Studies 2003 57 217 226 12888415 10.1080/0032472032000097119
Dyer O Iraq's hospitals struggle to provide a service BMJ 2003 326 899 12714460 10.1136/bmj.326.7395.899
Fleck F Iraq health minister plans future Iraqi health system Bulletin of the World Health Organization 2003 81 848 9 14758416
Furber AS Johnstone P Rebuilding health care in Iraq J Epidemiol Community Health 2004 58 890 2 15483301 10.1136/jech.2003.019216
Martone G Water and health in Iraq: A nurse reports on this crucial link as a country is rebuilt Am J Nurs 2003 103 46 7 14625424
Stone R Priorities for rebuilding civilian Iraqi science Science 2004 304 943 4 15143250 10.1126/science.304.5673.943b
UNICEF At a glance: Iraq 2003 Accessed November 15, 2004.
UNICEF Iraq: Restore public health system for malnourished children 2003 Accessed November 15, 2004.
UNICEF The Situation of Children in Iraq 2003 Accessed 15 November 2004
UNICEF The State of The World's Children 2003 Accessed 15 November 2004
Voelker R Iraq targets basic health needs JAMA 2004 292 1415 6 15383496 10.1001/jama.292.12.1415
| 15801984 | PMC1087850 | CC BY | 2021-01-04 16:28:57 | no | BMC Public Health. 2005 Mar 31; 5:29 | utf-8 | BMC Public Health | 2,005 | 10.1186/1471-2458-5-29 | oa_comm |
==== Front
BMC Public HealthBMC Public Health1471-2458BioMed Central London 1471-2458-5-371583679410.1186/1471-2458-5-37Research ArticleEpidemiology of leisure-time physical activity in socio-demographic, lifestyle and psychological characteristics of men and women in Greece: the ATTICA Study Pitsavos Christos [email protected] Demosthenes B [email protected] Yannis [email protected] Christodoulos [email protected] First Cardiology Clinic, School of Medicine, University of Athens, Athens, Greece2 Department of Nutrition and Dietetics, Harokopio University, Athens, Greece2005 18 4 2005 5 37 37 12 1 2005 18 4 2005 Copyright © 2005 Pitsavos et al; licensee BioMed Central Ltd.2005Pitsavos 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 evaluate the prevalence, frequency and type of leisure-time physical activity (LTPA) among adults in Greece, as well as its relationship with socio-demographic, lifestyle and clinical characteristics of these people.
Methods
From May 2001 to December 2002 we randomly enrolled 1514 men and 1528 women, without any evidence of cardiovascular or any other chronic disease. The sampling was stratified by the age – gender distribution of (census 2001) of the greater area of Athens. Weekly energy expenditure assessed by considering frequency, duration (in minutes) and intensity of sports related physical activity during a usual week.
Results
53% of men and 48% of women were classified as physically active. Men were more likely to be active as compared to women (p < 0.05), while the lowest activity rates were observed in 40 to 49 years old participants (p < 0.01). Physically active people had higher occupation skills, were more likely to live in rural areas, to be unmarried, non smokers and they were devoted to a healthier dietary pattern, as compared to sedentary, irrespective of age and sex (all p < 0.05). In addition, the cumulative risk factors score of obesity, hypertension, hypercholesterolemia and diabetes, was inversely associated with activity status (p < 0.001). Finally, physically active men and women were less likely to report depressive symptoms (p < 0.01), after various adjustments were made.
Conclusion
Half of the studied population reported physically inactive, indicating that sedentary lifestyle becomes a serious epidemic in Greece. High occupation skills, non-smoking, devotion to a healthier dietary pattern and a better cardiovascular risk factors profile were some of the determinants of physically active people.
==== Body
Background
Several observational and clinical studies suggest that physical activity substantially reduces the risk of dying of coronary heart disease, stroke, and colon cancer [1-4]. It also helps to control weight, contributes to healthy bones, muscles, and joints, reduces falls among older adults, helps to relieve the pain of arthritis, reduces symptoms of anxiety and depression and is associated with fewer hospitalizations, physician visits, and medications [3]. Despite the proven benefits of physical activity, the National Center for Chronic Disease Prevention and Health Promotion reports that more than one half of American adults do not get enough physical activity to provide health benefits, while 25% of adults are not active at all in their leisure time [3]. Worldwide, the World Health Organization estimates that over 60% of adults are not active enough to benefit their health [4]. Moreover, physical activity declines significantly with age, it is generally higher among females, and the overall inactivity trend is worse in poor urban areas. In addition, there are racial and ethnic differences in physical activity rates, particularly among women.
Data regarding the prevalence of physical activity and its association with various characteristics of Greek men and women are lacking. Therefore, the aim of this work is to evaluate the prevalence of leisure-time physical activity (LTPA) and to investigate its association with various socio-demographic, lifestyle and behavioural characteristics of Greek adults.
Methods
The "ATTICA" study is a health and nutrition survey, carried out in the province of Attica (including 78% urban and 22% rural areas), where Athens, is a major metropolis. The sampling was random, multistage and based on the city – gender – age distribution of the province of Attica (census of 2001). The study's design anticipates enrolling only one participant per household. The main goals of the ATTICA study were to record the distribution of several blood lipids, inflammatory, oxidation, coagulation, thrombotic and clinical factors and to explore the associations between these factors with several socio-demographic, lifestyle and psychological characteristics of the participants.
Study's participants
From May 2001 to August 2002, 4056 inhabitants from the above area were randomly asked to participate in the study. Of those, 3042 agreed to participate (75% participation rate); 1514 of the participants were men (50%, 20–87 years old) and 1528 were women (50%, 20–89 years old). Between 40 – 49 years old there were 48% men and 52% women. The number of the participants was determined by power analysis and chosen to evaluate two-sided differences between the normally distributed investigated parameters and physical activity groups greater than 10%, achieving statistical power > 0.80 at < 0.05 probability level (P-value). Also, those living in institutions were excluded from the sampling. The participants had no clinical evidence of cardiovascular or any other atherosclerotic disease, as well as chronic viral infections dental problems or any type of surgery in the past week. They also exhibited no signs of cold, flu, or any acute respiratory infection.
Physical activity ascertainment
We used a translated version of a validated questionnaire [5] of weekly energy expenditure that assessed: frequency (times per week), duration (in minutes) and intensity of sports related physical activity during a usual week. Intensity was graded in qualitative terms such as: light (expended calories < 4 Kcal/min, i.e. walking slowly, cycling stationary, light stretching etc.), moderate (expended calories 4–7 Kcal/min, i.e. walking briskly, cycling outdoor, swimming moderate effort etc.) and high (expended calories >7 Kcal/min, i.e. walking briskly uphill, long distance running, cycling fast or racing, swimming fast crawl etc.). Participants who did not report any physical activities were defined as sedentary. For the rest, we calculated a combined score by multiplying the weekly frequency, duration and intensity of physical activity. Then we calculated the tertiles of this score, and physically active participants were equally classified into three groups: low physical activity (1st tertile), medium physical activity (2nd tertile) and high physical activity (3rd tertile). Therefore, 1st, 2nd and 3rd tertile included 217 men and 205 women, respectively. We have also recorded the type of exercise, i.e. resistance (i.e. any technique that uses progressive resistance to increase muscular strength) or endurance training. The presence of occupational physical activity was recorded (41% of men and 27% of women reported occupational physical activity at least one time per week, i.e. mainly manual workers like builders, mason, labourer, technicians, etc), but was not taken into account for the analysis due to difficulties in evaluation and standardization. This exclusion may confound our findings, but the large sample size and the randomised procedure for the selection of the participants can spread the subjects who reported occupational exercise equally in both groups of the study.
Although there are some concerns with self-reported physical activity, the introduced evaluation is now considered reliable, valid and has been used by many other similar studies [6].
Socio-demographic and behavioural characteristics
In addition to the physical activity status, the study's questionnaire included demographic characteristics like age, gender, family status (married, divorced, widowed), financial status (average annual income during the past three years), and occupational status as well as education level. The educational level of the participants (as a proxy of social status) was measured in years of school. Occupation was also recorded and evaluated through a ten-point scale from unskilled – hand workers (lower values) to executive – skilled workers (higher values).
Information about smoking habits was collected using a standardized questionnaire developed for the Study. Current smokers were defined as those who smoked at least one cigarette per day. Never smokers those who have never tried a cigarette in their life and former smokers were defined as those who had stopped smoking more than one year previously. In all multivariate statistical analyses cigarette smoking habits were taken into account using the pack-years (cigarette packs per day × years of smoking). However, in order to correct for the amount of nicotine containment in various types of cigarettes smoked (i.e. light, heavy, very heavy), we assigned a weight in each different type of cigarette – pack, using the 0.8 mg/cigarette as the standard.
Consumption of non-refined cereals and products, vegetables, legumes, fruits, olive oil, dairy products, fish, pulses, nuts, potatoes, eggs, sweets, poultry, red meat and meat products were measured as an average per week during the past year through a validated food – frequency questionnaire (FFQ), from the Department of Nutritional Epidemiology of Athens Medical School [7]. The frequency of consumption was then quantified approximately in terms of the number of times a month a food was consumed. Alcohol consumption was measured by daily ethanol intake, in wineglasses (100 ml and 12 gr ethanol concentration). Based on the Mediterranean – diet pyramid [8], we calculated a special diet score ranged from 0 to 55. Higher values of this score indicates adherence to the traditional Mediterranean diet, characterized by moderate consumption of fat and high monounsaturated: saturated fat ratio), while lower values indicate adherence to the "Westernized" diet.
Depressive symptomatology was assessed using a translated and validated version of the Zung Self-Rating Depression Scale (ZDRS) [9]. The ZDRS is a well known and world-widely used self-rating scale for the measurement of depression. It is a self-reporting instrument and was originally developed in order to assess depression symptoms without the bias of an administrator affecting the results. Higher scores on this scale are indicative of more severe depression [9]. ZDRS consists of 20 items that cover affective, psychological, and somatic symptoms. The patient specifies the frequency with which the symptom is experienced (that is: a little = 1, some = 2, a good part of the time = 3, or most of the time = 4). A subject with ZDRS score below 50 is considered normal, with a score of 50–59 is considered to suffer from mild depression, with score 60–69 depression is considered moderate, while with a score of 70 or above depression is considered to be severe [9]. Previous investigations indicate that the ZDRS is a valid and sensitive measure of clinical severity in depressed patients and support its continued use as a research instrument [10]. Moreover, the sensitivity of the ZDRS was found adequate, and the scale was able to significantly differentiate four severity groups classified on the basis of the global rating.
Biochemical and clinical measurements
The blood samples were collected from the antecubital vein between 8 to 10 a.m., in a sitting position after 12 hours of fasting and avoiding of alcohol. The biochemical evaluation was carried out in the same laboratory that followed the criteria of the World Health Organization Lipid Reference Laboratories. Serum cholesterol was measured using chromatographic enzymic method in a Technicon automatic analyzer RA-1000. The intra and inter-assay coefficients of variation of cholesterol levels did not exceed 5%. Participants with total serum cholesterol levels greater than 200 mg/dl or those taking lipid-lowering agents were classified as hypercholesterolemic and those with blood sugar > 125 mg/dl or the use of antidiabetic medication were classified as diabetic. Arterial blood pressure was measured at end of the physical examination with subject in sitting position for 25–30 minutes. Three measurements were abtained from the right arm (ELKA using aneroid manometric sphygmometer, Von Schlieben Co, West Germany). Patients whose average blood pressure levels were greater or equal to 140/90 mm Hg or taking antihypertensive medication were classified as hypertensives. Body mass index was calculated as weight (in kilograms) divided by standing height (in meters squared). Obesity was defined as body mass index > 29.9 kg/m2. Also, the investigators of the study recorded a detailed medical history of the participants.
Details about the aims and methods of the ATTICA study have been presented elsewhere [11].
Statistical analysis
Continuous variables are presented as mean values ± standard deviation, while categorical variables are presented as absolute and relative frequencies. Associations between categorical variables were tested by the use of contingency tables and the calculation of chi-squared test. Correlations (or partial correlations) between continuous or discrete variables were tested by the calculation of Pearson's r or Spermans's rho (for skewed variables) correlation coefficients. Comparisons between normally distributed continuous variables and categorical were performed by the use of Analysis of co-Variance (multi- way ANCOVA), after controlling for homoscedacity, and potential confounders. Kolmogorov-Smirnov criterion was used for the assessment of normality. In the case of asymmetric continuous variables the tested hypotheses were based on the calculation of the Kruskal – Wallis test. Associations between the investigated variables and physical activity levels were tested through fixed effects models, after the adjustment for several potential confounders and its interactions with physical activity status. The assumptions of linearity and homoscedacity were graphically tested by plotting fitted values against standardised residuals. Multiple logistic regression analysis estimated the odds ratio of being at a specific health condition with respect to physical activity status. Deviance residuals evaluated models' goodness of fit. All reported P-values are based on two-sided tests and compared to a significance level of 5%. However, due to multiple significance comparisons we used the Bonferronni correction in order to account for the increase in Type I error. SPSS 11.0.5 software (SPSS Inc. 2002, Texas, USA) was used for all the statistical calculations.
Results
Eight hundred and two (53%) of men and 733 (48%) of women were classified as physically active. Table 1 presents the age distribution of physical activity in study's participants. Men were more likely to be physically active as compared to women, across all age groups (p = 0.001). The lowest physical activity rates were observed between 40 to 49 years old men and women (p = 0.01). Physically active men were devoted to more intense activities as compared to women (6.96 ± 1.2 vs. 6.23 ± 1.1 kcal/min, p = 0.02). In addition, men used to exercise 3.2 ± 1.7 times per week (for 85 ± 19 min/time) and women 2.7 ± 1.3 times per week (for 42 ± 21 min/time) (p = 0.03). An inverse relationship was observed between age and frequency, intention and time devoted to physical activities (partial r = -0.23, p = 0.02, r = -0.34, p = 0.001, and r = -0.21, p = 0.0001, respectively).
Table 1 Distribution of physical activity by age and sex
Physically active
Age group (men / women) Men (n = 1514) Women (n = 1528)
20 – 29 y 52.2% 48.0%
30 – 39 y 43.1% 41.0%
40 – 49 y 37.1% 35.1%
50 – 59 y 47.3% 37.5%
> 60 y 45.4% 38.2%
Furthermore, from physically active participants, 5% of men and 2% of women reported only resistance training, while 7% of men and 12% of women reported only endurance training. Thirty five percent of men and 25% of women reported that they were engaged in organised sports (sport clubs, or football, basketball and volleyball teams, etc.). When we compared participants in organised sports with non-participants, those who participated in organised sports were more often persistent exercisers as compared to those who did not (odds ratio = 11.6 for men, 95% confidence interval 7.2 to 15.1 and 7.9 for women, 95% confidence interval 5.5 to 11.0).
Physical activity and socio-demographic characteristics
Table 2 presents social characteristics of the participants by physical activity status. In particular, a positive association was observed between physical activity and occupation skills (unskilled to skilled) in men and in women (both p < 0.05). Moreover, people living in rural areas were more likely to be physically active as compared to people living in urban areas (55% vs. 46%, p = 0.02). Never married participants were more likely to be physically active as compared to married or divorced/widowed (57% vs. 44%, p < 0.001), irrespective of age and sex. No statistically significant associations were found between physical activity levels and other social status indices (education level and annual income).
Table 2 Socio – demographic characteristics by physical activity status and sex
Physical activity status
Men Sedentary 1sttertile 2ndtertile 3rdtertile P
Education (years of school) 12 ± 4 12 ± 4 13 ± 4 12 ± 5 0.18
Annual income (× 1000 Euros) 18,2 ± 3 16,9 ± 3 16,2 ± 5 15,8 ± 5 0.23
Occupational status⊥ (0 – 10) 5,1 ± 2 6,1 ± 3* 6,2 ± 3* 7,0 ± 5** 0.02
Women Sedentary 1sttertile 2ndtertile 3rdtertile P
Education (years of school) 11 ± 4 11 ± 3 13 ± 5 12 ± 2 0.36
Annual income (× 1000 Euros) 15,4 ± 7 15,8 ± 7 17,7 ± 6 14,2 ± 6 0.35
Occupational status⊥ (0 – 10) 4,9 ± 2 6,5 ± 3** 6,9 ± 3* 7,8 ± 3** 0.01
Variables are expressed as mean value ± SD
⊥Occupational status is measured in a ten-unit scale where lower values imply unskilled professions and higher values skilled.
P-values of the last column derived from stratified ANCOVA tests, after controlling for age and denote the significance of the association between physical activity status and education, income and occupational status of the participants.
* P < 0.05 and ** P < 0.01 Bonferronni corrected probability values for the comparisons between physical activity groups and sedentary participants.
Physical activity, lifestyle and behavioural characteristics
Table 3 illustrates the associations between physical activity and various lifestyle and behavioural characteristics of the participants. Particularly, we observed that physically active participants were less frequent to be active smokers as compared to sedentary. An inverse relationship was found between physical activity and dietary score, which implies that physically active participants showed greater adherence to a healthier dietary pattern, i.e. the Mediterranean diet (higher values of the diet score). In addition, active men and women consumed lower quantities of alcoholic beverages as compared to sedentary. Multivariate analysis (that used physical activity index as an independent variable and diet score as a dependent variable) showed that all the previous associations were independent from age, sex, education and financial status of the participants, as well as region of living (all p's < 0.05).
Table 3 Behavioural characteristics by physical activity status and sex
Physical activity status
Men Sedentary 1sttertile 2ndtertile 3rdtertile P
Current smoking 58% 54% 54% 48%* 0.01
Alcohol intake (ml/d) 280 ± 40 250 ± 55* 190 ± 95** 180 ± 90** 0.001
Diet score (0 – 55) 28 ± 10 32 ± 25* 38 ± 23** 41 ± 22** 0.001
ZDRS (0 – 100) 48 ± 23 41 ± 23 38 ± 16 32 ± 17 < 0.001
Women Sedentary 1sttertile 2ndtertile 3rdtertile
Current smoking 49% 44% 36%** 29%** 0.041
Alcohol intake (ml/d) 120 ± 40 110 ± 55 90 ± 55* 85 ± 50** 0.001
Diet score (0 – 55) 29 ± 11 36 ± 22* 42 ± 19** 43 ± 21** < 0.001
ZDRS (0 – 100) 52 ± 21 44 ± 13* 39 ± 13** 34 ± 19** 0.001
Variables are expressed as mean value ± SD or relative frequencies.
P-values of the last column derived from stratified ANCOVA tests, after controlling for age and denote the significance of the association between physical activity status and smoking, dietary habits and depression status of the participants.
* P < 0.05 and ** P < 0.01 Bonferronni corrected probability values for the comparisons between physical activity groups and sedentary participants.
An inverse relationship was observed between physical activity status and ZDRS, which indicates that men and women who used to exercise were less likely to have depressive symptoms as compared to sedentary (standardised beta-coefficient = -0.34, p = < 0.001), irrespective from the effect of various potential confounding factors like age, sex, educational, financial and occupational status. Moreover, participants in the highest tertile of physical activity had 41% lower odds (95% confidence interval 0.35 to 0.9) of having major depressive syndromes (i.e. ZDRS > 60), after controlling for the previous potential confounding factors.
Physical activity and clinical characteristics
An inverse relationship was observed between body mass index and physical activity score (partial r = -0.46, p < 0.001), after adjusting for age and sex. Participants in the highest tertile of physical activity had 48% lower odds of being obese (95% confidence interval 0.42 to 0.64), after adjusting for age, sex, smoking habits, blood pressure, total cholesterol and glucose levels. The prevalence of hypertension was 38% in men and 24% in women. Physical activity was also associated with lower odds of being hypertensive (odds ratio = 0.76, 95% confidence interval 0.59 to 0.98), irrespective of age, sex, smoking habits, total cholesterol and glucose levels. Moreover, compared to sedentary, people in the highest tertile of physical activity had on average 15 mm Hg lower systolic (p = 0.02) and 8 mmHg lower diastolic (p = 0.04) blood pressure levels, after adjusting for age, sex, body mass index and diet score. We have also observed that 54% of men and 60% of women had total serum cholesterol levels <200 mg/dl. No association was observed between physical activity and presence of hypercholesterolemia (p = 0.76).
Afterwards we calculated the cumulative distribution of four major cardiovascular risk factors, i.e. obesity, hypertension, hypercholesterolemia, and diabetes. We observed an inverse relationship between physical activity index and the cumulative score of risk factors (partial rho = -0.33, p < 0.001), after adjusting for age, sex and smoking habits. Figure 1 illustrates the inverse association between activity status and risk factors score in men and women, respectively. Additionally, we may observed that even moderate amounts of physical activity (i.e. second tertile) have a beneficial effect on the cumulative distribution of these cardiovascular risk factors.
Figure 1 Cumulative distribution of cardiovascular risk factors score by physical activity status.
Discussion
In this observational study of 3042 adult people from Attica region, Greece, we found that, roughly, one out of two men and women were physically active. Therefore, it could be speculated that, about, 4 million Greek adults are sedentary. In addition, men were more likely to be physically active as compared to women, while the lowest physical activity rates were observed in middle-aged participants (i.e. 40 to 49 years old). A description of the profile of the participants showed that physically active people had higher occupation skills, were more likely to live in rural areas, to be unmarried, to be non smokers and they were devoted to a healthier dietary pattern, as compared to sedentary, irrespective of age and sex. In addition, the cumulative score of four major cardiovascular risk factors, i.e. obesity, hypertension, hypercholesterolemia, and diabetes, was inversely associated with the applied physical activity index. Finally, physically active men and women were less likely to report depressive symptoms.
Physical activity and socio-demographic characteristics
Physical activity has been related with social status in some, but not all studies. For example, in a population-based sample from Copenhagen [12], the investigators observed that subjects with the lowest level of education (a proxy of social status) were most frequently physically inactive, as well as heavy smokers, heavy drinkers and obese. In another community health survey of Adelaide [13] the investigators reported that low physical activity status was strongly associated with low education and reduced mobility, number of social connections and degree of satisfaction with recreation facilities. On the other hand, Tammelin et al. [14] reported that infrequent participation at young age was associated with physical inactivity at older age, while low social class of the childhood family was associated with physical inactivity in adolescence but not with inactivity at older age. In our study, we observed a positive association of physical activity with some (occupation skills), but not all social status indices (education level and annual income). This may reflect various cultural characteristics of the participants, but it could not be further investigated by the present work.
We also found that participants who were devoted in organised sports and members in sports clubs were more frequently persistent exercisers as compared to those who did not. This is in accord to the reports from several other investigators, which observed that participation in organised sports was associated with the stability of physical activity [15-17]. Thus, we may suggest that the development of programs to promote participation in organised sports it could be a worthwhile investment for their health and well-being.
We observed that men were more likely to be physically active as compared to women, irrespective of age. These gender differences are pronounced in most countries. For example, it has been reported that the proportions of girls exercising actively is approximately half of that of boys among 15-year-olds in Greenland, Lithuania and Greece [18]. The issue of gender differences is also relevant, when we consider physical activity trends from adolescents to adults. In addition, we found that the gender difference in physical fitness was even higher than the difference in physical activity status, since men were more commonly persistently fit. The aforementioned gender differences have been reported in several studies [19-21], and may reflect behavioural and work-related differences, as well as differences in family responsibilities between sexes that could not investigated in this study. However, it should be noted that Barnekow-Bergkvist et al. [22] reported that, although more men than women participated in sports activities at the age of 16, there was no significant difference at the age of 34.
Regarding marital status we observed that never married participants were more likely to be physically active as compared to married or divorced/widowed, independent from age and sex of the participants. There are very sparse data in the literature concerning the relationship between marital status and physical activity. In a population-based study of men and women from the US [23], the investigators observed that the change from a married to a single state did not affect physical activity relative to remaining married, whilst the transition from a single to a married state resulted in significant positive changes in physical activity relative to remaining single. Thus, they concluded that marriage might potentially set the stage for natural changes in physical activity status. Family, work and other social responsibilities, as well as cultural difference may be responsible for the divergence of the results between the present and other studies.
Physical activity and depression
Depression constitutes a worldwide health problem, because is associated with functional impairment in normal life and at work, poor compliance with medical therapy and lifestyle risk factors interventions [24]. Much discussion has been made regarding the association between depression and physical activity status [25-31]. Few population-based studies reported that physical activity was not associated with depression [25-27], while other studies reported a protective effect [28-33]. For example, the results from the first National Health and Nutrition Examination Survey (NHANES) reported that the risk of depressive symptoms at follow-up were two-fold for women with little or no physical activity compared to women with much or moderate recreational physical activity [28]. Although it is difficult to compare findings from prospective with cross-sectional studies, we have found similar results with the NHANES since participants in the lowest tertile of physical activity had 1.6-times higher odds of reporting major depressive syndromes. Whether physical activity is associated with depression it is difficult to answered by the present study. Although we provided clear scientific evidence, further research is needed to state cause – effect relationships and to clarify the mechanisms underlying the association between physical activity and depression.
Physical activity and clinical characteristics
Furthermore, we confirmed previous reports concerning an inverse association of physical activity with the prevalence of obesity and hypertension [34-36], while no relationship was observed in our, as well as several other studies, regarding physical activity and hypercholesterolemia [37,38]. We also found an inverse relationship between presence of leisure time physical activities and the cumulative distribution of four major cardiovascular risk factors (i.e. obesity, hypertension, hypercholesterolemia, and diabetes). We expanded the previous findings by showing that even moderate amounts of physical activity were associated with the frequency of these cardiovascular risk factors levels. It is becoming more apparent that most of the health benefits at a minimal risk are derived from low to moderate intensity physical activities [39].
Limitations
This is a cross-sectional study that cannot provide causal relationships, but only state hypotheses for future research. Occupational physical activity which was not taken into account in the present study, and misreporting of physical activity status due to self-reports may confound, at least in part, the strength of the observed relationships. Another limitation is that LTPA has been related to a healthier lifestyle habits and consequently to a better health status. For example, adoption of a healthier dietary pattern or reduced smoking habits may be more common among individuals who were devoted to leisure time exercise. Thus, the benefits from LTPA on the prevalence of risk factors (Figure 1) cannot provide causal relationships and may be miss-interpreted at population level. Finally, although we took into account dietary and smoking habits of the participants the influence of the potential confounding effect of these factors cannot entirely be excluded.
Conclusion
A considerable proportion of the population studied reported physically inactive, indicating that sedentary lifestyle becomes a serious epidemic in Greece, and especially in women. We have also illustrated the socio-demographic profile of people that are physically active and we revealed the relationships between physical activity and various lifestyle behaviors, the presence of depressive symptoms and several metabolic disorders associated with cardiovascular risk. Despite the limitations of the cross-sectional design of the present study, our findings may carry an important public health message, i.e. to inform the community and especially middle-aged, low socio-economic status people living in urban areas, about the benefits derived from physical activity in order to prevent future cardiovascular events and other psychological disorders.
Competing interests
The author(s) declare that they have no competing interests
Authors' contributions
CP = drafted the paper, was the principal investigator, and had the concept, and design of the study, DBP = wrote the paper, was the principal investigator, and had the concept and design of the study, performed the data analysis, YL = drafted the paper, participated in the clinical evaluation of the participants and CS = drafted the paper.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The ATTICA Study is funded by research grants from the Hellenic Society of Cardiology (grant – 1, 2002).
The authors would like to thank the field investigators of "ATTICA" study: John Skoumas (principal field investigator), Natasa Katinioti (physical examination), Spiros Vellas (physical examination), Efi Tsetsekou (physical/psychological evaluation), Dina Masoura (physical examination), Lambros Papadimitriou (physical examination), as well as the technical team: Marina Toutouza (biochemical analysis), Carmen Vasiliadou (genetic analysis), Manolis Kambaxis (nutritional evaluation), Konstadina Paliou (nutritional evaluation), Chrysoula Tselika (biochemical evaluation), Sia Poulopoulou (biochemical evaluation) and Maria Toutouza (database management).
==== Refs
Wei M Kampert JB Barlow CE Nichaman MZ Paffenbarger RS JrBlair SN Relationship between low cardiorespiratory fitness and mortality in normal-weight, overweight, and obese men JAMA 1999 282 1547 1553 10546694 10.1001/jama.282.16.1547
Sesso H Paffenbarger R Lee IM Physical Activity and Coronary Heart Disease in Men. The Harvard Alumni Health Study Circulation 2000 102 975 10961960
US Department of Health and Human Services Physical Activity and Health: A Report of the Surgeon General Atlanta, Ga: US Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion 1996
Annual Global Move for Health Initiative; A Concept Paper: World Health Organization; WHO/NMH/NPH/PAH/031
Pate RR Pratt M Blair SN Haskell WL Macera CA Bouchard C Buchner D Ettinger W Heath GW King AC Physical activity and public health. A recommendation from the Centers for Disease Control and Prevention and the American College of Sports Medicine JAMA 1995 273 402 407 7823386 10.1001/jama.273.5.402
Taylor HL Jacobs DR JrSchucker B A questionnaire for the assessment of leisure time physical activities J Chronic Dis 1978 31 741 755 748370 10.1016/0021-9681(78)90058-9
Katsouyanni K Rimm EB Gnardellis C Trichopoulos D Polychronopoulos E Trichopoulou A Reproducibility and relative validity of an extensive semi-quantitative food frequency questionnaire using dietary records and biochemical markers among Greek schoolteachers Int J Epidemiol 1997 26 S118 S127 9126540 10.1093/ije/26.suppl_1.S118
Willett WC Sacks F Trichopoulou A Drescher G Ferro-Luzzi A Helsing E Trichopoulos D Mediterranean diet pyramid: a cultural model for healthy eating Am J Clin Nutr 1995 61 1402S 1406S 7754995
Zung W A Self-Rating Depression Scale Archives of General Psychiatry 1965 12 63 70 14221692
Biggs JT Wylie LT Ziegler VE Validity of the Zung Self-rating Depression Scale Br J Psychiatry 1978 132 381 385 638392
Pitsavos C Panagiotakos DB Chrysohoou C Stefanadis C Epidemiology of Cardiovascular risk factors in Greece; aims, design and baseline characteristics of the ATTICA study BMC Public Health 2003 3 1 9 12513697
Schnohr C HOjbjerre L Riegels M Ledet L Larsen T Schultz-Larsen K Petersen L Prescott E GrOnbaek M Does educational level influence the effects of smoking, alcohol, physical activity, and obesity on mortality? A prospective population study Scand J Public Health 2004 32 250 256 15370764 10.1080/14034940310019489
MacDougall C Cooke R Owen N Willson K Bauman A Relating physical activity to health status, social connections and community facilities Aust N Z J Public Health 1997 21 631 637 9470271
Tammelin T Nayha S Laitinen J Rintamaki H Jarvelin MR Physical activity and social status in adolescence as predictors of physical inactivity in adulthood Prev Med 2003 37 375 381 14507496 10.1016/S0091-7435(03)00162-2
Aarnio M Winter T Peltonen J Kujala UM Kaprio J Stability of leisure-time physical activity during adolescence – a longitudinal study among 16-, 17- and 18-year-old Finnish youth Scand J Med Sci Sports 2002 12 179 185 12135451 10.1034/j.1600-0838.2002.00250.x
Telama R Yang X Laakso L Viikari J Physical activity in childhood and adolescence as predictor of physical activity in young adulthood Am J Prev Med 1997 13 317 323 9236971
Telama R Leskinen E Yang X Stability of habitual physical activity and sport participation: a longitudinal tracking study Scand J Med Sci Sports 1996 6 371 378 9046549
Hickman M Roberst C de Matos C Currie Exercise and leisure time activities Health and health behaviour in school-aged children A WHO Cross-National Study (HBSC) International Report Health Policy for Children and Adolescents (HEPCA) Series No, 1 2000 Copenhagen: WHO 73 83
Sallis JF Zakarian JM Hovell MF Hofstetter CR Ethnic, socioeconomic, and sex differences in physical activity among adolescents J Clin Epidemiol 1996 49 125 134 8606313 10.1016/0895-4356(95)00514-5
Caspersen CJ Pereira MA Curran KM Changes in physical activity patterns in the United States, by sex and cross-sectional age Med Sci Sports Exerc 2000 32 1601 1609 10994912 10.1097/00005768-200009000-00013
Hicks MK Wiggins MS Crist RW Moode FM Sex differences in grade three students' attitudes toward physical activity Percept Mot Skills 2001 93 97 102 11693715
Barnekow-Bergkvist M Hedberg G Janlert U Jansson E Prediction of physical fitness and physical activity level in adulthood by physical performance and physical activity in adolescence. An 18-year follow-up study Scand J Med Sci Sports 1998 8 299 308 9809389
King AC Kiernan M Ahn DK Wilcox S The effects of marital transitions on changes in physical activity: results from a 10-year community study Ann Behav Med 1998 20 64 69 9989310
Carney RM Freedland KE Miller GE Jaffe AS Depression as a risk factor for cardiac mortality and morbidity. A review of potential mechanisms J Psychos Research 2002 53 897 902 10.1016/S0022-3999(02)00311-2
Lennox SS Bedell JR Stone AA The effect of exercise on normal mood J Psychosom Res 1990 34 629 36 2290135 10.1016/0022-3999(90)90106-E
Weyerer S Physical inactivity and depression in the community: evidence from the upper Bavarian field study Int J Sports Med 1992 13 492 496 1428382
Cooper-Patrick L Ford DE Mead LA Exercise and depression in midlife: a prospective study Am J Public Health 1997 87 670 673 9146452
Farmer ME Locke BZ Moscicki EK Physical activity and depressive symptoms: the NHANES I Epidemiologic Follow-up Study Am J Epidemiol 1988 128 1340 51 3264110
Camacho TC Roberts RE Lazarus NB Physical activity and depression: evidence from the Alameda County Study Am J Epidemiol 1991 134 220 31 1862805
Paffenbarger RS JrLee IM Leung R Physical activity and personal characteristics associated with depression and suicide in American college men Acta Psychiatr Scand Suppl 1994 377 16 22 8053361
Martinsen EW Physical activity and depression: clinical experience Acta Psychiatr Scand Suppl 1994 377 23 27 8053362
Dunn AL Trivedi MH O'Neal HA Physical activity dose-response effects on outcomes of depression and anxiety Med Sci Sports Exerc 2001 33 S587 97 11427783
Strawbridge WJ Deleger S Roberts RE Kaplan GA Physical activity reduces the risk of subsequent depression for older adults Am J Epidemiol 2002 156 328 334 12181102 10.1093/aje/kwf047
US Department of Health and Human Services Physical Activity and Health: A Report of the Surgeon General Atlanta, Ga: US Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion 1996
Executive Summary of the Third Report of The National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation and Treatment of High Blood Cholesterol In Adults (Adult Treatment Panel III) JAMA 2001 285 2486 2497 11368702 10.1001/jama.285.19.2486
Rennie KL McCarthy N Yazdgerdi S Marmot M Brunner E Association of the metabolic syndrome with both vigorous and moderate physical activity Int J Epidemiol 2003 32 600 606 12913036 10.1093/ije/dyg179
Bijnen FC Feskens EJ Caspersen CJ Physical activity and cardiovascular risk factors among elderly men in Finland, Italy and the Netherlands Am J Epidemiol 1996 143 553 561 8610672
Pols MA Peeters PH Twisk JW Physical activity and cardiovascular risk disease, risk profile, in women Am J Epidemiol 1997 146 322 328 9270411
Whaley MH Kampert JB Kohl HW Blair SN Physical Fitness and Clustering of Risk Factors Associated with the Metabolic Syndrome Med Sci Sports Exerc 1999 31 287 293 10063819 10.1097/00005768-199902000-00013
| 15836794 | PMC1087851 | CC BY | 2021-01-04 16:28:56 | no | BMC Public Health. 2005 Apr 18; 5:37 | utf-8 | BMC Public Health | 2,005 | 10.1186/1471-2458-5-37 | oa_comm |
==== Front
BMC Public HealthBMC Public Health1471-2458BioMed Central London 1471-2458-5-381583678410.1186/1471-2458-5-38Study ProtocolContextualizing and assessing the social capital of seniors in congregate housing residences: study design and methods Moore Spencer [email protected] Alan [email protected] Valerie [email protected] Therese [email protected] Carrie [email protected] Centre for Health & Policy Studies, Dept. of Community Health Sciences, University of Calgary, 3330 Hospital Dr. NW, Calgary, AB T2N 4T1, Canada2 Department of Sociology, 2500 University Dr. NW, University of Calgary, Calgary, AB T2N 1N4, Canada3 VicHealth Centre for the Promotion of Mental Health and Social Wellbeing, School of Population Health, University of Melbourne, level 5, 207 Bouverie St., Carlton, 3052, Australia4 Calgary Health Region, Healthy Living, SE Community Portfolio, Mental Health, NE Community Portfolio, Calgary, AB Canada2005 18 4 2005 5 38 38 25 2 2005 18 4 2005 Copyright © 2005 Moore et al; licensee BioMed Central Ltd.2005Moore 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
This article discusses the study design and methods used to contextualize and assess the social capital of seniors living in congregate housing residences in Calgary, Alberta. The project is being funded as a pilot project under the Institute of Aging, Canadian Institutes for Health Research.
Design/Methods
Working with seniors living in 5 congregate housing residencies in Calgary, the project uses a mixed method approach to develop grounded measures of the social capital of seniors. The project integrates both qualitative and quantitative methods in a 3-phase research design: 1) qualitative, 2) quantitative, and 3) qualitative. Phase 1 uses gender-specific focus groups; phase 2 involves the administration of individual surveys that include a social network module; and phase 3 uses anamolous-case interviews. Not only does the study design allow us to develop grounded measures of social capital but it also permits us to test how well the three methods work separately, and how well they fit together to achieve project goals.
This article describes the selection of the study population, the multiple methods used in the research and a brief discussion of our conceptualization and measurement of social capital.
==== Body
Background
Research in gerontology has long demonstrated the importance of social relationships and support for the health and well-being of seniors [1-4]. More recent research in the health sciences has demonstrated the effects of place or spatial context on individual health [5,6]. The current project unites the ideas of the social-network and spatial contexts to examine the effects of both on the health of seniors living in residential housing arrangements. The idea of the social-network context proposes that seniors are embedded in a structure of inter-personal ties and that this structure can influence their behavior and access to valued resources; the idea of the spatial context suggests that seniors reside in certain locations and have certain opportunity structures available to them as a result of this location. The effects of the two contexts on health may not be not mutually exclusive. Our long-term research program, which the pilot initiates, aims to examine how residential environments and social networks may act independently to influence the health and aging experiences of seniors and how they may act on one another to influence opportunity structures in residential environments and the shape of seniors' social networks.
Using a mixed method approach moving along a qualitative – quantitative – qualitative sequence, the pilot project initiates our research program in the social environments of aging. As a pilot project on social capital and seniors, the research will be concerned first of all with feasibility issues related to working with this population of seniors: 1) Are seniors resident in congregate housing capable and willing to participate in the different phases of the study? 2) How acceptable and feasible is it to obtain data on all aspects of a senior's networks both inside and outside congregate housing residences? (3) What will the participation and response rates be for follow-up anomalous case interviews? In addition to piloting our work with seniors in Calgary housing residencies, the project's mixed method approach will allow us to (1) develop contextualized understandings of social capital that are tailored to seniors in congregate housing, (2) measure empirically the social capital among seniors living in different seniors' housing facilities, and (3) examine how our explanatory models can be improved through follow-up anomalous case studies.
While there is a considerable gerontological literature on community-dwelling seniors, little research has been devoted to the social ties of the institutionalized elderly whose experiences can differ considerably from those of community-based seniors [7]. By focusing research on the social capital and aging experiences of seniors resident in non-profit and publicly subsidized housing, the pilot research targets a unique population of seniors who could be considered liminal to these two other populations. Such seniors are no longer community-dwelling per se but they are still expected to be independent with only minimal support; they thus access resources and maintain social ties both inside and outside their residence.
Why social capital?
Broadly speaking, there are two approaches to the concept of social capital: the communitarian and network. Perhaps, most evident in the work of the political scientist, Robert Putnam, and the public health research of Ichiro Kawachi, the communitarian approach to social capital emphasizes the shared values and norms of a group or a community [5,8,9]. As defined by Putnam, social capital refers to "the features of social organization, such as networks, norms, and trust that facilitate coordination and cooperation for mutual benefit [10]." Common to the communitarian approach is a vision of social capital as a common collective asset or public good equally available to all members of a group or community. In practice, the measurement of social capital becomes equated with levels of trust, reciprocity, or civic participation [8].
The network approach, on the other hand, views social capital as consisting of the resources to which individuals or groups have access through their own social networks: "social capital is the sum of resources, actual or virtual, that accrue to an individual or group by virtue of possessing a durable network of more or less institutionalized relationships of mutual acquaintance and recognition [11]." As evidenced primarily in the work of anthropologists and sociologists, the network approach to social capital emphasizes to varying degrees two elements: (1) the structure of an individual's or a group's network and (2) the resources to which individuals or groups have potential access given the structure and composition of those networks [8,12-15].
Despite the promise that social capital holds for health promotion, interventions seeking to improve social capital have been slow to develop. One possible reason for this is that social capital's operationalization and measurement in the health sciences has been inconsistent and controversial [16]. The measurement of social capital in the health sciences has commonly relied on readily available secondary data on trust, reciprocity, or civic participation. Individual-level responses are aggregated to the neighborhood, community, state, or national level to determine that unit's level of social capital. What appears lacking in the health sciences research are (1) measures of social capital that are tailored to the populations which they address and (2) measures of social capital that derive from network analysis methods. The project, although concerned with feasibility issues, is designed to improve overall research in the area of social capital and health by addressing these two weaknesses.
Methods/Design
Sample
The pilot study will be conducted in five of sixty-nine public and non-profit congregate housing residences in Calgary chosen in partnership with the Calgary Health Region. Inclusion criteria for sites will involve (1) the residence having a seniors resource nurse (SRN) who visits regularly and (2) the SRN having a strong working relationship with the residents and management of the residences. Based on these criteria along with recommendations from SRNs, residence managers will be contacted and asked if they would agree to participate. Since it is a one-year pilot project, the number of sites is limited to five.
Phase 1: focus groups
The first phase of the research will consist of gender-specific focus groups. Two gender-specific focus groups will take place in each facility. The decision to stratify according to gender is based on findings suggesting the differential distribution of social capital between genders [17,18]. In addition, preliminary field research has shown a majority of women at these locations. A stratified sample will allow us to ensure an adequate number of men.
The sampling frame for focus group participants will be the complete list of suite numbers of each facility. To generate a stratified random sample, suite numbers for each facility will be divided according to male or female resident to generate a stratified random sample. For suites containing couples, each resident will be listed separately. Random samples help reduce the biases that can result from using sign-up sheets and non-random methods [19]. Based on the list of randomly selected men and women, we will invite individuals from the top of the list until 15 – 20 residents for each group have agreed to participate or until we have exhausted the list.
The focus group script will consist of questions falling into three broad categories: 1) the meaning of social networks and social relations, 2) the identification and importance of social resources, and 3) perceptions of intra- and extra-facility trust and reciprocity. Focus group discussion will last for about 20 minutes in each category with the overall focus group lasting approximately 60 to 90 minutes. Discussions will be taped and transcribed in full.
Using content analysis, the focus group transcripts will first of all be examined to identify the key resources, individuals, and community-related events that seniors in congregate housing residences find important to them. These findings will be integrated into the resource generator module of the individual questionnaires. In addition, semantic and thematic analyses of the focus group transcripts will be undertaken to understand the meaning and importance of various types of social ties present in the residences, the levels of social support for seniors available, levels of community involvement, and other topics and themes that emerge from focus group discussions.
Phase 2: individual questionnaires
During the second phase of data collection, all seniors residing in the five facilities will be invited to complete a face-to-face interview. The interview will consist of six components containing questions about respondents': (1) sociodemographic characteristics (e.g. income and education), (2) sense of mastery, (3) move into the congregate housing residence, (4) social capital as measured from a communitarian approach, (5) social network and the resources that flow through this network, and (6) health status and health service utilization. The health status component will consist primarily of the Short Form 36 (SF-36) [20].
The two components of the survey on social capital – the communitarian and network modules – will allow us to explore the potential advantages and disadvantages of each approach for understanding the social capital of seniors and its effects on their health and well-being. The communitarian-approach module will generate data on trust, reciprocity, and civic participation. The General Social Survey (GSS) question used by Kawachi et al. "Do you think most people would take advantage of you if they got a chance?" [5] will be used to assess perceptions of trust. Following Kawachi, reciprocity will be assessed using the percentage who agree with the statement that "most people try to be helpful." The questions on trust and reciprocity will be supplemented by adding the phrase "in this residential home" to situate the measures of trust and reciprocity in that particular residential environment. Civic participation will be assessed by asking respondents about the types of associations to which they belong and their degree of participation in those associations. This question will be supplemented by an additional one asking respondents about their participation in associations or events in their residential facility. Following the coding and analysis of the focus group sessions, additional questions based on that analysis will be added to this component of the survey.
The network module of the questionnaire will consist of two instruments: 1) name generator/name interpreter sequences and 2) a resource generator. Name generators are questions that ask respondents to identify members of their social networks (alters). Different name generators tap different sectors of respondents' networks. To tap network members inside and outside the housing residence, we will use 4 name-generator questions: (1) who have you discussed important matters with in the last 6 months?, (2) who have you socialized with in the last 6 months?, (3) who do you feel attached to?, and (4) who do you know well enough to call up and talk with on the phone but who you don't know really well? Name interpreter questions are follow-up questions that generate information about 1) the characteristics of the respondent's alters (e.g., age, gender, marital status), 2) the nature of the respondent-alter relationship (e.g., role relationship, how often they see each other, how long they have known each other) and 3) the nature of the alter-alter relationships (e.g., do alters know each other).
Resource generators do not elicit the names of network members. Instead they ask respondents about access to specified resources through individuals whom they know in some capacity [21,22]. The resources that will be included in the resource generator will be identified through the phase one qualitative research and analysis. Resources that seniors may find important include "access to a car" or "knowing someone who can get them to a grocery store." Follow-up questions will ask whether the resources are accessed through family members, friends, or acquaintances. Resource generators have only recently been posited as an accurate instrument for assessing the social capital of individuals [21,22] and have yet to be tailored to senior populations.
Phase 3: anomalous-case interviews
During the third phase of data collection, a listing of anomalous cases will be created for each facility. We will use regression analysis to identify statistical outliers in the regression models used to test our hypotheses about social capital and its effects on the health of seniors. Statistical outliers might be identified on the following bases: their residual values, deviance from their predicted values, and influence over the regression models themselves [23]. For example, anomalous cases may be identified on the basis of their statistical deviance from the hypothesized social capital – health association. If certain individuals with low social capital maintain high reported health, we would like to know why. Were survey items valid? Were they reliable? Or, are there important features of social network and residential contexts that are currently being missed in the literature on social capital and health.
Once the list of individuals who represent anomalous cases has been created for each facility, a random sample of individuals will be taken from this list. Semi-structured interviews will be conducted with those selected and who agree to participate. While the specific questions to be included will only be developed after the analysis of phase one and phase two data, interview questions might center around participation in intra- or extra-residential events or ties to individuals who did not appear in the network questions, or their access to resources in Calgary. Follow-up interviews with anomalous cases in these areas will allow us to build theory, improve measures, and test our measurement instruments [23].
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
All authors participated in the design of the study. SM coordinated the study and drafted the manuscript. All authors read drafts of the manuscript, made comments and suggestions, and approved the final version.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We would like to thank all the seniors who chose to participate in the study as well as the managers of those residences in which we are conducting the study. We would also like to thank the Calgary Health Region, the Healthy Communities portfolio, and the Senior Resource Nurses (SRN) for their support.
We would also like to thank the Canadian Institutes for Health Research, Institute of Aging. The research for this pilot project was funded under the Institute of Aging, Canadian Institutes for Health Research. Drs. Moore and Shiell would also like to thank the Alberta Heritage Foundation for Medical Research.
==== Refs
Rowe JW Kahn RL Successful Aging 1998 New York: Pantheon
Bosworth H Schaie K The relationships of social environment, social networks, and health outcomes in the Seattle Longitudinal Study: Two analytic approaches J Gerontol B Psychol Sci Soc Sci 1997 52 P197 P205 9310088
Krause N Anticipated support, received support and economic stress among older adults J Gerontol B Psychol Sci Soc Sci 1997 52 P284 P293 9403517
Seeman T Berkman L Kohout F Lacroix A Glynn R Blazer D Intercommunity variations in the association between social ties and mortality in the elderly Ann Epidemiol 1993 3 325 335 8275207
Kawachi I Kennedy BP Lochner K Prothrow-Stith D Social capital, income inequality, and mortality Am J Public Health 1997 87 1491 8 9314802
Macintyre S Ellaway A Cummis S Place effects on health: how can we conceptualize, operationalize and measure them? Soc Sci Med 2002 55 125 139 12137182
Hirdes J Scott K Social relations in a chronic care hospital: a whole network study of patients, family, and employees Soc Networks 1998 20 119 133
Portes A Social Capital: Its Origins and Applications in Modern Sociology Annu Rev Sociol 1998 24 1 24
Navarro V A critique of social capital Int J Health Serv 2002 32 423 432 12211285
Putnam R Making Democracy Work 1993 Princeton: Princeton University Press
Bourdieu P Wacquant LD The Purpose of Reflexive Sociology An Invitation to Reflexive Sociology 1992 Chicago: The University of Chicago Press
Lin N Social networks and Status Attainment Ann Rev Sociol 1999 25 467 487
Erickson B Culture, Class, and Connections Am J Sociol 1996 102 217 251
Flap H Völker B eds The Creation and Returns of Social Capital 2004 London: Routledge
Haines V Beggs J Hurlbert J Levy J, Pescodillo B Exploring the Structural Contexts of the Support Process: Social networks, social statuses, social support, and psychological distress Social Networks and Health 2002 8 New York: Elsevier 269 292
Macinko J Starfield B The utility of social capital in research on health determinants Milbank Q 2001 79 387 427 11565162
Lin N Inequality in social capital Contemp Sociol 2000 29 785 795
Erickson B Flap H, Volker B The Distribution of Gendered Social Capital The Creation and Returns of Social Capital 2004 London: Routledge 27 50
Krueger RA King JA Involving Community Members in Focus Groups – Focus Group Kit 5 1998 London: Sage Publications
Ware JE Kosinski M Keller SD SF-36 physical and mental health summary scales: a user's manual 1994 Boston: The Health Institute
Van der Gaag M Snijders T Position Generator measures and their relationship to other social capital indicators
Flap H Snijders T Volker B van der Gaag M Measurement instruments for social capital of individuals
Pearce L Integrating Survey and Ethnographic Methods For Systematic Anomalous Case Analysis Sociol Methodol 2002 32 103 132
| 15836784 | PMC1087852 | CC BY | 2021-01-04 16:28:56 | no | BMC Public Health. 2005 Apr 18; 5:38 | utf-8 | BMC Public Health | 2,005 | 10.1186/1471-2458-5-38 | oa_comm |
==== Front
BMC Public HealthBMC Public Health1471-2458BioMed Central London 1471-2458-5-391584016210.1186/1471-2458-5-39Research ArticleAssessment of potential effects of the electromagnetic fields of mobile phones on hearing Uloziene Ingrida [email protected] Virgilijus [email protected] Egle [email protected] Viktoras [email protected] Institute for Biomedical Research of Kaunas University of Medicine, Eiveniu 4, Kaunas, Lithuania2 Department of Otolaryngology, Kaunas University of Medicine, Eiveniu 2, Kaunas, Lithuania2005 19 4 2005 5 39 39 18 12 2004 19 4 2005 Copyright © 2005 Uloziene et al; licensee BioMed Central Ltd.2005Uloziene 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 phones have become indispensable as communication tools; however, to date there is only a limited knowledge about interaction between electromagnetic fields (EMF) emitted by mobile phones and auditory function. The aim of the study was to assess potential changes in hearing function as a consequence of exposure to low-intensity EMF's produced by mobile phones at frequencies of 900 and 1800 MHz.
Methods
The within-subject study was performed on thirty volunteers (age 18–30 years) with normal hearing to assess possible acute effect of EMF. Participants attended two sessions: genuine and sham exposure of EMF. Hearing threshold levels (HTL) on pure tone audiometry (PTA) and transient evoked otoacoustic emissions (TEOAE's) were recorded before and immediately after 10 min of genuine and/or sham exposure of mobile phone EMF. The administration of genuine or sham exposure was double blind and counterbalanced in order.
Results
Statistical analysis revealed no significant differences in the mean HTLs of PTA and mean shifts of TEOAE's before and after genuine and/or sham mobile phone EMF 10 min exposure. The data collected showed that average TEOAE levels (averaged across a frequency range) changed less than 2.5 dB between pre- and post-, genuine and sham exposure. The greatest individual change was 10 dB, with a decrease in level from pre- to post- real exposure.
Conclusion
It could be concluded that a 10-min close exposure of EMFs emitted from a mobile phone had no immediate after-effect on measurements of HTL of PTA and TEOAEs in young human subjects and no measurable hearing deterioration was detected in our study.
==== Body
Background
Due to wide spread use of the Global System for Mobile Communications (GSM) mobile phones they have become indispensable as communication tools and therefore any consequent biological effects should be considered as a high-priority environmental health issue. However, to date, there is an inadequate knowledge on what biological systems could be affected by the use of these devices. Biological effects of radio-frequency electromagnetic fields (EMF) transmitted by mobile phones are still a matter of public and scientific discussion. Sensations of burning or warmth around the ear [1], headache [2], disturbance of sleep [3], alteration of cognitive functions and neural activity [4,5], as well as alteration of the blood-brain barrier and a relative decrease in regional cerebral blood flow have been reported as effects resulting from mobile phone use [6,7]. The potential tumorous effect of EMFs is still a subject of debates and research [8-11].
The hearing system is in the closest proximity to the device so that hearing is potentially the most affected target of thermal and non-thermal effects. Moreover, the hearing system and particularly the cochlear outer hair cells (OHC) are known to be highly sensitive to a great variety of exogenous and endogenous agents and externally applied electric and magnetic fields are known to be able to produce some hearing sensation [12]. Despite all these considerations and evidence, only recently, some studies have analyzed the effects of mobile phones on the auditory system [13,14]. However, the results are not completely consistent.
Only limited research data concerning interaction between EMF emitted by mobile phones and auditory function and possible impact on hearing, are available in the literature. The animal experiments using distortion product otoacoustic emissions (DPOAEs) did not show statistically significant changes on the OHC functionality of adult and developing rats exposed as long as 30 days 1–2 h per day to EMF at 900 MHz and 1800 MHz frequencies [15,16].
No measurable change in evoked otoacoustic emissions (OAEs) was detected and none of the subjects reported a deterioration in hearing threshold level after 10-min exposure to the EMFs emitted by mobile phones in a recent human study on possible effects of the EMF of mobile telephones on hearing [17]. Other studies based on the auditory brainstem response and middle latency response methods concluded that 30 min mobile phone use has no short-term adverse effects on the human auditory system [18,19]. The small amount of publications shows that there is a big gap in the knowledge of potential biological effects of cellular phone use on hearing.
The aim of the present study was to assess the acute potential changes in human hearing function as a consequence of exposure to low-intensity EMF's produced by mobile phones at frequencies of 900 and 1800 MHz and a sham-exposure under double-blind conditions as determined by changes in transient evoked otoacoustic emissions (TEOAEs) and hearing threshold levels (HTL) in pure tone audiometry (PTA).
Methods
The protocol of the study was elaborated in the frame of European Commission 5th Framework project "GUARD: potential adverse effects of GSM cellular phones on hearing".
The study group consisted of 30 healthy volunteers (mean age 23.6 ± 1.2 years; range 18 – 29 years) without any evidence of hearing or ear disorder. There were 18 males (mean age 22.8 ± 1.7 years; range 18 – 28 years) and 12 females (mean age 24.9 ± 1.8 years; range 18 – 29 years).
The participants satisfied the following inclusion criteria: age between 18 and 30 years, no history of otological disorder and/or familial hearing disorder, no self reported hearing difficulty or persistent tinnitus, no exposure to severe noise, no ototoxic drugs, no excess consumption of alcohol or drugs 24 hours prior the testing. Instrumental examination: normal appearance of tympanic membrane on otoscopy, hearing threshold levels (HTL) in both ears no worse that 20 dB(A) at any of the standard audiometric frequencies between 0.5 and 8 kHz (Interacoustics AC-40 audiometer, Denmark), no evidence of conductive hearing loss based on air-conduction and bone-conduction audiograms, normal tympanograms and acoustic reflexes present in both ears for stimulation using a 1 kHz tone at 100 dB HL (Interacoustics AT 235 h tympanometer, Denmark), presence of clear recordable TEOAE, defined as SNR greater than 6 dB in two or more half octave bands centred at 1, 2 and 3 kHz (Otodynamics ILO-88 system, London).
All participants attended two study sessions: genuine and sham exposures. The administration of genuine or sham exposure was double blind and counterbalanced in order. Genuine or sham EMF exposures were performed on separate days (at least 24 hours apart) with the tested participant and tester both blind to the condition being used.
The study session consisted of baseline audiological and TEOAE measurements, genuine or sham GSM mobile phone exposure, and followed immediate repeated audiological and TEOAE measurements. Post-exposure measurements were performed in the same order as pre-exposure for each participant.
Pure tone audiometry (PTA) consisted of air conduction using 2-dB steps in the test ear only. TEOAE measurement was used to record TEOAEs according to "linear" protocol using clicks at 80 dB SPL. (Analysis time was 20 ms, 260 stimuli). Each TEOAE measurement run included a minimum of 260 "sweeps". A TEOAE was defined as the response if its amplitude was more than 3 dB above the level of the noise floor. Reproducibility more than 60% was considered acceptable for the analysis at 3 successive frequency bands ranging from 1 to 3 kHz.
GSM exposure utilized the normal output of a consumer mobile phone (NOKIA 6310) at full power for 10 minutes. Fifteen participants received GSM exposure at 900 MHz (full power 2 W) and the other fifteen participants received GSM exposure at 1800 MHz (full power 1 W). The exposure consisted of speech at a typical conversational level via an insert earphone to one ear, plus GSM exposure in either genuine (test) or sham (control) conditions. All test were performed in a sound-treated 2.5 × 2.0 m booth.
The NOKIA 6310 GSM mobile phone without SIM card (checked for correct power output) was used for GSM exposure. Control for carrier frequency, output level, transmit/receive mode of the mobile phone was utilized using specialized PHOENIX software. Therefore the mobile phone was connected via serial data cable from the PC to the phone.
To simulate the normal use of a mobile phone the participants were simultaneously exposed to both GSM radiation and an acoustic stimulus (speech material 10 min of duration). However, to prevent any possible effects from using the speaker in the handset, the speech material was delivered via an insert phone (EAR tone 3A ABR). The insert phone was used without the ear tip inserted and, therefore, the tube was taped along the subject's jaw with the entrance of the tube placed at the tragal notch of the ear. The standard speech material duration of 10 minutes, read out by male speaker, was digitally recorded on minidisk (Sony MiniDisc Recorder MDS 101). Then the speech sample was filtered (Syntrilium "Cool Edit") amplifying the low frequencies of the speech to compensate to some extent for the frequency weighting of the tragal presentation, so that the speech did not sound too unnatural. Having re-recorded the filtered material, the speech sample was calibrated in 2-cc couples (Bruel & Kjer type 4152, Denmark) to the required level in order to produce a sound pressure level at the ear drum equivalent to free field speech level of approximately 60 dB(A). A sound replay system used to replay speech to the ear of the participant consisted of minidisk player, audiometer and insert earphone.
During the test session NOKIA 6310 mobile phone was fixed to the tested ear using special headband and positioning system, so that the centre of the radiated field should be over the entrance of the external ear canal. All parts of the positioning system were made by non-metallic plastic materials in order to avoid any perturbation of the EMF emitted by the mobile phone. The subjects tested were asked to perform an attention task so that they attend to the speech stimulus, such as counting the number of a specific word in the speech material. After the completion of the GSM exposure the participants were asked appropriate questions about the speech material and their experience of any possible subjective effect from the exposure.
The study has been acknowledged by An Independent Ethics Committee of Kaunas University of Medicine and is in compliance with the Helsinki Declaration.
Statistics
A statistical analysis was performed with the SPSS 12.0 (Statistical Package for Social Sciences) for Windows. Means of the groups and standard errors of the means (SEM) for each parameter were obtained for the genuine and sham GSM exposure. Confidence interval (CI) of 0.95 was chosen for statistical evaluation and significance level of 0.05 was chosen for testing statistical hypotheses. As all the participants underwent four-times testing (pre/post genuine exposure and pre/post sham exposure), four dependent samples of repeated measures were obtained. Therefore, for the comparison of the means of the audiological parameters in the total group (n = 30) Repeated Measures Analysis of Variance was used. An Exact Friedman Test was used for the comparison of the means of the audiological parameters in the groups tested for 900 MHz and 1800 MHz EMF frequncies (n = 15 each), respectively.
Results
The subjects tested in the study tolerated the EMF exposure of mobile phones quite well. There were no subjective complaints after the exposure.
The means of air conduction HTL of PTA throughout the testing frequencies of pre/post genuine and pre/post sham GSM exposure are presented in Fig. 1 for 900 MHz exposure, in Fig. 2 for 1800 MHz exposure and in Fig. 3 for the total group, respectively. The analysis of means of HTL of the PTA with the Repeated Measures Analysis of Variance and Exact Friedman Test did not reveal any statistically significant differences between pre/post genuine and pre/post sham exposure groups (p > 0.05).
Figure 1 Hearing threshold levels (900 MHz exposure subgroup). Mean ± SEM, p > 0.05
Figure 2 Hearing threshold levels (1800 MHz exposure subgroup). Mean ± SEM, p > 0.05
Figure 3 Hearing threshold levels (total group, n = 30). Mean ± SEM, p > 0.05
As all of the subjects tested in the study had a normal hearing, recordings of TEOAEs were obtained (6–10 dB) above the noise floor through 1 – 3 kHz test frequency range for all sessions. Reproducibility of TEOAEs was >60% (91 ± 3 % in average). The mean amplitude shifts of the TEOAE measurements of pre/post genuine and pre/post sham GSM exposure are presented in Fig. 4 for 900 MHz exposure, in Fig. 5 for 1800 MHz exposure and in Fig. 6 for the total group, respectively. Statistical analysis with the Repeated Measures Analysis of Variance and Exact Friedman Test did not reveal any statistically significant differences in mean TEOAEs amplitude shifts between genuine and sham exposure groups (p > 0.05). The data collected showed that average TEOAE levels (averaged across a frequency range) changed less than 2.5 dB between pre- and post-, genuine and sham exposure. The greatest individual change was 10 dB, with a decrease in level from pre- to post- real exposure.
Figure 4 TEOAE amplitude shifts (900 MHz exposure subgroup). Mean ± SEM, p > 0.05
Figure 5 Mean TEOAE amplitude shifts (1800 MHz exposure subgroup, n = 15). Mean ± SEM, p > 0.05
Figure 6 Mean TEOAE amplitude shifts (total group, n = 30,) Mean ± SEM, p > 0.05
The results of the present study suggest that 10-min exposure to EMFs emitted by GSM mobile phone did not cause any detectable alterations in neither PTA nor TEOAEs.
Discussion
GSM mobile phones have become very commonly used throughout the world within a short period of time. This has given rise to concern about potential influences of EMF emitted by mobile phones on health. Although there is no clear evidence to demonstrate harmful physiological effects of EMF at the levels used by mobile phones, however, there is a widespread public concern that there may be some potential harm [8-11,20-22].
Potential effects of mobile phone EMF radiation on hearing should be considered as one of the major priorities in the research of potential adverse effects of mobile phone use. Mobile phones are usually held in the closest proximity to the external ear and therefore, EMF exposure at the ear is high due to radiation from a remote earpiece. Since the cochlea is enclosed by very dense compact bone, located relatively deep, congested with the perilymph and endolymph, they all help to shield it from the mobile phone EMF [17]. However, the OHC of the inner ear are known to be the most sensitive and vulnerable elements of the auditory pathway. If subtle cochlear involvements occur, they might be detected through changes in TEOAEs, which directly reflect the function of cochlear OHC. Even minor changes in the functioning of OHCs, caused by various noxious factors, are known to considerably affect TEOAE amplitude [23-26]. On the other hand, TEOAEs represent acoustic responses of OHCs, which act like mechanoreceptor cells that generate force in their cell bodies to amplify sound and provide the exquisite sensitivity of the cochlea [27,29]. Consequently, the piezoelectric properties of OHCs that are essential for hearing might be relatively easily damaged by external EMFs emitted by mobile phones. Therefore, the present study in addition to conventional measures of HTL, which require a subjective response, employed an objective methodology of registration of TEOAE that is able to detect very small effects in hearing function with appropriate statistical power.
The experimental paradigm used in this study was the within-subject paradigm. This paradigm provided audiological measurements immediately before and immediately after exposure to EMF via a commercial mobile phone. As the procedure was conducted twice: one with a genuine exposure and one with a sham exposure, this approach maximized sensitivity to change, because between subjects' variations in the results were minimised by calculation of the difference between before and after measurements. A double-blind design of the procedure maximized objectivity of the experiment. EMF exposure dose used in the study was necessarily low but comparable with the use of the mobile phones in normal day life.
As stated above, statistical analysis of the results of the present study did not reveal any significant alterations of HTL after 10 min GSM mobile phone exposure. However, the main focus of this study was to analyse the effects of EMF of mobile phones on the TEOAE recorded before and immediately after a sham and a genuine exposure in a group of young subjects. The data collected showed that average TEOAE levels (averaged across a frequency range) changed less than 2.5 dB between pre- and post-, genuine and sham exposure. Therefore the variability of the TEOAE recorded before and after the GSM exposure and the individual variability appeared to be small and not statistically significant. This allows us to state that 10 min of EMF exposure at the maximum power (peak power: 1 W or 2 W according to the frequency) does not induce any measurable changes in the TEOAEs.
Some other unanswered questions of the present study should be mentioned. The measurements used in the study have been restricted by the frequency spectrum of the commercially available TEOAE equipment. Possibly, higher frequency instruments could be able to reveal more comprehensive information about the effects of EMF exposure [17]. As the study protocol was based on the comparison of the measures of the audiological tests obtained before and immediately after EMF exposure, only a relatively long-term or chronic alteration in hearing function could be detected by the present investigations. Some potential transitory, i.e. reversible, alterations in hearing function lasting for only a short time during the EMF exposure cannot be detected by these methods. Therefore, the simultaneous measurement of hearing function during the mobile phone's EMF emission would be of scientific interest. However, these implications could be considered as guidelines for further investigations.
Conclusion
It could be concluded that a 10-min close exposure of EMFs emitted from a mobile phone had no immediate after-effect on measurements of HTL of PTA and TEOAEs in young adult human subjects and no measurable hearing deterioration at least at outer and middle ear and cochlear levels was detected in our study.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
IU has been the principal investigator, participated in study design and coordination and helped to draft the manuscript. VU participated in the planning of the study and coordinated the writing of the manuscript. VS performed the statistical analysis. EG participated in acquisition of audiological data. All authors contributed to the interpretation of results, have read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This study was sponsored by European Commission 5FW project "GUARD: potential adverse effects of GSM cellular phones on hearing". Contract number: QLK4-CT-00150
==== Refs
Oftedal G Wilen J Sandstrom M Mild KH Symptoms experienced in connection with mobile phone use Occup Med (Lond) 2000 50 237 245 10912374
Frey AH Headaches from cellular telephones: are they real and what are the implications? Environ Health Perspect 1998 106 101 103 9441959
Borbely AA Huber R Graf T Fuchs B Gallmann E Achermann P Pulled high-frequency electromagnetic field affects human sleep and sleep electroencephalogram Neurosci Lett 1999 275 207 210 10580711 10.1016/S0304-3940(99)00770-3
Preece AW Iwi G Davies-Smith A Effect of a 915-MHz simulated mobile phone signal on cognitive function in man Int J Radiat Biol 1999 75 447 456 10331850 10.1080/095530099140375
Hamblin DL Wood AW Croft RJ Stough C Examining the effects of electromagnetic fields emitted by GSM mobile phones on human event-related potentials and performance during an auditory task Clin Neurophysiol 2004 115 171 178 14706485 10.1016/S1388-2457(03)00313-4
Fritze K Sommer C Schmitz B Effect of global system for mobile communication (GSM) microwave exposure on blood-brain barrier permeability in rat Acta Neuropathol (Berl) 1997 94 465 470 9386779 10.1007/s004010050734
Haarala C Aalto S Hautzel H Julkunen L Rinne JO Laine M Krause B Hamalainen H Effects of a 902 MHz mobile phone on cerebral blood flow in humans: a PET study Neuroreport 2003 14 2019 2023 14600490 10.1097/00001756-200311140-00003
Moulder JE Erdreich LS Malyapa RS Merritt J Pickard WF Vijayalaxmi Cell phones and cancer: what is the evidence for a connection? Radiat Res 1999 151 513 531 10319725
Blettner M Berg G Are mobile phones harmful? Acta Oncol 2000 39 927 930 11206998 10.1080/02841860050215891
Christensen HC Schuz J Kosteljanetz M Poulsen SH Thomsen J Johansen Christoffer J Cellular telephone use and risk of acoustic neuroma Am J Epidemiol 2004 159 277 283 14742288 10.1093/aje/kwh032
Lonn S Ahlbom A Hall P Feychting M Mobile phone use and the risk of acoustic neuroma Epidemiology 2004 15 653 659 15475713 10.1097/01.ede.0000142519.00772.bf
Watanabe Y Tanaka T Taki M Watanabe S FDTD Analysis of Microwave Hearing Effect IEEE Trans Microwave Theory Tech 2000 48 2126 2132 10.1109/22.884204
Kellenyi L Thurockzy G Faludy B Lenard L Effects of mobile GSM radiotelephone exposure on the auditory brainstem response (ABR) Neurobiology 1999 7 79 81 10746254
Marino C Cristalli G Galloni P Pasqualetti P Piscitelli M Lovisolo GA Effects of Micro-waves (900 MHz) on the cochlear receptor: exposure systems and preliminary results Radiat Environment Bioph 2000 39 131 136 10.1007/s004110000049
Parazzini M Marino C Galloni P Piscitelli M Tognola G Grandori F Ravazzani P Effects of electromagnetic fields on hearing: study of the influence of GSM cellular phones on the inner auditory system of Sprague-Dawley rats Proceedings 2nd Medical and Biological Engineering Conference, Vienna 1290 1 2002, December 4–8
Kizilay A Ozturan O Erdem T Kalcioglu T Miman MC Effects of chronic exposure of electromagnetic fields from mobile phones on hearing in rats Auris Nasus Larynx 2003 30 239 245 12927285 10.1016/S0385-8146(03)00054-3
Ozturan O Erdem T Miman MC Kalcioglu MT Oncel S Effects of the electromagnetic field of mobile telephones on hearing Acta Otolaryngol 2002 122 289 293 12030576 10.1080/000164802753648178
Arai N Enomoto H Okabe S Yuasa K Kamimura Y Ugawa Y Thirty minutes mobile phone use has no short-term adverse effects on central auditory pathways Clin Neurophysiol 2003 114 1390 1394 12888020 10.1016/S1388-2457(03)00124-X
Bak M Sliwinska-Kowalska M Zmyslony M Dudarewicz A No effect of acute exposure to the electromagnetic field emitted by mobile phones on brainstem auditory potentials in young volunteers Int J Occup Med Environ Health 2003 16 201 209 14587532
Weinberger Z Richter ED Cellular telephones and effects on the brain: the head as an antenna and brain tissue as a radio receiver Med Hypotheses 2002 59 703 705 12445512 10.1016/S0306-9877(02)00298-0
Kompis M Hausler R Electromagnetic interference of bone-anchored hearing aids by cellular phones revisited Acta Otolaryngol 2002 122 510 512 12206260 10.1080/00016480260092318
Samkange-Zeeb F Berg G Blettner M Validation of self reported cellular phone use J Expo Anal Environ Epidemiol 2004 14 245 248 15141153 10.1038/sj.jea.7500321
Skellett RA Crist JR Fallon M Bobbin RB Chronic low-level noise exposure alters distortion product otoacoustic emissions Hearing Res 1996 98 68 76 10.1016/0378-5955(96)00062-7
Janssen T Boege P Oestreicher E Arnold W Tinnitus and 2f1-f2 distortion product otoacoustic emissions following salicylate overdose J Acoust Soc Am 2000 107 1790 1792 10738835 10.1121/1.428578
Frank AM Alexiou C Hulin P Janssen T Arnold W Trappe AE Non-invasive measurement of intracranial pressure changes by otoacoustic emissions (OAEs) – a report of preliminary data Zentralbl Neurochir 2000 61 177 180 11392287 10.1055/s-2000-15597
Ravazzani P Tognola G Parazzini M Grandori F Principal component analysis as a method to facilitate fast detection of transient-evoked otoacoustic emissions IEEE Trans Biomed Eng 2003 50 249 252 12665039 10.1109/TBME.2002.807645
Parazzini M Radiofrequency electromagnetic fields produced by mobile phones on the auditory system: study on biological effects and numerical dosimetry Doctoral thesis, Istituto di Ingegneria Biomedica CNR, Milan 2004
Dallos P The active cochlea Neurosci 1992 12 4575 4585
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
| 15840162 | PMC1087853 | CC BY | 2021-01-04 16:28:57 | no | BMC Public Health. 2005 Apr 19; 5:39 | utf-8 | BMC Public Health | 2,005 | 10.1186/1471-2458-5-39 | oa_comm |
==== Front
BMC Pulm MedBMC Pulmonary Medicine1471-2466BioMed Central London 1471-2466-5-61582631410.1186/1471-2466-5-6Study ProtocolInhaled corticosteroids for abnormal pulmonary function in children with a history of Chronic Lung Disease of Infancy: study protocol [ISRCTN55153521] Alotaibi Saad [email protected] David [email protected] Mark [email protected] Reginald [email protected] Sheldon [email protected] Pediatric Pulmonologist, Pediatric Department, Farwanyah Hospital, Kuwait2 Pediatric Emergency Consultant, Pediatric Emergency Department, Alberta Children Hospital, University of Calgary, Calgary, Canada3 Pediatric Pulmonology Consultant, Pediatric Pulmonology Division, Alberta Children Hospital, University of Calgary, Calgary, Canada4 Community Health Sciences, Faculty of MEDICINE (Medical School) Graduate Science Education, Paediatrics, University of Calgary, Calgary, Canada5 Head of Pediatric Pulmonology Division, University of Calgary, Alberta Children Hospital, Calgary, Canada2005 12 4 2005 5 6 6 11 3 2005 12 4 2005 Copyright © 2005 Alotaibi et al; licensee BioMed Central Ltd.2005Alotaibi 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 considerable evidence from the literature that children with chronic lung disease of infancy (CLD) have abnormal pulmonary function in childhood and this could have an impact on their life quality and overall health. There are similarities between CLD and asthma, and corticosteroids are the mainstay treatment for asthma. Many physicians use inhaled corticosteroids in children with CLD with no evidence. Therefore we wish to conduct a randomized double-blinded placebo controlled trial to test for the role of inhaled corticosteroids in children aged from3 to 9 years with a history of CLD. Our primary hypothesis will be that inhaled corticosteroids are beneficial in children with CLD.
Methods
Our primary hypothesis is that using inhaled steroids; Beclomethasone Dipropionate (QVAR) 100 mcg 2 puffs 2 times a day for 6 weeks will improve the respiratory system resistance and the quality of life in children with CLD.
Discussion
We propose that Beclomethasone Dipropionate (QVAR) will affect the pulmonary function after 6 weeks of treatment. In summary we think that our study will highlight knowledge on whether the use of inhaled steroids is clinically effective for CLD.
==== Body
Background
Children with a history of severe CLD were left with significant pulmonary sequelae at school age compared with prematurely born children who had required ventilator assistance and supplemental oxygen in the neonatal period but who did not go on to have CLD [1]. It was also documented that prematurely born children with respiratory distress syndrome (RDS) have long term pulmonary sequelae consisting of airway obstruction and airway hyper reactivity, which are related to both the severity of the initial pulmonary insult and its treatment as well as to a familial predisposition to airway hyper reactivity. The role of inhaled corticosteroids in children with chronic lung disease of infancy or bronchopulmonary dysplasia was not previously studied although most if not all physicians use this treatment for these children whenever they develop chest infections or wheezy episodes. This is a randomized double blinded placebo controlled study using inhaled corticosteroid (CFC free Beclomethasone dipropionate, QVAR) for 6 weeks and measuring the effect on pulmonary function and quality of life.
Methods
We will use a double blind randomized parallel design. Patients will be invited to the study by sending a letter from the perinatal clinic office inviting them to contact us. Consent will be handed to the parents at the beginning of the study. Subject randomization will be carried out using a computer generated random numbers. Medications and placebo will be kept locked in a cabinet in the clinic area, one of the clinic nurses will be responsible for dispensing the medication or placebo. The medication or placebo will be taken twice daily, at 08:00 and 20:00. The Placebo will have the same taste, shape and color of the study drug. There will be a run in period for 1 week before the study, for teaching the patients, and checking proper technique, and filling out the Child Health Questionnaire (CHQ). In the first visit, informed consent filled, inclusion and exclusion criteria will be obtained. Medical and surgical history, physical exam, inhalation and pulmonary function technique will be done. We will review the daily diary, and prepare the placebo for all patients in the run in period. One week after the run in period, randomization carried out and patients entered into the study. Phone calls will be carried out every two weeks to optimize compliance. After six weeks of treatment, CHQ will be filled. Forced expiratory volume at 1 second (FEV1) (for children older than 6 years of age) and respiratory system resistance at 5 Hz frequencies will be measured.
The Impulse Oscillometry tool to be used is developed by Jaeger Company (Wurzburg, Germany). In short rectangular mechanical impulses containing the whole frequency spectrum are applied to the respiratory system by a mouthpiece, while the patient is breathing quietly. The resulting pressure and volume signals are analysed for their amplitude and phase difference to determine the Rrs at 5 Hz of the total respiratory system [2]. One investigator will perform all pulmonary function tests. After a short explanation, the children will perform some practice trials. When the child is feeling comfortable, a measurement of 25–35 seconds will be started. Patients will be asked to breathe quietly through a round mouthpiece. A nose clip will be used and the cheeks will be supported by the hands of the investigator to minimize pressure loss through the upper airway shunt. Measurement will be discarded if the time-flow pattern showed apnea, speaking, swallowing or air leak, as described by Bisgraaed and Klug. The following parameter Rrs at 5 Hz will be recorded before and 20 minutes after administration of 200 mcg salbutamol. The spacer device used to administer the medication will be an aerochamber with mask for children under the age of 4.7 yrs old and with a mouth piece for children above the age of 4.7 yrs [3-5].
The Study of a patient should be delayed six weeks if developed a viral respiratory illness. Bronchodilators should not be used on the day of the study visit (at least 24 hours for long acting, 6 hours for short acting). Neonatal variables to be collected from medical records: gender, birth weight, birth length, head circumference, gestational age, neonatal treatment, total days on oxygen, maximal FiO2, days on CPAP, days on IPPV, use of steroids (how much and how long), and the use of surfactant. Maternal data: parity, living offspring's, and smoking days during pregnancy. Skin prick testing for the 10 common allergens will be done at the start of the study. Atopy will be defined by one or more positive reactions (3 mm wheal) to 10 common aeroallergens. Parents will fill out daily symptom scores. Social and environmental data to be collected are smoking habits, pets, family history of asthma and asthma in the child. Methacholine challenge testing will be done in the start of the study to define those patients with airway hyper reactivity, which might be a sensitive indicator of response to steroids [6,7]. Respiratory function will be measured using Spirometry (FEV1) and Impulse Oscillometry (RS @ 5 Hz). Quality of life will be measured using the CHQ. Treatment group will receive QVAR 100 mcg 2 puffs twice daily; placebo will be given to the other arm for 6 weeks, after which the respiratory function and quality of life will be measured.
Aim of the study
To study the effect of inhaled steroids on school- aged children with CLD.
Primary hypothesis
Inhaled corticosteroids, Beclomethasone Dipropionate will improve respiratory system resistance after 6 weeks of treatment.
Primary outcome measures
Improvement in respiratory system resistance at 5 Hz frequency measured by impulse oscillometry after 6 weeks of inhaled corticosteroids in CLD.
Secondary outcome measures
Improvement in Quality of life system score using Child Health Questionnaire (CHQ) and FEV1 response to bronchodilators (Salbutamol) after inhaled corticosteroids treatment for 6 weeks in children with CLD.
The inclusion criteria: children between 3 and 9 years of age with a history of CLD with the following criteria.
1. Premature birth at 36 weeks gestation or less.
2. Clinical and radiological diagnosis of CLD. Patient charts will be reviewed and the x-rays will be reviewed, even if there is no x-ray findings patients will be included if they fulfil the other criteria.
3. Requirement of supplemental oxygen to maintain saturation of 90%, for at least 36 wk corrected gestation.
The exclusion criteria:
1. Cardiovascular disease other than PDA (Patent Ductus Arteriosus)
2. Those with severe neurological disease, developmental delay. Oxygen requirement for respiratory diagnosis other than CLD e.g. congenital diaphragmatic hernia and aspiration. An inability to perform spirometry or IOS, or noncompliance with therapy and those with pneumonia.
Statistical consideration
The required number of patients was estimated as n = 74 (37 in each group) and to allow 10% dropout we need 82 patients. The sample size estimation was based on a type I error rate of = .05 and a type II error rate of = 0.20. The improvement in airway resistance at 5 HZ as primary end point was expected to be 15% for the treatment group. The coefficient variation for the impulse oscillometry measurement is 10%. The primary statistical analysis of efficacy data is based on the intention-to-treat patients, who were randomly assigned to treatment and received at least one dose of medication. Baseline characteristics will be compared using descriptive statistics including confidence intervals. Imbalance between the groups on key variables, which might influence the observed outcome, will indicate that an adjusted analysis is necessary using these variables. The differences between airway resistance at 5 HZ response will be tested with a 2-tailed paired test on a significance level of = .05. Analysis of FEV1 improvement will be performed by two-tailed T test (= .05). All tests, except those for CHQ, will be done by SSPS procedures. For CHQ analysis, the differences of sum scores at week 6 after treatment to baseline will be tested by Wilcoxon rank sum test using StatXa.
Discussion
Bronchopulmonary Dysplasia (BPD) is a chronic lung disease related to injury induced by mechanical ventilation and oxygen therapy in prematurely born infants. William Northway, a radiologist at Stanford University, described this pathology in 1967[8]. More recently the term chronic lung disease (CLD) has been used since nowadays preterm infants with birth weights between 500 and 750 grams have at least a 50 percent chance of survival at 24 weeks gestation age[9]. The clinical diagnosis of this CLD is made at one month of age in prematurely born infants who have undergone mechanical ventilation for at least one week, who have symptoms of persistent respiratory distress, are dependent on supplemental oxygen, and who have chest radiographs that show rounded radio lucent areas in the lung alternating with thinner strands of radiodensity.
The pathogenesis of CLD is believed to be multifactorial, with pulmonary oxygen toxicity and barotrauma having important roles [10]. More recently inflammatory mediators, cytokines, and proteolytic enzymes released by inflammatory cells, whose influx into the lungs is the consequence of oxygen exposure and prenatal or postnatal sepsis, were described [11]. The release of neuropeptides such as substance P from afferent nerve endings in the airways or lung parenchyma may accelerate this inflammatory cascade[12]. Coates et al found decreased expiratory flows in children 7 and 8 years old who had survived the respiratory distress syndrome, and later suggested that oxygen therapy could have been at least partly responsible for these sequelae [13]. CLD has become the most common form of chronic lung disease in infants in the United States with approximately 7000 new cases occurring each year [14].
It has been noted that infants with BPD were twice as likely to be hospitalized in the first 2 years of life compared with preterm infants without BPD [15]. But at age 7 years both groups had more cough and wheeze than the term group. Pulmonary function including lung volumes and expiratory airflow were worse in the BPD group. This group showed more evidence of gas trapping and hyperinflation and significant expiratory airflow obstruction [15]. This was confirmed by the study done by Jacob et al who compared children of BPD at 11 years old with children in a control group who were born preterm but who did not have BPD [9]. This was also supported by the findings of Coates et al who confirmed that premature birth and the development of the RDS of the newborn have been associated with long term pulmonary sequelae [13,16]. Is there any treatment available for those children?
The literature suggested that there is significant morbidity and mortality from chronic lung disease in infancy into adult and elderly ages. In a follow up study of men born during 1911–1930 whose birth weight, weight at 1 year, and childhood respiratory illnesses were recorded at that time by health visitors [17], Barker et al found that lower birth weight was associated with worse adult lung function. Bronchitis, pneumonia, or whooping cough in infancy further reduced adult lung function, they concluded that death rates from chronic obstructive disease, but not from another disease related to smoking or lung cancer fell with increasing birth weight and weight at 1 year. This was supported by the findings of Peto et al [18] in their follow up study of men whose lung function and respiratory symptoms had been assessed in adult life and showed that the risk of death from chronic obstructive airway disease was strongly correlated with the initial FEV1.
Liggins and Howie have demonstrated that the administration of a corticosteroid leads to a decreased incidence of RDS in human premature infants[19]. It is well known that inhaled and systemic steroids are used for the prevention of CLD in infancy [20]. Any pulmonary growth retardation would be expected to have an effect on the airways and result in increased airway resistance and lower flows. It was pointed out that the only major changes in airways from birth to adulthood are in their size [21]. High incidence of asthma has been noted in families of neonates in whom the RDS was complicated by CLD [22]. Other clinical studies suggest an increased incidence of airway hyper reactivity in prematurely born children [23-26]. Recurrent wheeze in childhood and readmission to hospital for bronchiolitis like illness has been observed among prematurely born children who survived the RDS [24]. Most patients with CLD also have increased airway lability and improvement in expiratory flow rates following bronchodilator inhalation suggesting that such therapy may improve pulmonary function in these children [15]. It was anticipated that the incidence of CLD would decrease after the introduction of surfactant therapy, clinical trials by Horbar and Long failed to confirm this [27,28].
So far there are no studies have been published studying the clinical response to inhaled corticosteroids in children CLD. Pulmonary function testing for these children can be done using different techniques. Pennings et al [29] used forced oscillation technique (FOT) to detect changes in airway bronchial tone of asthmatic patients, that were not detected by conventional flow volume recording. It was found that the changes of the diameter of small airways of children could be missed by routine spirometry. The apparent increase incidence of bronchiolitis in RDS survivor's points to these small airways as a source of pathology [30]. In 1956, Dubios et al described the use of FOT as a measure of total respiratory system impedance, which was non invasive and required minimal cooperation from the subject and it can be applied to infants and older children. Laberge et al used FOT to measure respiratory system resistance in 3 to 6 years old asthmatic children and they concluded that the dry powder inhaler (Turbuhaler) and the metered dose inhaler with spacer (Nebuhaler) are equally effective in administering terbutaline [31].
Conclusion
We propose that Beclomethasone Dipropionate (QVAR) will affect the pulmonary function after 6 weeks of treatment. In summary we think that our study will highlight knowledge on whether the use of inhaled steroids is clinically effective for CLD. This study was approved by the conjoint Health Research Ethics Board and Child Health Research Committee at the University of Calgary, Alberta, Canada.
List of abbreviations
CLD = chronic lung disease of infancy.
CHQ = child health questionnaire.
FEV1 = forced expiratory volume at 1 second.
Rsr = respiratory system resistance.
IOS = impulse oscillometry.
Competing interests
The authors declare that they have no competing interests. This study had a gift of fund from 3 M pharmaceuticals-Canada in the form of a payment of CAD$ 6,400.00 as well as 100 canisters of 3M QVAR™ 100 ug and 100 canisters of placebo.
Authors' contributions
The first author as research pulmonology fellow collected the literature and writing, editing the manuscript and as organiser of this study and patient recruitment. Second author gave statistical help of this study protocol. Third author helped in the design of the study. Forth author gave ethical help for the study and recruitment of patients form perinatal clinic. Fifth author was the principal investigator in this clinical fellow research study protocol, also contributed to the editing and writing of the manuscript. All authors contributed to the study protocol from all aspects.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
For the University of Calgary Ethics committee for the approval of this study protocol and for 3M pharmaceutical company for their support in providing funding in the form of a payment of CAD$ 6,400.00 as well as 100 canisters of 3M QVAR™ 100 ug and 100 canisters of placebo.
==== Refs
Bertrand JM Riley SP Popkin J Coates AL The long-term pulmonary sequelae of prematurity: the role of familial airway hyperreactivity and the respiratory distress syndrome N Engl J Med 1985 312 742 745 3974651
Hellinckx J De Boeck K Bande-Knops J van der Poel M Demedts M Bronchodilator response in 3-6.5 years old healthy and stable asthmatic children Eur Respir J 1998 12 438 443 9727798 10.1183/09031936.98.12020438
Klug B Bisgaard H Specific airway resistance, interrupter resistance, and respiratory impedance in healthy children aged 2-7 years Pediatr Pulmonol 1998 25 322 331 9635934 10.1002/(SICI)1099-0496(199805)25:5<322::AID-PPUL6>3.0.CO;2-K
Klug B Bisgaard H Measurement of lung function in awake 2-4-year-old asthmatic children during methacholine challenge and acute asthma: a comparison of the impulse oscillation technique, the interrupter technique, and transcutaneous measurement of oxygen versus whole-body plethysmography Pediatr Pulmonol 1996 21 290 300 8726154 10.1002/(SICI)1099-0496(199605)21:5<290::AID-PPUL4>3.0.CO;2-R
Bisgaard H Klug B Lung function measurement in awake young children Eur Respir J 1995 8 2067 2075 8666102 10.1183/09031936.95.08122067
Crapo RO Casaburi R Coates AL Enright PL Hankinson JL Irvin CG MacIntyre NR McKay RT Wanger JS Anderson SD Cockcroft DW Fish JE Sterk PJ Guidelines for methacholine and exercise challenge testing-1999. This official statement of the American Thoracic Society was adopted by the ATS Board of Directors, July 1999 Am J Respir Crit Care Med 2000 161 309 329 10619836
Nielsen KG Bisgaard H The effect of inhaled budesonide on symptoms, lung function, and cold air and methacholine responsiveness in 2- to 5-year-old asthmatic children Am J Respir Crit Care Med 2000 162 1500 1506 11029368
Northway WHJ Moss RB Carlisle KB Parker BR Popp RL Pitlick PT Eichler I Lamm RL Brown BWJ Late pulmonary sequelae of bronchopulmonary dysplasia N Engl J Med 1990 323 1793 1799 2247118
Jacob SV Coates AL Lands LC MacNeish CF Riley SP Hornby L Outerbridge EW Davis GM Williams RL Long-term pulmonary sequelae of severe bronchopulmonary dysplasia J Pediatr 1998 133 193 200 9709705
Farrell PA Fiascone JM Bronchopulmonary dysplasia in the 1990s: a review for the pediatrician Curr Probl Pediatr 1997 27 129 163 9145287
Groneck P Gotze-Speer B Oppermann M Eiffert H Speer CP Association of pulmonary inflammation and increased microvascular permeability during the development of bronchopulmonary dysplasia: a sequential analysis of inflammatory mediators in respiratory fluids of high-risk preterm neonates Pediatrics 1994 93 712 718 8165067
Agani FH Kuo NT Chang CH Dreshaj IA Farver CF Krause JE Ernsberger P Haxhiu MA Martin RJ Effect of hyperoxia on substance P expression and airway reactivity in the developing lung Am J Physiol 1997 273 L40 5 9252538
Coates AL Bergsteinsson H Desmond K Outerbridge EW Beaudry PH Long-term pulmonary sequelae of premature birth with and without idiopathic respiratory distress syndrome J Pediatr 1977 90 611 616 839378
Stevenson DK Wright LL Lemons JA Oh W Korones SB Papile LA Bauer CR Stoll BJ Tyson JE Shankaran S Fanaroff AA Donovan EF Ehrenkranz RA Verter J Very low birth weight outcomes of the National Institute of Child Health and Human Development Neonatal Research Network, January 1993 through December 1994 Am J Obstet Gynecol 1998 179 1632 1639 9855609
Gross SJ Iannuzzi DM Kveselis DA Anbar RD Effect of preterm birth on pulmonary function at school age: a prospective controlled study J Pediatr 1998 133 188 192 9709704
Coates AL Chronic lung disease in infants--long-term pulmonary sequelae Pediatr Pulmonol Suppl 1997 16 40 42 9443190
Barker DJ Godfrey KM Fall C Osmond C Winter PD Shaheen SO Relation of birth weight and childhood respiratory infection to adult lung function and death from chronic obstructive airways disease Bmj 1991 303 671 675 1912913
Peto R Speizer FE Cochrane AL Moore F Fletcher CM Tinker CM Higgins IT Gray RG Richards SM Gilliland J Norman-Smith B The relevance in adults of air-flow obstruction, but not of mucus hypersecretion, to mortality from chronic lung disease. Results from 20 years of prospective observation Am Rev Respir Dis 1983 128 491 500 6614643
Liggins GC Howie RN A controlled trial of antepartum glucocorticoid treatment for prevention of the respiratory distress syndrome in premature infants Pediatrics 1972 50 515 525 4561295
Avent ML Gal P Ransom JL The role of inhaled steroids in the treatment of bronchopulmonary dysplasia Neonatal Netw 1994 13 63 69 7512192
Hogg DJ.Landau LI PPD Age as a factor in the distribution of lower-airway conductance and in the pathological anatomy of obstructive lung disease. N Engl J Med 1870 282 1283
Nickerson BG Taussig LM Family history of asthma in infants with bronchopulmonary dysplasia Pediatrics 1980 65 1140 1144 7375238
Shepard FM Johnston RBJ Klatte EC Burko H Stahlman M Residual pulmonary findings in clinical hyaline-membrane disease N Engl J Med 1968 279 1063 1071 5686463
Outerbridge EW Nogrady B Beaudry PH Stern L Idiopathic respiratory distress syndrome. Recurrent respiratory illness in survivors Am J Dis Child 1972 123 99 104 5025869
Smyth JA Tabachnik E Duncan WJ Reilly BJ Levison H Pulmonary function and bronchial hyperreactivity in long-term survivors of bronchopulmonary dysplasia Pediatrics 1981 68 336 340 6792584
Bader D Ramos AD Lew CD Platzker AC Stabile MW Keens TG Childhood sequelae of infant lung disease: exercise and pulmonary function abnormalities after bronchopulmonary dysplasia J Pediatr 1987 110 693 699 3572620
Horbar JD Soll RF Sutherland JM Kotagal U Philip AG Kessler DL Little GA Edwards WH Vidyasagar D Raju TN A multicenter randomized, placebo-controlled trial of surfactant therapy for respiratory distress syndrome N Engl J Med 1989 320 959 965 2648150
Long W Thompson T Sundell H Schumacher R Volberg F Guthrie R Effects of two rescue doses of a synthetic surfactant on mortality rate and survival without bronchopulmonary dysplasia in 700- to 1350-gram infants with respiratory distress syndrome. The American Exosurf Neonatal Study Group I J Pediatr 1991 118 595 605 2007938
Pennings HJ Wouters EF Effect of inhaled beclomethasone dipropionate on isocapnic hyperventilation with cold air in asthmatics, measured with forced oscillation technique Eur Respir J 1997 10 665 671 9073002
Lewis S A follow-up study of the respiratory distress syndrome Proc R Soc Med 1968 61 771 773 5673400
Laberge S SSDSPTJP Comparison of inhaled terbutaline administered by either the Turbuhaler dry powder inhaler or a metered-dose inhaler with spacer in preschool children with asthma. J Pediatr 1994 124 815 817 8176575
| 15826314 | PMC1087854 | CC BY | 2021-01-04 16:30:12 | no | BMC Pulm Med. 2005 Apr 12; 5:6 | utf-8 | BMC Pulm Med | 2,005 | 10.1186/1471-2466-5-6 | oa_comm |
==== Front
BMC Plant BiolBMC Plant Biology1471-2229BioMed Central London 1471-2229-5-41579677810.1186/1471-2229-5-4Research ArticleImaging plant cell death: GFP-Nit1 aggregation marks an early step of wound and herbicide induced cell death Cutler Sean R [email protected] Chris R [email protected] Department of Botany, University of Toronto, 25 Willcocks St., Toronto, Ontario, M5S 3B2, Canada2 Carnegie Institution, Department of Plant Biology, 260 Panama Street, Stanford, CA 94305, USA2005 29 3 2005 5 4 4 9 12 2004 29 3 2005 Copyright © 2005 Cutler and Somerville; licensee BioMed Central Ltd.2005Cutler and Somerville; 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 great deal is known about the morphological endpoints of plant cell death, but relatively little is known about its sequence of events and / or its execution at the biochemical level. Live cell imaging using GFP-tagged markers is a powerful way to provide dynamic portraits of a cellular process that can in turn provide a descriptive foundation valuable for future biochemical and genetic investigations.
Results
While characterizing a collection of random GFP-protein fusion markers we discovered that mechanical wounding induces rapid aggregation of a GFP-Nitrilase 1 fusion protein in Arabidopsis cells directly abutting wound sites. Time-lapse imaging of this response shows that the aggregation occurs in cells that subsequently die 30 – 60 minutes post-wounding, indicating that GFP-Nit1 aggregation is an early marker of cell death at wound sites. Time-lapse confocal imaging was used to characterize wound-induced cell death using GFP-Nit1 and markers of the nucleus and endoplasmic reticulum. These analyses provide dynamic portraits of well-known death-associated responses such as nuclear contraction and cellular collapse and reveal novel features such as nuclear envelope separation, ER vesiculation and loss of nuclear-lumen contents. As a parallel system for imaging cell death, we developed a chemical method for rapidly triggering cell death using the herbicides bromoxynil or chloroxynil which cause rapid GFP-Nit1 aggregation, loss of nuclear contents and cellular collapse, but not nuclear contraction, separating this response from others during plant cell death.
Conclusion
Our observations place aggregation of Nitrilase 1 as one of the earliest events associated with wound and herbicide-induced cell death and highlight several novel cellular events that occur as plant cells die. Our data create a detailed descriptive framework for future investigations of plant cell death and provide new tools for both its cellular and biochemical analysis.
==== Body
Background
Events that compromise the integrity of a single cell can threaten the well-being of an entire multi-cellular organism, a fact reflected by the diverse mechanisms that organisms use to eliminate potentially dangerous cells [1]. For example, when injured beyond repair by high doses of irradiation or by treatment with cytotoxic agents, regulated cell death may be activated to remove damaged cells. This process is best understood in animal cells from extensive analyses of the mechanisms that underpin apoptosis, a regulated process marked by a set of stereotyped changes in cellular architecture that culminate in cell death [2,3].
In plant cells, regulated cell death occurs in numerous contexts, such as during development of xylem elements [4] or as part of the hypersensitive response (HR) to pathogen attack [5-7]. Although the hypersensitive response is triggered by highly specific plant-pathogen interactions and is under genetic control [7-9], the HR exhibits many similarities to plant responses triggered by wounding. Both processes activate localized mechanisms for forming protective barriers through lignification, cross-linking of cell wall proteins and other modifications of the extracellular matrix [10,11], which are thought to limit pathogen access. The HR and wound response display extensive overlaps in their transcriptional responses [12] and pathogen-derived elicitors of the HR can activate wound-induced kinase activities [13] suggesting that the two responses share some regulatory components. Thus, both wound and hypersensitive responses activate related defenses local to and transmitted from their primary sites of initiation, but differ in their activating signals.
There has been intensive research on the signaling mechanisms that regulate the myriad responses elicited by wounds and pathogens. Several lines of evidence suggest that the localized and transmitted components of both responses are regulated, in part, by hydrogen peroxide-mediated signaling events [14-17]. It has been proposed that H202 is a broad spectrum signaling molecule that triggers local processes like cell wall protein cross linking [14] and cell death [6] as well as long distance effects such as gene induction [11]. Although the precise mechanisms by which H202 signals are initiated locally and then transmitted are incompletely known, they involve activation of an NADPH oxidase that generates H202 via superoxide production [18], analogous to the oxidative defenses mounted by macrophages of the mammalian immune system. NO signaling also appears to participate in this oxidative response [19,20].
In contrast to the signal-mediated events that trigger these responses, relatively little is known about the downstream events that execute the orderly patterns of cellular deconstruction that accompany cell death. In animal cells many of the characteristic events are attributed to the activity of caspase proteases, which initiate and execute a cascade of proteolytic events that participate in subcellular deconstruction [21]. Parallels between plant cell death and animal apoptosis have been suggested from observations of cellular contraction, nuclear contraction and fragmentation of DNA during HR cell death. Plants have a family of caspase-related proteins, designated as metacaspases, and numerous studies have implicated caspase-like proteases in the control of cell death activation in plants [22]. Recently, a vacuolar protease unrelated to caspases at the amino acid sequence level has been found to posess caspase protease activity and has been demonstrated to be required for viral induced hypersentive cell death in tobacco [23], showing that caspase activities mediate some components of plant cell death. Nonetheless, the targets of this and other caspases that mediate the changes in subcellular architecture during plant cell death still remain elusive.
In the course of a screen for useful GFP fusion proteins [24], we observed that a GFP-Nitrilase 1 fusion protein exhibited a change of aggregation state in response to wounding. In order to place this phenomenon in a larger context, we initiated a series of live-cell imaging experiments to identify and characterize markers associated with the execution of physically and chemically induced cell death. Our studies reveal a series of previously undescribed events associated with wound-induced cell death in plants and raise new mechanistic questions regarding the execution of plant cell death.
Results
GFP-Nit1 aggregation marks an early step of wound induced cell death
Our investigations of cell death arose from the discovery of a wound responsive GFP marker, GFP-Nit1, which was identified in a previous random marker screen [24]. The GFP-Nit1 construct encodes the full length coding sequence for the enzyme nitrilase 1 fused to the C-terminus of the coding sequence for GFP. In non-wounded cells, the fusion protein was present throughout the cytoplasm and nucleoplasm (comparable to soluble GFP) and was visible as diffuse fluorescence (results not presented). Following a puncture wound, GFP-Nit1 rapidly translocates from the cytosol to organelle-associated aggregates in cells adjacent to (Fig. 1A,B) and at a distance from wounds (Fig. 1C, and see additional file 1), suggesting that the marker responds to wound induced signals. Based on observations described below, we consider the local response at the wound site (defined here as the cells directly abutting cells damaged by the wound) as distinct from the transmitted response, which can spread many cell layers from the wound site.
Figure 1 Localized and transmitted wound responses of the GFP-Nit1 marker. (A, B) Aggregation in cells abutting a wound site. A 35S GFP-Nit1 leaf petiole was severed by a razor blade incision at a 45° angle to the main axis of the petiole. Three mesophyll cells exposed by the wound are indicated by the white arrows (panel B) – note the extensive redistribution of the marker in these cells to organelle-associated aggregates, which is particularly evident around chloroplasts (red fluorescence). (C) Transmitted wound response. A 35S GFP-Nit1 seedling was mounted in 0.5X MS and punctured above the root meristem (wound site marked with an arrow). Imaging was initiated after an ~45 – 60 second delay between the wound and alignment of the root tip with field of view. Aggregates become evident several cell layers removed from the wound site (marked with an asterisk). Times shown are in minutes. The image shown in A and B are 3-dimensional reconstruction of a Z-series taken 5 min after wounding. Chloroplasts are visible because of their red autofluorescence. Images in C are single optical sections. Scale bars: A 10 μm; B, C 25 μm.
To characterize the local response, we imaged cells in the direct vicinity of puncture wounds administered to hypocotyls or cotyledon tissue of transgenic GFP-Nit1 plants. We observed that a variable number of cells in the wound site displayed GFP-Nit1 aggregates and rounded nuclei (Fig. 2A). Time lapse imaging of several wound-zones demonstrated that after forming GFP-Nit1 aggregates, these cells displayed nuclear contraction (Fig. 2A) followed by dramatic cellular collapse (Fig. 2B and see additional file2) (n = 6 independent wound experiments). These two responses (cellular collapse and nuclear contraction) are morphological hallmarks of cell death during the HR [11,25,27], and suggested that the local GFP-Nit1 response marks one component of a more general wound-triggered cell death response. Unlike the local response, we found that the transmitted response was reversible, with cytoplasmic streaming resuming concomitant with disappearance of GFP-Nit1 aggregates (see additional file 3). Thus, our observations suggest GFP-Nit1 aggregation does not mark a point of commitment to cell death in all cells. It does however mark an early stage of a distinct cell death process that operates in cells abutting the wound site.
Figure 2 Nuclear and cellular collapse during the wound response. The GFP-Nit1 line N1P2E was wounded and a confocal Z-series was immediately collected at 60 sec intervals from a region abutting the wound. (A) Nuclear contraction of a nucleus (adjacent to the white arrow) in a wound-proximal cotyledon epidermal cell that after wounding displayed the aggregation response (aggregates indicated by red arrows) then collapsed. (B) Cellular collapse by one of several cells in the wound site that displayed GFP-Nit1 aggregates and contractions. The red arrow points to the edge of the contracting cell and the white arrow to the wall of an adjacent cell. Note the detachment of the cell from the wall (most evident at 48 min after wounding). (C) Reversibility of transmitted response. An agarose-mounted 35S GFP-Nit1 plant was wounded above the root meristem and imaged by the acquisition of Z-series at 2 min intervals for 60 min (see supplemental data for complete image series). In the first time point after wounding, the aggregation response had spread throughout much of the root meristem. Over time, the aggregates reverted back to a cytoplasmic distribution pattern and cytoplasmic streaming was evident. Images in A are single confocal optical sections, B and C are reconstructions from Z-series stacks. The numbers in the lower right of each panel indicate time in min from the start of imaging. Scale bars A, B = 25 μm; C = 20 μm.
To explore the possibility that the GFP-Nit1 fusion protein might induce an aberrant cell-death response, we examined whether the process occurred independently of the GFP-Nit1 fusion protein. We examined the nuclear contraction observed with GFP-Nit1 during the wound response using a previously isolated nuclear marker (N6) [24]. Hypocotyls of N6 GFP transgenic plants were wounded and cells adjacent to wounds continuously imaged over a one-hour period (Fig. 3). As observed with GFP-Nit1, a subset of cells in the wound site possessed swollen, hypertrophic nuclei that progressed to a contracted state (Fig. 3A and see additional file 4). Propidium iodide was included in the imaging buffer to monitor plasma membrane integrity during the response. Intense staining of nuclei occurred subsequent to the major nuclear contraction, suggesting that degeneration of plasma membrane integrity is a late step. Thus nuclear contraction during wound induced cell death is not an artifact induced by the GFP-Nit1 fusion protein, a conclusion supported by results with another nuclear marker and markers of the ER (see below). In addition, as noted below, we have developed a biochemical assay to probe Nit1 relocalization aggregation and find that this occurs in wild type plants (i.e. in the complete absence of any GFP fusion protein).
Figure 3 Wound induced nuclear collapse in nuclear marker lines. (A) Nuclear response in line N6. Following wounding of a hypocotyl, the nucleus (arrow) first swells then contracts. The asterisk at time = 0 illustrates the normal lentoid shape of hypocotyl nuclei. Contraction is followed by a decrease in GFP fluorescence, most evident between the 16 and 20-min time points. This is followed later by intense nuclear propidium iodide staining (shown in red, adjacent to arrow). (B) Nuclear response in a wounded hypocotyl of an N7 line. Contraction of the nucleus (red arrow), decrease in nuclear fluorescence and release of nuclear fluorescence into the cytoplasm are evident. Note the formation of lobes (marked with a white arrow) around the contracting nucleus (red arrow). The time series in A and B are brightest-point reconstructions. Times shown are in min. Scale bars: A = 25 μm, B = 10 μm.
Nuclear leakage and lobing accompany nuclear contractions
In addition to nuclear contraction, a wound-induced decrease in the intensity of GFP nuclear fluorescence in the N6 marker line was evident within less than 20 min of wounding (Fig. 3A). Several explanations for the decrease in nuclear fluorescence seemed possible, including GFP destruction, alterations in pH that quenched fluorescence, or the expulsion of nuclear contents into the cytoplasm. While these possibilities are not mutually exclusive, the latter seemed particularly attractive since contraction might cause extrusion of nuclear contents. The N6 marker possesses a low background level of cytoplasmic GFP localization that might mask the release of nuclear contents and also may be partially immobilized from the bulk nucleoplasm by an association with DNA (this marker illuminates chromosomes during mitosis, [26]). Consequently, we examined wound responses using the nuclear marker line N7, which displays a lower level of background cytoplasmic fluorescence and does not appear to be associated with chromatin as assessed by time-lapse imaging of mitoses (data not shown). Hypocotyls of plants carrying the N7 marker were wounded and imaged. Time-lapse imaging of cells proximal to wound sites revealed that a gradual illumination of the cytoplasm occurs concomitantly with the decrease in nuclear fluorescence and nuclear contraction, suggesting that the nuclear label is lost or expelled into the cytoplasm (Fig. 3B). The simultaneous illumination of nucleoplasm and cytoplasm facilitated visualization of the formation of lobes on the nuclei, revealed by negative contrast, as the interior of these lobes excludes the GFP label (See arrows in Fig. 3).
Lobing is caused by nuclear envelope separation
Since the nucleus is bound by the nuclear envelope, a membrane system contiguous with the ER (for review of the plant ER see [28]) lobing could conceivably occur if the inner membrane of the nuclear envelope contracted without a corresponding contraction of the outer membrane – i.e. if the two membranes separated. Exclusion of cytosolic GFP from the interior of the lobes might be expected if this were true, since the space between the inner and outer nuclear envelopes, the peri-nuclear space, is contiguous with the lumen of the ER and thus isolated from the cytoplasm. This hypothesis is supported by observations of two ER markers, Q4 an ER membrane marker and mGFP5, a KDEL-tagged lumenal ER marker. Figure 4 displays time points from a time-lapse wound imaging experiment using the Q4 marker line. Figure 4A shows three-dimensional reconstructions of one image series, and illustrates the characteristic lobes that form during nuclear contraction. Consistent with the hypothesis of nuclear envelope separation, optical sections through the mid-plane of the same contracting nucleus shown in 4A reveal a clear double membrane structure that separates concomitantly with nuclear contraction (Fig. 4B). We next tested the behavior of an ER-lumenal GFP marker (mGFP5) during nuclear contraction and lobing and found that nuclei at differing stages of contraction and lobing contain the ER-lumenal marker mGFP5 within the inter-lobal space (Fig. 4D,E).
Figure 4 Nuclear lobes are bound by the ER membrane and contain ER lumen contents. A hypocotyl of the ER membrane marker line Q4 was wounded and imaged at 2-minute intervals by confocal microscopy. The simultaneous contraction and lobing of the nucleus are evident as a separation of the nuclear envelope. (A) 3D reconstruction of the contracting nucleus made with a brightest-point reconstruction of the acquired data set. (B) The same data set as in Fig. 4A at a single optical section through the mid-plane of the contracting nucleus, illuminating the double membrane structure (pointed to by arrows) and its separation. (C) A single optical section through the contracted nucleus shows propidium iodide staining (shown in red) of the interior, demonstrating that the nuclear lumen is interior to the lobes. (D, E) Hypocotyls of a line expressing the ER-lumenal GFP marker mGFP5 were wounded. Shown are contracting epidermal nuclei (white arrows) from two independent wound experiments using mGFP5 approximately 10 min after wounding. The inter-lobal space contains mGFP5 label (red arrows), indicating that this compartment is contiguous with the lumen of the ER. GFP fluorescence is shown in green and propidium iodide in red. Scale bars = 10 μm.
Based on our observations that the inter-lobal space excludes cytosolic GFP, contains the lumenal ER marker mGFP5, and that the lobes are demarcated by a membrane system contiguous with the ER, we believe that the simplest explanation for the observed nuclear lobing is that the inner and outer membranes of the nuclear envelope separate during nuclear contraction. Thus, nuclear contraction does not reflect a contraction of the entire nucleus proper, but rather a contraction of the nuclear lumen and its tightly associated inner envelope.
Degeneration of the cortical-ER network accompanies cell death
Both the ER tubules and the nuclear envelope contain closely appressed double-membrane structures, assemble with similar properties in vitro [28,43], and are contiguous membrane systems. Thus, the death response might be expected to additionally alter cortical ER tubules. To examine this, we imaged wound-proximal cells of the Q4 ER-membrane marker line (Fig. 5 and Additional file 5). About 20 min after wounding, extensive nuclear contraction and lobing was evident (marked with an asterisk in Fig. 5), part of the ER membrane system had formed bubble-like structures but much of the ER-tubular network was still intact. Shortly thereafter, the tubular ER network rapidly degenerated, contracting from tubules into small vesicular structures (Fig. 5). Thus, alterations in the integrity of both the nuclear envelope and cortical ER tubules occur during the local wound-induced cell death response, although apparently with different kinetics. Prior to ER tubule vesiculation, but concurrent with nuclear contraction and lobing, we observed that plastids in the cell moved away from the cell cortex (data not shown). Since plastid movements are dependent on a well-characterized actin-based motility system [29] these observations suggest that this system is at least partly intact during the earlier stages of wound induced cell death. Further investigation of this awaits imaging experiments using cytoskeletal markers.
Figure 5 ER vesiculation during wound induced cell death. A wound-proximal petiole epidermal cell of ER membrane marker line Q4 was imaged. The panel shows time points (in min) after nuclear contraction. Prior to the times shown, part of the ER network formed bubble-like structures rapidly after wounding. The remaining intact tubular ER subsequently degenerated, forming large numbers of vesicular structures. Images are brightest point reconstructions of Z-series collected at 2 minutes intervals after wounding. Times shown are in minutes. Scale bar = 10 μm.
Benzonitrile herbicides induce rapid cell death
To probe the specificity of the local wound responses we sought a simple method for chemically inducing cell death on a time frame similar to that used for our wounding experiments (45 – 90 min). Many chemically-induced forms of cell death have been characterized, but typically their effects operate on the time scale of hours to days as opposed to minutes. To identify a fast acting molecule we used the GFP-Nit1 aggregation response to screen a number of herbicides and other small-molecules for rapid activity, focusing on nitriles, which have been used to trigger cell death in other experiments (i.e., KCN, [5]). Potassium cyanide, while obviously toxic, did not induce the rapid GFP-Nit1 aggregation, cell death and cellular collapse on a 1 hour time scale that we sought [30]. We found that the hydroxy-benzonitrile herbicide bromoxynil (2,4 di-bromo-para-hydroxybenzonitrile), and its analog chloroxynil (2,4 di-chloro-para-hydroxybenzonitrile), caused rapid cell death when applied to Arabidopsis tissue at concentrations between 250 – 500 μM. Within 1 hour of application to seedlings bromoxynil and chloroxynil induced rapid GFP-Nit1 aggregation (Fig. 6A), extensive cell shrinkage and cell death (additional file 6 ). Like wounding, these herbicides also induced callose deposition (Fig. 6B), which was evident as early as 8 min after application to roots. The specific mode of action of these herbicides is unclear but they have well documented effects on mitochondrial respiration [31,32] and also bind to photosystem II proteins [33]. Whatever their mechanism of action, they proved convenient rapid inducers of woundless cell death for use in our imaging experiments.
Figure 6 GFP-Nit1 aggregation and callose induction by bromoxynil. (A) 8 day old GFP-Nit1 plants (N1P2E) seedlings were imbibed in 500 μM Bromoxynil (+Bx) or mock solutions (-Bx) for 1 h then imaged by confocal microscopy, note the exstensive accumulation of the fusion protein in aggregates after herbicide treatment. (B) Callose accumulation following bromoxynil treatment. Wild type seedlings were treated with 500 μm bromoxynil (+Bx) or a mock-solution (-Bx). Roots were examined at regular intervals under UV illumination using conventional epifluorescence and a DAPI filter set; the 15 min time point is shown. Callose deposition was visualized using the callose specific stain sirofluor, added at 100 ug/ml at the beginning of the experiment. Sirofluor fluorescence was first visible approximately 8 min after addition of bromoxynil. Images were acquired on a digital camera using identical exposure settings.
Nuclear and cellular contraction can be uncoupled during cell death
To probe nuclear dynamics during chemically-induced cell death we examined hypocotyl cells of the N7 marker line after chloroxynil application. Time lapse imaging experiments showed that prior to the late stage inductions of cellular collapse and propidium iodide staining of nuclei, the N7 label leaves the nuclei, illuminating the cytoplasm analogously to wound induced cell death (Fig. 7B). However the characteristic lentoid shape of the hypocotyl nuclei persisted throughout, demonstrating that this hallmark of wound induced cell death does not occur during this particular type of chemically induced cell death (Fig. 7A). Notable in these experiments was the rapid loss of the GFP fluorescence, which was not seen to the same extent in our wounding experiments. We next examined the response in root cells, which also failed to show nuclear contraction but clearly show nuclear leakage of the GFP marker (Fig. 7B and additional file 6). Thus, in both cell types, nuclear leakage occurs without concomitant contractions of the nucleus, showing that these two responses can be separated. Importantly, this observation suggests that the simple hypothesis that nuclear leakage is caused by contraction-induced extrusion is unlikely.
Figure 7 Chloroxynil induced death – collapse of cells without nuclear contractions. (A) The nuclear marker line N7 was mounted in agarose, and the bromoxynil-analog chloroxynil was added to a final concentration of 500 μM and a Z-series of the hypocotyl-root junction was collected at 2 min intervals. Loss of nuclear label into the cytoplasm and a generalized loss of GFP fluorescence were followed by staining of nuclei by propidium iodide and massive cellular collapse. The shape of the nuclei remained constant throughout the experiment and did not display the characteristic contractions seen during wound induced cell death. The white asterisk identifies the same nucleus, and the green arrows to two cells in this plane of view that displayed massive plasmolysis. (B) Chloroxynil induced release of N7 marker from nucleoplasm in 12 day old N7 plants mounted in agarose and overlaid with chloroxynil to a final concentration of 500 μM. An initiating lateral root was imaged by the acquisition of Z-series at 120 sec intervals. A gradual release of nuclear label and illumination of the cytoplasm occurs without the characteristic nuclear contractions induced by wounding. Scale bars = 25 μm. Inset numbers indicate min after herbicide addition.
Fusiform-body spherulation accompanies wounding and chemical cell death
Markers of the ER in Arabidopsis illuminate a cigar-shaped accessory organelle called the fusiform body [34,35]. Recent evidence suggests that this organelle may be involved in cell death, but its specific functions are still unclear [36]. This structure is evident in Arabidopsis plants expressing the ER markers mGFP5 and Q4 (Fig. 8A). In our time-lapse observations of ER tubule and nuclear envelope degeneration using the Q4 marker line, the fusiform bodies were noticeably absent. This prompted us to examine them more closely using mGFP5, which brightly illuminates these structures [37,35]. In wound-proximal mGFP5-expressing cells, the most evident structures from the first time points after wounding are spheres, with fusiform bodies absent (Fig. 8A). This observation suggests that the fusiform bodies rapidly lose their characteristic morphology in response to wounding. To examine if this response occurs during herbicide-induced death, we imaged the Q4 ER-membrane marker line after application of chloroxynil. Shortly after addition of chloroxynil, cytoplasmic streaming stopped and this was quickly followed by the transformation of fusiform bodies into spherical structures (Fig. 8B,C and additional file 7). Thus, both wounding and herbicide treatment act at an early stage to effect component(s) required to maintain the distinctive shape of this organelle.
Figure 8 ER Fusiform body alterations during wound and chloroxynil induced cell death. (A) Hypocotyl cell of an unwounded plant showing the normal morphology of the fusiform bodies. (B) Wound proximal hypocotyl cell. Note the spherical bodies in this cell, distinct from the normally cigar-shaped structures illuminated by lumenal ER markers. (C) Chloroxynil-induced transitions. Epidermal hypocotyl cells were treated with 500 μM chloroxynil. Inset numbers indicate time (in minutes) after herbicide addition. An individual fusiform body throughout the experiment is marked by the red arrow. By 4 min after herbicide application, streaming ceases, evident by the lack of movement of fusiform bodies and the tubular ER network, which are usually highly dynamic. Shortly thereafter, fusiform bodies transform into spherical structures. Scale bar = 10 μm.
GFP-Nit1 aggregates translocate into a pellet fraction during cell death
The extensive aggregation of GFP-Nit1 triggered by benzonitrile herbicides facilitated biochemical characterization of the process. GFP-NIT1 in unwounded plants is soluble and does not pellet during low-speed centrifugation (Fig 9). By contrast, we observed that the aggregates in GFP-Nit1 plants were stable during tissue disruption by sonication and partitioned into a low-speed Triton-X 100 insoluble pellet fraction by centrifugation (Fig. 9A). This low-speed pellet fraction was also found to contain the actin-binding protein cofillin, but not the membrane protein PIP2A, which resided in the low-speed supernatant; we therefore used these two proteins as markers for each fraction and as loading controls (Fig. 9B). The low-speed pelleting of Nit1 enabled us to test for herbicide-induced change in the solubility of nitrilase in wild type plants. A polyclonal antibody to recombinant Nit1 was produced and used to probe proteins isolated from the nit1-3 null mutant [38]. The crude serum recognizes two proteins, one of which is absent in the nit1-3 mutant (Fig. 9B), showing that the antibody faithfully recognizes Nit1 protein, the identity of the upper band is unknown and was not detected after affinity purification of the antibody against immobilized recombinant Nit1. Affinity-purified antibody was used to probe plants that were treated with bromoxynil under conditions that caused extensive aggregation in the GFP-Nit1 transgenic lines. In wild type plants treated with bromoxynil at concentrations that triggered aggregation of GFP-Nit1 in vivo, the Nit1 protein partitioned into the pellet fraction (Fig. 9C). Western blots of the GFP-Nit1 fusion protein showed comparable accumulation of the fusion protein in the low-speed pellet fraction (data not shown). These experiments provide an independent biochemical assay for a bromoxynil-induced change in solubility of Nit1 and suggest that the bromoxynil- and wound-induced aggregation of the GFP-Nit1 marker is not a fusion protein artifact.
Figure 9 Biochemical analysis of Nit1 aggregation. (A) Extracts from bromoxynil-treated plants were made 4% in Triton X100 and centrifuged for 10 min at 10,000 g and the pellet was imaged by confocal microscopy. (B) Western blot of protein extracts from leaf tissue of wild type and the Nit1-3 mutant showing that the antibody recognizes the Nit1 protein in wild type but not the null mutant. (C) Aggregation of Nit1 into the low speed pellet fraction during herbicide-induced cell death. Three week old seedlings were imbibed in solutions of bromoxynil (0, 125, 250, 500 μM) for 1 h then extracted, adjusted to 4% Triton X100 and centrifuged at 10,000 g for 10 min. Proteins in the supernatant and pellet were separated by SDS-PAGE and various proteins detected by western blotting. Antibodies against PIP2A (plasma membrane intrinsic protein) and cofillin were used to show complete separation of the two fractions.
Discussion
This study was stimulated by the fortuitous observation that a GFP-Nit1 fusion protein rapidly underwent a change of state, from soluble to granular, in cells adjacent to a puncture wound. Because of technical difficulties associated with reproducibly producing puncture wounds, we surveyed a number of chemicals for their ability to induce GFP-Nit1 aggregation. We found that several benzonitrile herbicides caused effects that were similar to puncture wounds. Using a variety of GFP-based markers, we observed that cells abutting the wound site, or following treatment with benzonitriles, display a suite of dynamic morphological events that culminate in cellular contraction and cell death. Using an ER membrane marker we were able to simultaneously observe alterations in cortical ER tubules, fusiform bodies, the nuclear envelope and the nucleus. Similarly, using GFP-Nit1, we observed that the formation of nitrilase aggregates precedes nuclear and cellular contraction. Collectively, our data is consistent with a series of events that we have synthesized into a descriptive scheme (Fig. 10).
Figure 10 Cytological events associated with wound-induced cell death. (A) Unwounded cell. Fusiform bodies outlined in blue, cytoplasmic GFP-Nit1 in red, ER and nuclear envelope membranes shown in green, ER lumen contents shown in gray and nuclear lumenal contents shown in yellow, plasma membrane in black. (B – F) Post-wound events. (B)The earliest events documented are the collapse of fusiform bodies and nuclear hypertrophy. (C) Shortly after these early rapid events, aggregation of GFP-Nit1 occurs and the nucleus initiates contraction concurrent with separation of the nuclear envelope. (D) As contraction and lobing continue, contents of the nucleoplasm leak into the cytosol. It is currently unknown if this release is general or if there is specificity to the particular contents lost. (E) Later, after nuclear contraction has largely ceased, the tubular ER network that is still intact undergoes a dramatic loss of integrity and degenerates, forming vesicular structures. (F) This is followed later by intense staining of nuclei by propidium iodide and extensive cell shrinkage. Propidium iodide staining of nuclei is indicated in (F) by the red color of the nucleus.
The earliest events after wounding were fusiform body alterations, aggregation of GFP-Nit1, and nuclear hypertrophy. Following herbicide treatment, we observed that fusiform body alterations preceded GFP-Nit1 aggregation (data not shown). Shortly after these early events, nuclear contraction ensues with a concomitant separation of the nuclear envelope and nucleoplasm leakage into the cytosol. It is currently unknown if this reflects a generalized loss of nucleoplasm or if there is specificity to the contents lost. Later, after nuclear contraction has largely ceased, the remaining tubular ER undergoes a dramatic loss of integrity and degenerates, forming vesicular structures. Intense staining of nuclei by propidium iodide and extensive cell shrinkage follows this. Since propidium iodide is a charged fluorophore that is largely precluded from entering vital cells, we interpret this late event to represent a major degeneration in the integrity of the plasma membrane and consider this the likely point of cell death, although this is an operational definition. Observations of a plasma-membrane marker suggest that the plasma membrane is intact at a gross level during the early stages of the wound response, although atypical involutions and extreme photo-sensitivity have been observed (data not shown).
The events we documented occur on a relatively rapid time scale. Typically, within minutes of wounding we observed detectable effects on the shapes of nuclei, fusiform bodies and the aggregation of GFP-Nit1. Within one to two hours of wounding, cells displayed intense staining of their contracted nuclei by propidium iodide. Although we did not obtain evidence for programmed cell death (PCD), this rapid response is compatible with evidence that plant cells undergo PCD by constitutively expressed molecular machinery [39]. It seems plausible that, by limiting possible pathogen access through a wound site, wound-induced cell death functions for the benefit of the organism as a whole.
Although we are far from a mechanistic picture of these events, our observations show that nuclear contraction can be uncoupled from cellular collapse. The known effects of benzonitrile herbicides on mitochondrial function may suggest that nuclear contraction is energy dependent. Since cellular collapse occurred after chloroxynil treatment, this may suggest that contraction is energy independent, perhaps unsurprising, since turgor pressure is generated in part through vacuolar ATPases and proton pumps.
Our focus in this study was on placing the aggregation of GFP-Nit1 in the context of cellular responses during wound-induced cell death. We could not identify any precedent in the literature for a similar transformation. We do not understand the mechanism by which GFP-Nit1 is converted from a soluble protein to a granular form. The observation that nitrilase 1 undergoes a change in sedimentation coefficient after chemically-induced injury in wild type plants suggests that the phenomenon is not an artifact of the GFP fusion protein. We do not know if nitrilase 1 is the only protein that exhibits such properties. However, we did not observe a similar response during the characterization of several hundred other GFP-protein fusions [24] (and unpublished results). The observation that the GFP-Nit1 aggregates have a relatively constant size that is roughly comparable to that of a secretory vesicle may suggest that the basis of aggregation is encapsulation or binding of nitrilase to a vesicular type of structure. However, the observation that treatment with Triton X100 does not disperse the aggregates indicates that they are not simple membrane-bound vesicles.
Conclusion
Wounding and benzonitrile herbicides both induce rapid forms of cell death that yield collapsed cells with disrupted plasma membranes. One of the earliest responses in both of these cell death systems is the redistribution of Nit1 into aggregates that are visible in vivo as GFP-Nit1 granules or biochemically as a low-speed pelletable form of Nit1. Nit1 aggregation precedes nuclear contraction and cellular collapse, two classic cytological features of plant cell death. The use of GFP-Nit1 and other markers to image plant cell death in vivo has revealed novel subcellular responses that to our knowledge have not been previously described, such as nuclear envelope separation, formation of nuclear lobes and release of nuclear contents into the cytosol. Our data have enabled a new detailed descriptive model for plant cell death and raise new mechanistic questions.
Methods
Plant growth and preparation for microscopy
Plants were grown on agar-solidified media consisting of 0.5X MS salts (Gibco-BRL) and 0.8 % Agar (Research Organics). Prior to germination, seeds were chilled for 4 days on MS plates at 4°C then transferred to a growth cabinet and grown under continuous illumination at 300 μE m-2s-1. Unless otherwise stated, imaging experiments were performed on whole 4 – 8 day old seedlings mounted in 0.5X MS or an imaging buffer (IMB) composed of 0.5X MS salts, 25 μg/ml propidium iodide (Sigma) and 0.01% Silwet (Lehle Seeds, Tucson, AZ), which was added to facilitate propidium iodide penetration of the epidermal cuticle. To prevent specimen drift during time lapse experiments, whole seedlings in 0.5X MS were mounted by adding 2 volumes of molten 2% low melting point agarose (Research Organics), 2% high resolution 3:1 agarose (FMC) and 0.5X MS salts (Gibco-BRL) at 42°C. Cells showed active streaming after mounting, suggesting minimal stress induced by the mounting process.
Targeted GFP lines
Most of the transgenic lines used for imaging in this paper have been described previously [24,37], but are described here for clarity. For GFP-cDNA lines, experiments were performed on T3 seed derived from the primary transgenic plants. Unless indicated otherwise, these lines contained a single GFP-cDNA fusion. The GFP fusions are available from the Arabidopsis Biological Resource Center (ABRC) at Ohio State University .
Nuclear markers
The N7 marker line (ABRC #CS84731) contains a fusion protein between GFP and the carboxy terminus of an ankyrin-repeat containing transcription-factor-like protein (Genbank Accession CAA16704). The N6 marker line (ABRC #CS84815) contains a near full-length fusion between an HMG-delta protein (Genbank accession Y14075) and GFP. In dividing cells it illuminates chromosomes aligned along the metaphase plate and thus likely associates with chromatin (26). In interphase cells, it illuminates the nucleoplasm. This line also contains a second PCR-detectable insert that contains an out-of-frame cDNA fusion to GFP.
ER markers
The mGFP5 line was generated by Wolf Scheible by transformation of the mGFP5 construct on plasmid pmGFP5-ER [37] into the Columbia ecotype of Arabidopsis. The Q4 ER membrane marker line (ABRC #CS84728) is a fusion between GFP and a novel protein with a predicted carboxy-terminal trans-membrane (Genbank Accession AAB71445).
Wound induced aggregates
The GFP-Nit1 marker comprises a full-length nitrilase 1 [38] (Genbank Accession U47114) fused to the C-terminus of GFP. Two lines were used in this study 35S-GFP-Nit1, in which the fusion protein is driven by the 35S promoter and, N1P2E, in which expression is driven by 1.8 kb of sequence upstream from the Nit1 start codon [30]. These two different lines possess similar expression levels in transgenic plants, however in comparison to 35S-GFP-Nit1, the N1P2E line shows reduced expression in root epidermal cells and guard cells
Imaging
Agarose mounted seedlings were wounded by creating cuts through plant tissue with a razor blade or sharp forceps tips. Image data was collected after a brief (45 – 60 second) period of aligning the wound site with the field of view. During and prior to imaging experiments, agarose embedded specimens were covered with a humidifying dome to prevent desiccation. The majority of image series were obtained by collecting 20 – 25 μM deep Z-series (typically 15X 1.5 μm z-steps) for 30 – 60 time points at 120 sec intervals. 3-D time series were made from these data sets by making brightest-point reconstructions of each Z-series using the BioRad software package LaserSharp (BioRad, Hercules, CA). For some time series, reconstructions were performed manually in NIH image using brightest point reconstructions (Wayne Rasband, RSB, NIH, Bethesda Maryland). A Nikon inverted fluorescence microscope equipped with a Nikon 60X 1.2 numerical aperture water immersion objective and a Bio-Rad MRC 1024 confocal head with a krypton-argon laser. EGFP was excited at 488 nm and emitted fluorescence was collected through a 525-30 band pass filter. Chloroplast autofluorescence and propidium iodide fluorescence were obtained by excitation at 568 nm and collecting emitted fluorescence through a 596 – 615 nm band pass filter.
Chemical treatments
Chloroxynil has a higher solubility in ethanol and water than bromoxynil and was used in imaging experiments. Both herbicides show similar toxicity to Arabidopsis using a germination assay [30]. A 1 M stock of Chloroxynil (ChemServices, West Chester, PA) was made in ethanol. For imaging experiments, a 1 mM solution of Chloroxynil in IMB was made and 100 μl of this solution was overlaid onto to seedlings mounted in 100 μl of agarose, yielding a final concentration of 500 μM.
Nit1 antibody production
Recombinant Nit1 was produced in E. coli as a His-tagged thioredoxin fusion protein using the vector pET32(A) (Novagen, Madison, WI) and purified by chromatography on a nickel-chelate resin, (Novagen, Madison, WI). Polyclonal serum against recombinant Nit1 protein was made by a commercial provider (Cocalico, Reamstown, PA). The Nit1 antibodies were affinity purified against 2 mg of purified recombinant Nit1 adsorbed onto a 6 × 8 cm piece of nitrocellulose membrane and eluted off the membrane by treatment with 200 mM glycine, 150 mM NaCl pH 2.0. The eluted antibodies were neutralized to pH 7.0 with Tris base, dialysed overnight against Tris-buffered saline and concentrated to 2 mg/ml using a spin column concentrator with a 15 kDa cutoff, the concentrated antibody was diluted to 1 mg/ml with glycerol.
Nit1 pelleting assays
Herbicide-treated plants were transferred into a 3X-weight volume of ice-cold granule isolation buffer (GEB, 400 mM sucrose, 75 mM KCl, 50 mM PIPES pH 6.9, 10 mM MgCl2, 1 mM EGTA, 1 mM DTT, 1 mM PMSF, 1 μg/ml leupeptin, 1 μg/ml aprotinin). They were then sonicated, filtered through Miracloth and Triton-X 100 was added to a final concentration of 4%. After 10 min incubation on ice, the extract was centrifuged at 10,000 g for 10 min. The supernatant was removed and retained and the pellet was washed by with GEB + 4% Triton-X 100. Protein from the supernatant fraction was precipitated using a chloroform-methanol extraction protocol [40]. The pellet fraction and precipitated low-speed supernatant fractions were resuspended in equal volumes of SDS-PAGE sample buffer, typically 500 μl per 100 mg input tissue (fresh weight).
For herbicide induced-pelleting, 2–3 week old Arabidopsis seedlings were transferred into a solution of ddH20, 0.01% Silwet and varying doses of Bromoxynil. After 1 h incubation at room temperature with gentle shaking, seedlings were transferred to an eppendorf tube, placed on ice and disrupted by sonication in GEB as described above.
Protein analysis
SDS-PAGE was performed using 8 – 20 % gradient mini-gels, prepared and run as described in Ausubel et al. [41]. For western analysis, proteins were transferred onto nitrocellulose membranes using a semi-dry electroblotter in Towbin transfer buffer [42]. Westerns were developed using Pierce SuperSignal HRP substrate (Pierce, Rockville, IL). The cofillin and Pip2A used as control antibodies were obtained from Rose Biotechnology [44] and used at 1/2000 dilutions. The Nit1 antibodies were used at 1/1000 dilution.
Supplementary Material
Additional File 1
Wound induced granulation of GFP-Nit1 demonstrating action at a distance A 35S GFP-Nit1 root was wounded above the root meristem with a syringe tip, quickly aligned for microscopy and then imaged at a single focal plane at 4 second intervals for 8 minutes (the wound site is out of the focal plane of the time series, but is visible as the dark area of separated cells in the center of the root tip). Initially, the GFP-Nit1marker is visible as diffuse fluorescence. Over the course of several minutes, bright granular structures become evident concomitant with a decrease in cytoplasmic fluorescence, suggesting a redistribution of the marker into granules. The total time of the image series is 8 minutes. In other time series which have imaged the process over a longer time period, we have observed reversion of the marker to cytoplasmic distribution with resumption of cytoplasmic streaming (see additional file3). The action-at-a distance observed with this marker suggests that the marker is responsive to a motile signal is sent after wounding. The data shown in figure1A were derived from this data set.
Click here for file
Additional File 2
Collapse of cells within the wound zone – wound induced granulation of GFP-Nit1 followed by cellular contraction and detachment from neighboring cells The cotyledons of agarose imbedded GFP-Nit1 lines (N1P2E) were wounded with razor incisions and subsequently imaged by collecting confocal Z-series of a region abutting the wound at 60 second intervals. (2A) Shown is a detail of a time-series made from brightest point 3-dimensional reconstructions of the acquired Z-series. Immediately after wounding, the center cell shown in the image series displayed GFP-Nit1 granules. After a short period of time, the cell started contracting with a extensive detachment of the cell from its cell walls. (2B) The full frame image data that Movie 2A was derived from. Note the numerous cells in the wound zone displaying granulation after wounding and their subsequent contraction over the course of the time series. Figure 2B in the manuscript was derived from this data set (see additional file 8).
Click here for file
Additional File 3
Reversion of GFP-Nit1 granulation during the transmitted response An agarose mounted 35S GFP-Nit1 plant was wounded above the root meristem and imaged by the acquisition of Z-series at 2 minute intervals for 90 minutes. The wound site was above the field of view. By the time of acquiring the first time point, the granulation response had spread throughout much of the root meristem. Over time, the granules revert back to a cytoplasmic distribution pattern and the cells resume cytoplasmic streaming. Images are brightest point reconstructions of Z-series collected at 2 minutes intervals after wounding.
Click here for file
Additional File 4
Nuclear contraction during wound induced cell death The hypocotyl of an agarose imbedded nuclear marker line (N6) was wounded with a razor blade and them imaged at 2 minute intervals by collecting confocal Z-series. Propidium iodide was included in the imaging buffer to monitor plasma membrane integrity. Following wounding,, the nucleus on the left displays hypertrophy relative to the normal lentoid shape of unwounded cells (see the nuclei in cells on the right hand side in the movie). Over the course of the time series, the nucleus contracts concomitant with a decrease in GFP fluorescence, most evident in the time series between the 16 and 20 minute time points shown. This is followed later by intense staining of the contracted nucleus by propidium iodide.
Click here for file
Additional File 5
Degeneration of the cortical ER network during wound induced cell death A wound-proximal petiole epidermal cell of the ER membrane marker line Q4 was imaged. Part of the ER network formed bubble-like structures rapidly following wounding, and the remaining intact tubular ER then degenerates, forming large numbers of vesicular structures. Images are brightest point reconstructions of Z-series collected at 2 minutes intervals after wounding.
Click here for file
Additional File 6
Chloroxynil induced cell death – extensive cellular collapse without nuclear contraction The nuclear marker line N7 was mounted in agarose, and chloroxynil was added (final concentration 500 μM) and imaging was initiated collecting Z-series of the hypocotyl-root junction at 2 minute intervals. Loss of nuclear label into the cytoplasm and a generalized loss of GFP fluorescence were followed by staining of nuclei by propidium iodide and massive cellular collapse. The shape of the nuclei remain constant throughout the experiment and do not display the characteristic contractions seen during wound induced cell death..
Click here for file
Additional File 7
Fusiform body alterations during Chloroxynil triggered cell death Epidermal hypocotyl cells of an agarose mounted Q4 seedling were imaged by acquiring z-series at 120 second intervals. After the three z-series had been acquired, chloroxynil was added to a final concentration of 500 μM. Active streaming is evident during the first 3 time points. By 4 minutes after herbicide application, streaming ceases, evident by the lack of movement of fusiform bodies and the tubular ER network, which are usually highly dynamic. Shortly thereafter, fusiform bodies transform into spherical structures, which also occurs during wound-induced cell death. The data shown in figure 8 were derived from this data set.
Click here for file
Additional File 8
Collapse of cells within the wound zone – wound induced granulation of GFP-Nit1 followed by cellular contraction and detachment from neighboring cells. The full frame image data that Movie 2A was derived from. Note the numerous cells in the wound zone displaying granulation after wounding and their subsequent contraction over the course of the time series. Figure 2B in the manuscript was derived from this data set.
Click here for file
Acknowledgements
We thank David Ehrhardt for assistance with confocal microscopy and for fruitful discussions. We also thank Farhah Assaad for valuable comments on the manuscript and figures. This work was supported by a grant from the U.S. Department of Energy (DOE-FG02-00ER20133).
==== Refs
Vaux DL Haecker G Strasser A An evolutionary perspective on apoptosis Cell 1994 76 777 779 8124715 10.1016/0092-8674(94)90350-6
Kerr JF Wyllie AH Currie AR Apoptosis: a basic biological phenomenon with wide-ranging implications in tissue kinetics Br J Cancer 1972 26 239 57 4561027
Lockshin RA Zakeri Z Tilly JL When cells die Wiley-Liss 1998
Fukuda H Programmed cell death of tracheary elements as a paradigm in plants Plant Mol Biol 2000 44 245 53 11199386 10.1023/A:1026532223173
Ryerson DE Heath MC Cleavage of nuclear DNA into oligonucleosomal fragments during cell death induced by fungal infection or by abiotic treatments Plant Cell 1996 8 393 402 12239388 10.1105/tpc.8.3.393
Levine A Pennell RI Alvarez ME Palmer R Lamb C Calcium-mediated apoptosis in a plant hypersensitive disease resistance response Curr Biol 1996 6 427 37 8723347 10.1016/S0960-9822(02)00510-9
Greenberg JT Programmed cell death in plant-pathogen interactions Annu Rev Plant Physiol Plant Mol Biol 1997 48 525 545 15012273 10.1146/annurev.arplant.48.1.525
Greenberg JT Guo A Klessig DF Ausubel FM Programmed cell death in plants: a pathogen-triggered response activated coordinately with multiple defense functions Cell 1994 77 551 63 8187175 10.1016/0092-8674(94)90217-8
Durner J Wendehenne D Klessig DF Defense gene induction in tobacco by nitric oxide, cyclic GMP, cyclic ADP-ribose Proc Natl Acad Sci USA 1998 95 10328 33 9707647 10.1073/pnas.95.17.10328
Bostock RM Stermer BA Perspectives on wound-healing in resistance to pathogens Annu Rev Phytopath 1989 27 343 371 10.1146/annurev.py.27.090189.002015
Lamb C Dixon RA The oxidative burst in plant disease resistance Annu Rev Plant Physiol Plant Mol Biol 1997 48 251 275 15012264 10.1146/annurev.arplant.48.1.251
Durrant WE Rowland O Piedras P Hammond-Kosack KE Jones JDG cDNA-AFLP reveals a striking overlap in race-specific resistance and wound response gene expression profiles Plant Cell 2000 12 963 977 10852940 10.1105/tpc.12.6.963
Zhang S Liu Y Klessig DF Multiple levels of tobacco WIPK activation during the induction of cell death by fungal elicitins Plant J 2000 23 339 47 10929127 10.1046/j.1365-313x.2000.00780.x
Bradley DJ Kjellbom P Lamb CJ Elicitor-induced and wound-induced oxidative cross-linking of a proline-rich plant-cell wall protein : a novel rapid defense response Cell 1992 70 21 30 1623521 10.1016/0092-8674(92)90530-P
Levine A Tenhaken R Dixon R Lamb C H2O2 From the oxidative burst orchestrates the plant hypersensitive disease resistance response Cell 1994 79 583 593 7954825 10.1016/0092-8674(94)90544-4
Orozco-Cardenas M Ryan CA Hydrogen peroxide is generated systemically in plant leaves by wounding and systemin via the octadecanoid pathway Proc Natl Acad Sci USA 1999 96 6553 7 10339626 10.1073/pnas.96.11.6553
Orozco-Cardenas ML Narvaez-Vasquez J Ryan CA Hydrogen peroxide acts as a second messenger for the induction of defense genes in tomato plants in response to wounding, systemin, and methyl jasmonate Plant Cell 2001 13 179 191 11158538 10.1105/tpc.13.1.179
Torres MA Dangl JL Jones JD Arabidopsis gp91phox homologues AtrbohD and AtrbohF are required for accumulation of reactive oxygen intermediates in the plant defense response Proc Natl Acad Sci USA 2002 99 517 22 11756663 10.1073/pnas.012452499
Delledonne M Xia Y Dixon RA Lamb C Nitric oxide functions as a signal in plant disease resistance Nature 1998 394 585 588 9707120 10.1038/29087
Durner J Wendehenne D Klessig DF Defense gene induction in tobacco by nitric oxide, cyclic GMP and cyclic ADP-ribose Proc Natl Acad Sci USA 1998 95 10328 10333 9707647 10.1073/pnas.95.17.10328
Chang HY Yang X Proteases for cell suicide: functions and regulation of caspases Microbiol Mol Biol Rev 2000 64 821 46 11104820 10.1128/MMBR.64.4.821-846.2000
Watanabe N Lam E Recent advance in the study of caspase-like proteases and Bax inhibitor-1 in plants: their possible roles as regulator of programmed cell death Plant Pathol 2004 5 65 70 10.1111/j.1364-3703.2004.00206.x
Hatsugai N Kuroyanagi M Yamada K Meshi T Tsuda S Kondo M Nishimura M Hara-Nishimura I A plant vacuolar protease, VPE, mediates virus-induced hypersensitive cell death Science 2004 305 855 8 15297671 10.1126/science.1099859
Cutler SR Ehrhardt DW Griffitts JS Somerville CR Random GFP:: cDNA fusions enable visualization of subcellular structures in cells of Arabidopsis at a high frequency Proc Natl Acad Sci USA 2000 97 3718 3723 10737809 10.1073/pnas.97.7.3718
Alvarez ME Pennell RI Meijer PJ Ishikawa A Dixon RA Lamb C Reactive oxygen intermediates mediate a systemic signal network in the establishment of plant immunity Cell 1998 92 773 784 9529253 10.1016/S0092-8674(00)81405-1
Cutler SR Ehrhardt DW Polarized cytokinesis in vacuolated cells of Arabidopsis Proc Natl Acad Sci USA 2002 99 2812 2817 11880633 10.1073/pnas.052712299
Heath MC The enigmatic hypersensitive response: induction, execution, and role Physiol Mol Plant Pathol 1999 55 1 3 10.1006/pmpp.1999.0217
Staehelin LA The plant ER: A dynamic organelle composed of a large number of discrete functional domains Plant J 1997 11 1151 1165 9225461 10.1046/j.1365-313X.1997.11061151.x
Kandasamy MK Meagher RB Actin-organelle interaction: Association with chloroplast in Arabidopsis leaf mesophyll cells Cell Motil Cytoskeleton 1999 44 110 118 10506746 10.1002/(SICI)1097-0169(199910)44:2<110::AID-CM3>3.0.CO;2-O
Cutler SR Live cell analysis in Arabidopsis using randomly generated markers 2001 Stanford University
Zottini M Scoccianti V Zannoni D Effects of 3, 5-dibromo-4-hydroxybenzonitrile (Bromoxynil) on bioenergetics of higher-plant mitochondria (Pisum-sativum) Plant Physiol 1994 106 1483 1488 12232423
Carpenter K Heywood BJ Herbicidal action of 3, 5-dihalogeno-4-hydroxybenzonitriles Nature 1963 200 28 29
Giardi MT Marder JB Barber J Herbicide binding to the isolated photosystem-II reaction center Biochim Biophys Acta 1988 934 64 71
Gunning BES The identity of mystery organelles in Arabidopsis plants expressing GFP Trends Plant Sci 1998 3 417 417 10.1016/S1360-1385(98)01336-3
Ridge RW Uozumi Y Plazinski J Hurley UA Williamson RE Developmental transitions and dynamics of the cortical ER of Arabidopsis cells seen with green fluorescent protein Plant Cell Physiol 1999 40 1253 61 10682347
Hayashi Y Yamada K Shimada T Matsushima R Nishizawa NK Nishimura M Hara-Nishimura I A proteinase-storing body that prepares for cell death or stresses in the epidermal cells of Arabidopsis Plant Cell Physiol 2001 42 894 899 11577182 10.1093/pcp/pce144
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
Normanly J Grisafi P Fink GR Bartel B Arabidopsis mutants resistant to the auxin effects of Indole-3-acetonitrile are defective in the nitrilase encoded by the Nit1 gene Plant Cell 1997 9 1781 1790 9368415 10.1105/tpc.9.10.1781
Elbaz M Avni A Weil M Constitutive caspase-like machinery executes programmed cell death in plant cells Cell Death Diff 2002 9 726 733 10.1038/sj.cdd.4401030
Wessel D Flugge UI A method for the quantitative recovery of protein in dilute solution in the presence of detergents and lipids Anal Biochem 1984 138 141 3 6731838 10.1016/0003-2697(84)90782-6
Ausubel FM Brent R Kingston RE Moore DD Seidman JG Smith JA Struhl K Current protocols in molecular biology Wiley-interscience 1993
Towbin H Staehelin T Gordon J Electrophoretic transfer of proteins from polyacrylamide gels to nitrocellulose sheets: procedure and some applications Proc Natl Acad Sci USA 1979 76 4350 4 388439
Dreier L Rapoport TA In vitro formation of the endoplasmic reticulum occurs independently of microtubules by a controlled fusion reaction J Cell Biol 2000 148 883 898 10704440 10.1083/jcb.148.5.883
Rose Biotechnololgy
| 15796778 | PMC1087855 | CC BY | 2021-01-04 16:37:14 | no | BMC Plant Biol. 2005 Mar 29; 5:4 | utf-8 | BMC Plant Biol | 2,005 | 10.1186/1471-2229-5-4 | oa_comm |
==== Front
BMC Struct BiolBMC Structural Biology1472-6807BioMed Central London 1472-6807-5-81580435010.1186/1472-6807-5-8Research Article2Statistically significant dependence of the Xaa-Pro peptide bond conformation on secondary structure and amino acid sequence Pahlke Doreen [email protected] Christian [email protected] Dietmar [email protected] Dirk [email protected] Forschungsinstitut für Molekulare Pharmakologie Robert-Rössle-Str. 10, D-13125 Berlin, Germany2 Biotechnologisches Zentrum Universität Dresden, AG Zelluläre Maschinen Tatzberg 47–51, D-01307 Dresden, Germany2005 1 4 2005 5 8 8 3 8 2004 1 4 2005 Copyright © 2005 Pahlke 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 reliable prediction of the Xaa-Pro peptide bond conformation would be a useful tool for many protein structure calculation methods. We have analyzed the Protein Data Bank and show that the combined use of sequential and structural information has a predictive value for the assessment of the cis versus trans peptide bond conformation of Xaa-Pro within proteins. For the analysis of the data sets different statistical methods such as the calculation of the Chou-Fasman parameters and occurrence matrices were used. Furthermore we analyzed the relationship between the relative solvent accessibility and the relative occurrence of prolines in the cis and in the trans conformation.
Results
One of the main results of the statistical investigations is the ranking of the secondary structure and sequence information with respect to the prediction of the Xaa-Pro peptide bond conformation. We observed a significant impact of secondary structure information on the occurrence of the Xaa-Pro peptide bond conformation, while the sequence information of amino acids neighboring proline is of little predictive value for the conformation of this bond.
Conclusion
In this work, we present an extensive analysis of the occurrence of the cis and trans proline conformation in proteins. Based on the data set, we derived patterns and rules for a possible prediction of the proline conformation. Upon adoption of the Chou-Fasman parameters, we are able to derive statistically relevant correlations between the secondary structure of amino acid fragments and the Xaa-Pro peptide bond conformation.
==== Body
Background
The peptide bond has a partial double bond character which results in the plane arrangement of the six backbone atoms Cα(i-1), C'(i-1), O(i-1), N(i), H(i), Cα(i). The angles Ψ, Φ and Ω readily describe the arrangement of the six atoms in three-dimensional space: Ψ defines the angle of the N – Cα bond and Φ is given by the Cα-C' bond of the same residue whereas the Ω angle is defined between the C' and N atoms of adjacent residues. For Ω only two conformations are energetically and sterically preferred. The trans conformation is defined by an Ω angle of 180° while the cis conformation ideally displays an Ω angle of 0°. The cis conformation occurs rarely in polypeptides because of the higher intrinsic energy compared to the trans conformation [1]. In contrast to other amino acids proline has a higher propensity for the cis conformation. This can be explained by the smaller energy difference between the cis and trans isomer which is 2 kcal/mol higher for the cis as compared to the trans imide bond. The functional relevance of the proline cis/trans equilibrium is supported by the existence of special enzymes called peptidyl-prolyl isomerases which catalyze the cis/trans isomerization of Xaa-Pro bond [2,3]. The action of these enzymes is thought to be important for the proper functioning of biological processes such as protein folding [4,5] and splicing [6]. In addition, cis prolines act possibly as molecular switches [7] and are frequently present in turn regions of water-soluble proteins [8].
Nuclear magnetic resonance (NMR) experiments have shown that the cis/trans ratio depends on the amino acid sequence adjacent to the proline. More specifically, a correlation has been found between the isomerization rate and the bulkiness of the side chain of the residue preceding proline. The isomerization rate becomes smaller as the bulkiness of the side chain increases. For example, aromatic residues cause an approximately tenfold reduction in isomerization rate in comparison to alanine [9]. Further NMR studies [10] demonstrated that the cis/trans ratio is influenced by the nature of the succeeding amino acid. Positively charged side chains seem to destabilize cis relative to trans whereas aspartate, asparagine and glycine stabilize the cis form.
Cis and trans isomers show different tendencies to be present in certain secondary structure elements: While the trans conformations can be found in all classes of secondary structure (in helices at the beginning or the end and in the center of helices causing a sharp kink of ≥ 20°) [11], the cis isomer is usually confined to bend and turn regions within proteins. However, a systematic analysis of these findings has been elusive.
The goal of our study is to derive statistically relevant propensities for predicting the relative occurrence of cis and trans proline conformations in proteins with respect to their sequential and structural properties.
For the analysis of the data set, different statistical methods such as calculation of the Chou-Fasman parameters [12] and occurrence matrices were used.
Results and discussion
Chou-Fasman parameters
Figure (1) shows the comparison of the relative natural occurrence of the secondary structure types for cis and trans conformations of the Xaa-Pro peptide bond based on the PDB analyses (see Materials and Methods). Interestingly, cis prolines mostly occur in bend structures and almost never in helix or strand whereas trans prolines are mostly found in coil and to a smaller degree in turn, bend, helix or strand structures.
This observation is confirmed by analysis of the data with the modified Chou-Fasman parameters (equations 1–3 in the Methods section). The values obtained from this analysis (Table (1)) show that the parameter values are close to one for the trans proline, indicating no significant preference for any given secondary structure type. In contrast, the value is high for cis prolines within the bend structure type (4.249) while it is very small for helix (0.036) or strand (0.051) elements.
In addition, the Chou-Fasman parameters of the secondary structures of the two amino acids adjacent to proline (i-1 and i+1) are calculated and illustrated in Table (2). Similar to the results for the central proline position (Table (1)), the trans conformation of the proline is almost equally distributed amongst the different secondary structure types for these two positions (values close to 1). For cis prolines, the Chou-Fasman parameter at positions (i-1) and (i+1) show a strong bias against helical secondary structure, albeit not as strong as for the proline itself (0.086 and 0.321, respectively). The low propensity for the strand structure seen for the central cis proline is not observed for the preceding residue and is above 1 for the residue following proline. The preference for the bend structure is attenuated for the adjacent residues (2.054 and 1.263, respectively) as compared to the central cis proline (4.249). Interestingly, the propensity for the residue preceding cis proline to be part of a turn structure is higher than to be part of a bend structure (2.261 and 2.054, respectively). For the cis proline itself, the order is reverse.
Occurrence matrix
A matrix of residue occurrence of the five-piece fragments (Table (3)) reveals the different preferences of residues adjacent to proline. On the left side all residue combinations with a trans proline in the mid-position are listed and on the right side all combinations with a cis proline in this position are reported. Rows indicate the amino acid types and columns contain the absolute and relative occurrences of the amino acids in the five-pieces fragments with proline in the central position.
Table (3) illustrates the significant changes of the natural occurrence of the amino acids at different positions. The calculated natural occurrences (column 2) correspond to the results by [13].
The observed relative occurrences were normalized in respect to the natural occurrence at each position for each amino acid type (Table (4)). The normalized occurrences for cis and trans (trans proline conformation Htrans and cis proline conformation Hcis) show the preference for certain amino acid sequence patterns. Amino acids with a relative occurrence greater one are shown in bold face. We considered a relative occurrence difference as almost equal (≈) if ΔH was smaller than 0.05. ΔH values larger than 0.5 were of particular significance (>>). The following observations can be made: For the normalized occurrences (Htrans/cis) of the trans conformations at the position i-2 the amino acids P≈G>F>Y≈W>T = N≈L are preferred in comparison to the cis conformation where C>>Q>P>S≈M>K≈G>V are favored in position i-2. For position i-1 the amino acids C>N>I = H>T = D≈L≈K≈Q≈R are predominant for the trans conformation while the order W>Y>>G>F>C>N>P≈Q≈A was found for the cis conformation. For the succeeding residue of Pro (residue i+1) the preferences E>Q = A≈D≈V = G≈R are observed for the trans conformation and the order Y≈ F>A>W = C>H>G≈S≈P = L≈Q resulted for the cis conformation. The most favorable amino acids are P≈W = G≈S≈M≈A≈F≈L for the trans conformation (although the highest ΔH value of P is only 1.09 indicating that this position has almost no preference for any amino acid) and P>C>>T>N>G≈R = K≈V for the cis conformation at position i+2.
By analyzing the individual positions, it appears that certain amino acids occur in the cis and in the trans sets with high propensity (in position i-2: G and P, in position i-1: C, N and Q, in position i+1: G, A and Q and in position i+2: P and G). Besides those amino acids a number of amino acids occur more exclusively at distinct positions for either the cis or the trans conformation. Interestingly, certain properties seem to dominate at particular positions. For cis, aromatic residues are very likely at position i-1 and i+1 whereas for trans aromatic residues occur with high propensity at position i-2.
In Table (5) the occurrences of the secondary structure types of proline in the cis and trans conformation are shown. For the five secondary structure elements (bend, coil, helix, strand and turn) we analyzed the surrounding of the fixed proline in the mid-position. Rows indicate the secondary structure type with fixed secondary structure of the proline. Columns show the relative occurrence of the secondary structure at the 4 positions around proline. Based on the relative occurrence of the secondary structure of the proline we can specify typical secondary structure pattern for cis and trans Xaa-Pro peptide bond conformation.
For example, if the cis proline is located in a bend or coil secondary structure type the amino acid at the position i-1 is never an element of a helix or turn. Furthermore, in the rare case where cis proline is part of a helix it is never found at the beginning or the end of the helix. In the case of the cis proline being present in a turn we never found a helix or bend structure element at the position i-1. However, residues in a turn occurred relatively often (378 times at the position i-1).
The 10 most frequently occurring secondary structure fragments with a cis proline at the mid position are presented in the second and third column of Table (6). For comparison, the occurrences of the corresponding fragments with trans proline are shown in column 4 and 5. In contrast, Table (7) shows the 10 most frequently occurring secondary structure fragments of trans proline and the corresponding occurrence of cis prolines. Here, the amount of helical regions in the environment of proline increases dramatically whereas in the cis case the helix structure occurs very rarely. Most of the trans prolines of the five-piece fragments in helical regions appear as the first helix element in contrast to the helical cis prolines which do never occur as the first element of a helix (Table (5)).
All previous results from the different approaches are compared with the content of the "Top 10" tables (Tables (6)(7)) in order to confirm the observed dependencies of sequence and secondary structure on the cis/trans peptide bond conformation of proline.
As a control parameter for the estimation of cis and trans conformations the sum of the Chou-Fasman parameter can be used (Table (1) and (2)). The sum of the at the positions (i-1, i, i+1) leads to the same ranking in comparison to the "TOP 10" tables considering only 3 positions. For the pattern bbc in the cis case the sum is 7.746 and for trans conformation it is 2.541. In contrast the pattern ccc leads to 3.071 for trans and 2.275 for the cis conformation. On the basis of separated data sets for cis and trans proline conformation we derived the occurrence for the secondary structure elements and sequence pattern from so-called occurrence matrices (Table (3)(4) and (5)).
The output from the "TOP 10" tables can be verified in the occurrence matrices (Table (5)). The cbbcc pattern is most probable for the cis conformation and for trans conformation it is the ccccc pattern. They coincide as most probable secondary structure combination in Table (5) as highlighted by bold face. For example, for the secondary structure element bend in position i, ccbbc results as most probable for the trans proline and the cbbcc pattern for cis. For coil as secondary structure element in the position i ccccs results for the cis case whereas ccccc occurs for trans.
In order to evaluate the significance of Table (3) and (4), we analyzed the ccccc secondary structure pattern for the middle proline in trans peptide bond conformation in terms of the most probable amino acid in the 4 positions (H > 1). We found the following preferences: position i-2: P>M>Q>A>Y>T>R>S, position i-1: P>S>A>Q>R>K>M>C>L>T, position i+1: P>R>T>E>S>Q>A>V>K position i+2: P>S>K>R>Y>A>V>E. However, if those amino acids are compared with the most probable amino acids for trans or cis peptide bond conformation with proline in position i, no unambiguous sequence pattern remains.
We conclude that the influence of the secondary structure on a possible prediction of the Xaa-Pro peptide bond conformation is more significant than the sequence information of the residues surrounding proline.
Analysis of the solvent accessible surface
The relationship between the relative solvent accessibility and the relative occurrence of prolines in our PDB data set was investigated. In the range from 0% to 20% of solvent accessibility 62.7% of the trans entries can be found compared to 56.1% cis entries whereas 43.9% cis instances occur in the range above the threshold of 20 % accessibility rate in contrast to 37.3% trans. These numbers suggest that proline in the cis conformation is slightly more frequently found in surface accessible areas compared to the trans proline.
The reason for the difference can be explained by the finding that cis prolines are more frequently found in solvent-exposed turn and bend structures, whereas trans prolines mostly occur in either helix or strand secondary structure elements. The relatively high frequency of exposed cis-prolines in conjunction with the relatively low energy barrier for the cis to trans conversion as compared to other amino acids mark proline as a preferred site for conformational switch mechanism in proteins. For example, PPIases catalyze the isomerization rate and thereby may regulate biological responses by alternative conformations of loop regions [14].
Conclusion
In this work, we have analyzed more than 15000 proline residues within PDB-deposited protein structures in regard to their peptide bond conformation (cis or trans). We extracted fragments of 5 residues in length with proline in the mid-position. The PDB-derived secondary structure and the sequence information were used in a further statistical analysis of the 15778 fragments.
The calculation and interpretation of occurrence matrices reveal distinct preferences for the cis and for the trans conformation in dependence of secondary structure types. By the use of the modified Chou-Fasman parameter at the positions (i-1), (i) and (i+1) (equations 1–3) propensities for the proline peptide bond conformation can be derived from the secondary structure pattern. It is conceivable that an implementation of the modified Chou-Fasman parameters can be used for the prediction of proline peptide bond conformation in fragments with known secondary structure. An application of the new Chou-Fasman parameters for other naturally occurring amino acids is possible and leads to statistically relevant propensities for the prediction of the peptide conformation of any of the 20 amino acids in proteins. A prediction algorithm is presented on our website .
Finally the relationship between the relative solvent accessibility of proline and its peptide bond conformation shows that cis prolines occur more frequently in surface accessible areas compared to the prolines in trans conformation.
Methods
Materials
For data acquisition the PDB [15] was used by iterating over those protein entries who's PDB IDs are in a set of 3722 nonredundant proteins. All these proteins fulfill the following conditions: they share a maximum sequence identity of 25%, they have been solved to a resolution of 4.0 Å or less, they display a maximum R-value of 1.0 and a maximal chain length of 10000 amino acids (given by the PISCES [16] protein sequence culling service at .
From these proteins the coordinate section was extracted to calculate the Ω dihedral angle between adjacent residues. A Perl script using the angle calculation algorithm of the PDB tool dihedrl.for was used for this calculation. A peptide bond was defined to be in cis conformation if the Ω angle was between -30° and +30° whereas angles outside of this range are assumed to be trans. The resulting file contains the PDB ID, the chain notation, the position of the cis residue in the sequence, its amino acid three letter code and the calculated Ω angle.
The resulting set of the calculation comprised 954 proteins containing at least one peptide backbone conformation of cis. The adjacent residues and the secondary structure were extracted from the locally installed PDB files pdb_seqres.txt and ss.txt. In this way fragments of five amino acids were created with two residues flanking the proline at the mid-position on each side. The secondary structure of these five-residue segments were derived from the ss.txt of the respective PDB file and was calculated on the basis of the hydrogen bonding pattern by DSSP [17] denoting H as helix, B as beta bridge, E as strand, G as 3.1 helix, I as pi-helix, T as turn, S as bend and a blank space as coil structure. We grouped the DSSP derived structural information into the following five types: {b(end) = S, c(oil) = {B, }, h(elix) = {H, G, I}, s(trand) = E, t(urn) = T}. The resulting data set contained 15778 entries including 1390 fragments containing cis prolines. Each entry of the data set comprised the amino acid types at each position, the secondary structure information and the classification (cis/trans) (for example "R, S, P, F, T, c, b, b, c, t, cis").
Chou-Fasman parameter
The correlation between secondary structure type and the conformation of proline can be calculated from the Chou Fasman parameters. We have applied the Chou-Fasman algorithm to elucidate this correlation by using the following formulas:
where fs denotes the occurrence of a certain secondary structure type with proline whether in the cis or trans conformation relative to the total occurrence of the same structure type. fclass is the relation of the number of prolines in a specific conformation and the total number of prolines in the data set. then describes the altered Chou-Fasman parameter for the probability of the cis or trans conformation of proline to be present in the individual secondary structure types.
Solvent accessible area
The solvent accessible area was calculated for the whole protein with DSSP [17]. The DSSP algorithm provides information of the secondary structure based on the hydrogen bond pattern and of the solvent accessible area [18] of the different amino acids in the sequences. For the normalization of the accessible surface area of the prolines, the values obtained by DSSP were divided by the maximum solvent accessible area of an isolated proline residue (269 A2). The calculated relative accessibilities were in the range from 0% to 78%.
Authors' contributions
DP, DLe, DLa, CF are responsible for data mining, statistical analysis and the manuscript preparation. DP & DLa programmed the scripts. All authors read and approved the final manuscript.
Acknowledgements
We are grateful for the financial support of this work by the BMBF-Leitprojekt "Strukturanalyse mit hohem Durchsatz für medizinisch relevante Proteine – Proteinstrukturfabrik" (Fk.01GG9812). C.F. acknowledges a grant from the Volkswagen Stiftung related to this work (I/77955). The authors thank Urs Wiedeman for helpful discussion and careful reading of the manuscript.
Figures and Tables
Figure 1 Occurrence of secondary structure. Relative occurrence of the secondary structure types compared for cis and trans prolines.
Table 1 Chou-Fasman parameters (i). Chou-Fasman parameters for the cis/trans classes at position i regarding the secondary structure of proline.
Structure
Bend 4.249 0.683
Coil 0.389 1.060
Helix 0.036 1.094
Strand 0.051 1.093
Turn 1.452 0.956
Table 2 Chou-Fasman parameters (i-1) (i+1). Chou-Fasman parameters for the cis/trans classes regarding the secondary structure of proline's predecessors and successors.
structure i-1 i+1
bend 2.054 0.897 1.263 0.974
coil 0.462 1.052 1.437 0.957
helix 0.086 1.089 0.321 1.066
strand 0.794 1.020 1.207 0.980
turn 2.261 0.877 0.786 1.021
Table 3 Occurrence matrix. Occurrence matrix for the amino acid combinations of trans and cis proline in the fixed position (i). The total number Ntotal is the occurrence of an amino acid in all investigated sequence fragments. The number of the proline at the position (i) in the trans conformation is 14388 and in the cis conformation it is 1390. The absolute and relative occurrence of one amino acid is shown at all positions (rel and abs). The relative occurrence is the relation of the number of the amino acid at the position to the total number of all amino acid at this position.
trans Pro cis Pro
Position position
AA Total i-2 i-1 i+1 i+2 i-2 i-1 i+1 i+2
abs rel abs rel abs rel abs rel abs rel abs rel abs rel abs rel abs rel
A: 4343 0.075 1024 0.071 1030 0.072 1169 0.081 1120 0.078 90 0.065 111 0.080 134 0.096 79 0.057
C: 810 0.014 204 0.014 256 0.018 166 0.012 184 0.013 38 0.027 25 0.018 24 0.017 33 0.024
D: 3565 0.062 818 0.057 972 0.068 956 0.066 819 0.057 57 0.041 59 0.042 52 0.037 83 0.060
E: 3604 0.063 799 0.056 685 0.048 1208 0.084 912 0.063 70 0.050 84 0.060 47 0.034 61 0.044
F: 2426 0.042 656 0.046 562 0.039 596 0.041 612 0.043 59 0.042 82 0.059 96 0.069 58 0.042
G: 4275 0.074 1202 0.084 824 0.057 1113 0.077 1136 0.079 112 0.081 154 0.111 112 0.081 106 0.076
H: 1418 0.025 323 0.022 418 0.029 313 0.022 364 0.025 32 0.023 29 0.021 40 0.029 32 0.023
I: 3178 0.055 780 0.054 915 0.064 711 0.049 772 0.054 68 0.049 38 0.027 54 0.039 62 0.045
K: 2804 0.049 654 0.045 745 0.052 703 0.049 702 0.049 76 0.055 66 0.047 57 0.041 70 0.050
L: 5279 0.092 1336 0.093 1423 0.099 1179 0.082 1341 0.093 96 0.069 90 0.065 134 0.096 96 0.069
M: 1131 0.020 292 0.020 272 0.019 263 0.018 304 0.021 35 0.025 21 0.015 22 0.016 28 0.020
N: 2683 0.047 684 0.048 816 0.057 608 0.042 575 0.040 65 0.047 79 0.057 56 0.040 70 0.050
P: 3121 0.054 896 0.062 687 0.048 691 0.048 847 0.059 98 0.071 82 0.059 78 0.056 149 0.107
Q: 2105 0.037 480 0.033 556 0.039 574 0.040 495 0.034 72 0.052 56 0.040 53 0.038 45 0.032
R: 2591 0.045 644 0.045 675 0.047 655 0.046 617 0.043 61 0.044 49 0.035 54 0.039 64 0.046
S: 3596 0.062 886 0.062 887 0.062 877 0.061 946 0.066 108 0.078 84 0.060 91 0.065 79 0.057
T: 3479 0.060 876 0.061 948 0.066 818 0.057 837 0.058 84 0.060 61 0.044 77 0.055 102 0.073
V: 4165 0.072 1021 0.071 1036 0.072 1073 0.075 1035 0.072 101 0.073 65 0.047 97 0.070 101 0.073
W: 815 0.014 221 0.015 169 0.012 206 0.014 219 0.015 18 0.013 44 0.032 23 0.017 19 0.014
Y: 2164 0.038 592 0.041 512 0.036 509 0.035 551 0.038 50 0.036 111 0.080 89 0.064 53 0.038
Table 4 Normalized relative occurrences. Normalized relative occurrences of all amino acids at the different positions. For the normalization we used the coefficient of each observed occurrence and the natural occurrence for each amino acid at each position. Highlighted in bold are the most probable amino acid (H > 1) occurring at the four positions.
AA H
trans
i-2 H
cis
i-2 H
trans
i-1 H
cis
i-1 H
trans
i+1 H
cis
i+1 H
trans
i+2 H
cis
i+2
A: 0.95 0.87 0.96 1.07 1.08 1.28 1.04 0.76
C: 1.00 1.93 1.29 1.29 0.86 1.21 0.93 1.71
D: 0.92 0.66 1.10 0.68 1.06 0.60 0.92 0.97
E: 0.89 0.79 0.76 0.95 1.33 0.54 1.00 0.70
F: 1.10 1.00 0.93 1.40 0.98 1.64 1.02 1.00
G: 1.14 1.09 0.77 1.50 1.04 1.09 1.07 1.03
H: 0.88 0.92 1.16 0.84 0.88 1.16 1.00 0.92
I: 0.98 0.89 1.16 0.49 0.89 0.71 0.98 0.82
K: 0.92 1.12 1.06 0.96 1.00 0.84 1.00 1.02
L: 1.01 0.75 1.08 0.71 0.89 1.04 1.01 0.75
M: 1.00 1.25 0.95 0.75 0.90 0.80 1.05 1.00
N: 1.02 1.00 1.21 1.21 0.89 0.85 0.85 1.06
P: 1.15 1.31 0.89 1.09 0.89 1.04 1.09 1.98
Q: 0.89 1.41 1.05 1.08 1.08 1.03 0.92 0.86
R: 1.00 0.98 1.04 0.78 1.02 0.87 0.96 1.02
S: 1.00 1.26 1.00 0.97 0.98 1.05 1.06 0.92
T: 1.02 1.00 1.10 0.73 0.95 0.92 0.97 1.22
V: 0.99 1.01 1.00 0.65 1.04 0.97 1.00 1.01
W: 1.07 0.93 0.86 2.29 1.00 1.21 1.07 1.00
Y: 1.08 0.95 0.95 2.11 0.92 1.68 1.00 1.00
Table 5 Secondary structure matrix. The occurrence of the secondary structure types of proline in the cis and trans conformation is shown. For the five defined secondary structure elements (bend, coil, helix, strand and turn) we analyzed the surrounding of the fixed proline in the mid-position. Rows indicate the secondary structure type with fixed secondary structure of the proline. Columns show the relative occurrences of the secondary structure at the 4 positions around proline. Bold face highlights the most probable secondary structure. For the cis peptide bond conformation of the central proline in helix and strand not enough entries could be collected from the PDB to reach significance.
secondary structure structure of trans Pro structure of cis Pro
position position
i-2 i-1 bend i+1 i+2 i-2 i-1 bend i+1 i+2
bend: 257 0.21 317 0.26 1211 684 0.56 229 0.19 176 0.24 467 0.63 742 143 0.19 105 0.14
coil: 462 0.38 813 0.67 1211 334 0.28 545 0.45 329 0.44 180 0.24 742 433 0.58 321 0.43
helix: 77 0.06 2 0.00 1211 22 0.02 107 0.09 31 0.04 0 0.00 742 28 0.04 70 0.09
strand: 254 0.21 74 0.06 1211 135 0.11 228 0.19 164 0.22 95 0.13 742 104 0.14 174 0.23
turn: 161 0.13 5 0.00 1211 36 0.03 102 0.08 42 0.06 0 0.00 742 34 0.05 72 0.10
i-2 i-1 coil i+1 i+2 i-2 i-1 coil i+1 i+2
bend: 1401 0.22 1614 0.25 6505 952 0.15 901 0.14 41 0.18 80 0.35 231 29 0.13 37 0.16
coil: 2966 0.46 4430 0.68 6505 3749 0.58 2656 0.41 112 0.48 137 0.59 231 127 0.55 81 0.35
helix: 377 0.06 8 0.00 6505 532 0.08 862 0.13 9 0.04 0 0.00 231 9 0.04 14 0.06
strand: 934 0.14 360 0.06 6505 788 0.12 1312 0.20 41 0.18 13 0.06 231 62 0.27 91 0.39
turn: 827 0.13 93 0.01 6505 484 0.07 774 0.12 28 0.12 1 0.00 231 4 0.02 8 0.03
i-2 i-1 helix i+1 i+2 i-2 i-1 helix i+1 i+2
bend: 361 0.14 193 0.07 2603 1 0.00 12 0.00 0 0.00 0 0.00 8 0 0.00 1 0.13
coil: 752 0.29 913 0.35 2603 1 0.00 38 0.01 0 0.00 0 0.00 8 0 0.00 2 0.25
helix: 722 0.28 935 0.36 2603 2597 1.00 2489 0.96 8 1.00 8 1.00 8 8 1.00 0 0.00
strand: 165 0.06 43 0.02 2603 0 0.00 6 0.00 0 0.00 0 0.00 8 0 0.00 0 0.00
turn: 603 0.23 519 0.20 2603 4 0.00 58 0.02 0 0.00 0 0.00 8 0 0.00 5 0.63
i-2 i-1 strand i+1 i+2 i-2 i-1 strand i+1 i+2
bend: 124 0.09 98 0.07 1352 123 0.09 202 0.15 0 0.00 2 0.22 9 1 0.11 1 0.11
coil: 268 0.20 157 0.12 1352 391 0.29 322 0.24 4 0.44 2 0.22 9 2 0.22 2 0.22
helix: 10 0.01 0 0.00 1352 80 0.06 115 0.09 0 0.00 0 0.00 9 0 0.00 0 0.00
strand: 788 0.58 1096 0.81 1352 703 0.52 603 0.45 3 0.33 4 0.44 9 6 0.67 4 0.44
turn: 162 0.12 1 0.00 1352 55 0.04 110 0.08 2 0.22 1 0.11 9 0 0.00 2 0.22
i-2 i-1 turn i+1 i+2 i-2 i-1 turn i+1 i+2
bend: 334 0.12 247 0.09 2717 12 0.00 450 0.17 57 0.14 0 0.00 400 54 0.14 39 0.10
coil: 777 0.29 1266 0.47 2717 26 0.01 835 0.31 142 0.36 9 0.02 400 88 0.22 138 0.35
helix: 534 0.20 127 0.05 2717 239 0.09 414 0.15 34 0.09 0 0.00 400 57 0.14 74 0.19
strand: 447 0.16 155 0.06 2717 20 0.01 192 0.07 38 0.10 13 0.03 400 20 0.05 69 0.17
turn: 625 0.23 922 0.34 2717 2420 0.89 826 0.30 129 0.32 378 0.95 400 181 0.45 80 0.20
Table 6 10 most frequent cis secondary structure combinations. Comparison of the 10 most frequent cis secondary structure combinations versus the corresponding trans secondary structure combinations.
Structure cis Pro Trans Pro
absolute relative(%) absolute relative(%)
cbbcc 93 6.6906 24 0.1668
bbbcc 58 4.1727 28 0.1946
cbbss 45 3.2374 7 0.0487
cttcc 31 2.2302 3 0.0209
ssbcc 30 2.1583 12 0.0834
ttttt 28 2.0144 166 1.1537
ccbcc 25 1.7986 48 0.3336
cbbbc 25 1.7986 22 0.1529
bbbss 25 1.7986 21 0.146
sbbcc 24 1.7266 5 0.0348
Table 7 10 most frequent trans secondary structure. Comparison of the 10 most frequent trans secondary structure combinations versus the corresponding cis secondary structure combinations.
structure trans Pro Cis Pro
absolute relative(%) absolute relative(%)
ccccc 1311 9.1118 23 1.6547
cchhh 524 3.6419 0 0
hhhhh 398 2.7662 0 0
tthhh 291 2.0225 0 0
ccttc 251 1.7445 1 0.0719
cccss 241 1.675 9 0.6475
bbccc 221 1.536 5 0.3597
hthhh 218 1.5152 0 0
ccchh 213 1.4804 3 0.2158
ccttt 210 1.4595 2 0.1439
==== Refs
Ramachandran G Mitra A An explanation for the rare occurrence of cis peptide units in proteins and polypeptides J Mol Biol 1976 107 85 92 1003461
Fischer G Bang H Mech C Determination of enzymatic catalysis for the cis-trans-isomerization of peptide binding in proline-containing peptides Biomed Biochim Acta 1984 43 1101 11 6395866
Pal D Chakrabarti P Cis peptide bonds in proteins: residues involved, their conformations, interactions and locations J Mol Biol 1999 294 271 88 10556045 10.1006/jmbi.1999.3217
Lang K Schmid F Fischer G Catalysis of protein folding by prolyl isomerase Nature 1987 329 268 70 3306408 10.1038/329268a0
Wedemeyer W Welker E Scheraga H Proline cis-trans isomerization and protein folding Biochemistry 2002 41 14637 44 12475212 10.1021/bi020574b
Horowitz D Lee E Mabon S Misteli T A cyclophilin functions in pre-mRNA splicing EMBO J 2002 21 470 80 11823439 10.1093/emboj/21.3.470
Reimer U Fischer G Local structural changes caused by peptidyl-prolyl cis/trans isomerization in the native state of proteins Biophys Chem 2002 96 203 12 12034441 10.1016/S0301-4622(02)00013-3
Stewart D Sarkar A Wampler J Occurrence and role of cis peptide bonds in protein structures J Mol Biol 1990 214 253 60 2370664
Grathwohl C Wuthrich K Nmr studies of the rates of proline cis-trans isomerization in oligopeptides Biopolymers 1981 20 2623 2633 10.1002/bip.1981.360201209
Dyson H Rance M Houghten R Wright P Lerner R Folding of immunogenic peptide fragments of proteins in water solution. II. The nascent helix J Mol Biol 1988 201 201 17 3418697 10.1016/0022-2836(88)90447-0
MacArthur M Thornton J Influence of proline residues on protein conformation J Mol Biol 1991 218 397 412 2010917 10.1016/0022-2836(91)90721-H
Chou P Fasman G Prediction of protein conformation Biochemistry 1974 13 222 45 4358940 10.1021/bi00699a002
Jones D Taylor W Thornton J The rapid generation of mutation data matrices from protein sequences Comput Appl Biosci 1992 8 275 82 1633570
Brazin K Mallis R Fulton D Andreotti A Regulation of the tyrosine kinase Itk by the peptidyl-prolyl isomerase cyclophilin A Proc Natl Acad Sci U S A 2002 99 1899 904 11830645 10.1073/pnas.042529199
Berman H Westbrook J Feng Z Gilliland G Bhat T Weissig H Shindyalov I Bourne P The Protein Data Bank Nucleic Acids Res 2000 28 235 42 10592235 10.1093/nar/28.1.235
Wang G Dunbrack R PISCES: a protein sequence culling server Bioinformatics 2003 19 1589 91 12912846 10.1093/bioinformatics/btg224
Kabsch W Sander C Dictionary of protein secondary structure: pattern recognition of hydrogen-bonded and geometrical features Biopolymers 1983 22 2577 637 6667333 10.1002/bip.360221211
Shrake A Rupley J Environment and exposure to solvent of protein atoms. Lysozyme and insulin J Mol Biol 1973 79 351 71 4760134 10.1016/0022-2836(73)90011-9
| 15804350 | PMC1087856 | CC BY | 2021-01-04 16:03:52 | no | BMC Struct Biol. 2005 Apr 1; 5:8 | utf-8 | BMC Struct Biol | 2,005 | 10.1186/1472-6807-5-8 | oa_comm |
==== Front
Biomed Eng OnlineBioMedical Engineering OnLine1475-925XBioMed Central London 1475-925X-4-251581997310.1186/1475-925X-4-25ResearchFinite volume analysis of temperature effects induced by active MRI implants with cylindrical symmetry: 1. Properly working devices Busch Martin HJ [email protected] Wolfgang [email protected] Jörg [email protected]önemeyer Dietrich HW [email protected] Research and Development Center for Microtherapy (EFMT), D-44799 Bochum, Germany2 TFH University of Applied Sciences, D-13353 Berlin, Germany3 Institut für Radiologie, Charité, Medizinische Fakultät, Humboldt-Universität zu Berlin, D-10117 Berlin, Germany4 Grönemeyer Institute for Microtherapy, University of Witten/Herdecke, D-44799 Bochum, Germany2005 8 4 2005 4 25 25 16 11 2004 8 4 2005 Copyright © 2005 Busch et al; licensee BioMed Central Ltd.2005Busch 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
Active Magnetic Resonance Imaging implants are constructed as resonators tuned to the Larmor frequency of a magnetic resonance system with a specific field strength. The resonating circuit may be embedded into or added to the normal metallic implant structure. The resonators build inductively coupled wireless transmit and receive coils and can amplify the signal, normally decreased by eddy currents, inside metallic structures without affecting the rest of the spin ensemble. During magnetic resonance imaging the resonators generate heat, which is additional to the usual one described by the specific absorption rate. This induces temperature increases of the tissue around the circuit paths and inside the lumen of an active implant and may negatively influence patient safety.
Methods
This investigation provides an overview of the supplementary power absorbed by active implants with a cylindrical geometry, corresponding to vessel implants such as stents, stent grafts or vena cava filters. The knowledge of the overall absorbed power is used in a finite volume analysis to estimate temperature maps around different implant structures inside homogeneous tissue under worst-case assumptions. The "worst-case scenario" assumes thermal heat conduction without blood perfusion inside the tissue around the implant and mostly without any cooling due to blood flow inside vessels.
Results
The additional power loss of a resonator is proportional to the volume and the quality factor, as well as the field strength of the MRI system and the specific absorption rate of the applied sequence. For properly working devices the finite volume analysis showed only tolerable heating during MRI investigations in most cases. Only resonators transforming a few hundred mW into heat may reach temperature increases over 5 K. This requires resonators with volumes of several ten cubic centimeters, short inductor circuit paths with only a few 10 cm and a quality factor above ten. Using MR sequences, for which the MRI system manufacturer declares the highest specific absorption rate of 4 W/kg, vascular implants with a realistic construction, size and quality factor do not show temperature increases over a critical value of 5 K.
Conclusion
The results show dangerous heating for the assumed "worst-case scenario" only for constructions not acceptable for vascular implants. Realistic devices are safe with respect to temperature increases. However, this investigation discusses only properly working devices. Ruptures or partial ruptures of the wires carrying the electric current of the resonance circuits or other defects can set up a power source inside an extremely small volume. The temperature maps around such possible "hot spots" should be analyzed in an additional investigation.
==== Body
Background
Metallic implants often cause distortions inside Magnetic Resonance (MR) images. These effects arise either from the different susceptibility of tissue and metal, generating a discontinuity of the local field strength at the interface, or from the Faraday cage effect, which is set up by induced eddy currents on the metallic implant structure [1-4]. An advantageous solution to overcome the eddy current shielding is to amplify the transmitted signal to and the detected signal from the spin ensemble appropriately for the Faraday cage. A local amplification of excitation and detection for a magnetic resonance signal can be achieved by using a resonator tuned to the resonance frequency of the MR system [5-8]. Such a resonator can be integrated into the metallic structure of the implant itself or can be added around the normal implant structure. The great advantage of such an active Magnetic Resonance Imaging implant (aMRIi) is the local amplification of the signal only where it is needed. The decreased excitation pulse can be amplified for the spin ensemble inside the faraday cage without affecting the signal and contrast behavior of the rest of the excited volume. The possible advantages of active MRI implants are demonstrated in Figure 1 showing the possible amplification inside a Faraday cage and the potential to get functional information out of such an implant.
Figure 1 Commercial Stent with additional resonator inside a water bath (a) and an implanted prototype in a rabbit aorta (b) with velocity curve over the cardiac cycle (c). a) shows an MR image of a commercially available nitinol stent with an additional resonator around a part of the metallic structure immersed in a water bath. The MR signal inside the stent lumen decreases outside the resonator (arrows) in contrast to the enhanced signal inside the resonator. b) shows one example of an in vivo image with a prototype implanted into a rabbit aorta [9, 10]. c) shows the measured velocity curve inside the prototype of image b and the spline approximation delivered by the the MRI System.
The amplification is adjustable by the quality factor Q of the resonance circuit and allows the imaging of the inner volume of such metallic implants as well as gathering functional information from this location, for example, velocity or flow. For these special investigations inside the implant volume the quality factor should be as high as possible to increase the signal to noise ratio for this region and to get as low signal as possible from spins outside the volume of interest. For an overall overview of the implant inside the anatomic surrounding it is only necessary to raise the decreased signal inside the Faraday cage to the normal level, which is possible with a lower quality factor. The active implant locally increases the absorbed power during MR investigations. The temperature increase induced by this extra power loss around vessel implants with a cylindrical geometry is calculated with a finite volume analysis for a "worst-case scenario". For most of the simulations only thermal heat conduction is assumed as "worst case" without cooling due to blood perfusion and without cooling due to blood flow inside the active implant. Partly also a blood flow is considered carrying energy out of the simulation volume. The calculations are carried out with respect to power loss, time of rf-exposure and size of the aMRIi. The aim of this investigation was to perform a detailed assessment of the safety for such implants under "worst case" assumptions.
Methods
Theory
Specific Absorption Rate (SAR)
The local specific absorption rate (SAR; [W/kg]) for electromagnetic radiation is defined as power absorbed per unit of mass of an object at location r [11-16],
where σ [S/m] is the electric conductivity, ρ [kg/m3] is the mass density, E [V/m] is the amplitude of a sinusoidal time dependent electric field and r is the position vector. Assuming that the amplitude of the electric field only arises by induction from a uniform, linearly polarized magnetic field with amplitude Bm [T] and angular frequency ω [rad/s] in a homogeneous body of rotational symmetry, the following equation holds [11]
with r as component of cylindrical coordinates and r = 0 on the rotational symmetry axis aligned parallel to the magnetic field Bm. Combining Eq. (1,2) the spatially averaged SAR over a cylinder with radius Rcyl and using dV = 2 π r h dr yields
For pulsed MR sequences two correction factors are necessary,
The duty cycle factor cdc equals N*τ/TR, where TR [s] is the MR-sequence repetition time, N is the number of identical excitations during TR and τ [s] is the duration of one excitation pulse. cdc equals to the time ratio of "rf-on" to "rf-on + rf-off" during an MR scan. cpwm corrects for rf-pulses which are not rectangular in shape and is the ratio between the energy of the MR excitation pulse and the energy of a rectangular shaped pulse with identical amplitude and identical duration.
Principles of B1-field amplification inside a resonator
A circularly polarized magnetic field with magnitude B1 or a linearly polarized magnetic field with amplitude Bm for excitation is amplified inside an aMRIi with respect to the used resonance circuit. Because of the shape of vascular implants and to simplify calculations, a solenoid is chosen as inductor of the resonator and is investigated as a magnetic antenna. For the theoretical estimation, the axis of the solenoid resonator is assumed to be parallel to a sinusoidal linearly polarized magnetic field with amplitude Bm, or equivalently to be in the plane of a circularly polarized magnetic field with magnitude B1. For the resonance case (2πν0 = ω0 = (LC)-1/2), where the overall resistance of the resonator is just R, the following equation can be derived from the law of induction, Ohms law and the definition of the quality factor Q of a resonator,
where Vind [V] and Vself [V] are the induced and self-induced voltage respectively, Z [Ω] is the impedance, R [Ω] is the resistance, L [H] is the inductance of the inductor and αres and α are the flip angles inside the resonator inductance and outside the resonator respectively. The total magnetic field inside the resonator arises from both components Bm and Bres. For large quality factors (Q>>1) Bres dominates Bm. With Eqs(3, 4) it is possible to calculate the magnetically induced SAR and with the SAR the corresponding power loss inside a resonator. But this takes into account only power losses due to eddy currents and not the total power loss of a resonator.
Total power loss of a resonator
The overall power loss P [W] also includes the electric losses on and around the resonator. It is given by the following basic equation [16],
where Wtotal [J] is the energy stored inside the resonator.
The total energy Wtotal can be expressed by the energy stored in the magnetic field of maximum amplitude using the following basic relations of a solenoid inductance with nr turns (),
where V, A, ℓ with index imp are the volume, cross section and length of the implant inductor. For a pulsed MR sequence with excitation magnitude B1 the total power loss can be calculated by combining Eqs. (4, 5, 6) to
A comparison of Eqs. (3b, 7) shows that for a certain resonator tuned to the resonance frequency of a specific MR field strength, the supplementary absorbed power is proportional to the SAR of the MR sequence due to the identical dependence on cdc, cpwm and B12. For a specific sequence on a MRI system with a definite resonance frequency, the extra power is proportional to the quality factor Q and the volume Vimp of the inductance (Eq. (7)). The proportionality to the quality factor Q may be surprising, because it states that for better quality factors the power loss is larger. This is due to the total energy of the resonator, which depends quadratically on the field strength inside the resonator. This field strength is proportional to the exciting field B1 (or Bm) as well as the quality factor Q. Combining Eq. (5) and Eq. (6) alters the inverse proportionality to Q (Eq. (5)) to a proportionality (Eq. (7)) with respect to the field established by the transmit coil. Examining the power loss with respect to the magnetic field inside the resonator confirms the inverse proportionality to Q.
Finite volume analysis
Principles
All temperature increases are calculated as temperature difference maps by using a simulation volume divided into many small simulation cells. The energy exchange ΔE [J] between two simulation cells with a specific contact area A [m2], temperature difference ΔT [K] and heat diffusion path length Δx [m] for a time interval Δt [s] is given by the equation for heat conduction as,
where λ [W/(m K)] is the thermal conductivity.
The total energy change ΔEtot of one cell during a time interval Δt is the sum of exchanges of this cell with all adjacent cells with non zero contact area and the energy change due to the power loss pcell inside the cell. From this total energy change the temperature difference ΔT* can be calculated,
where c [J/(kg K)] is the specific thermal capacity of the cell material and Vcell is the cell volume.
This investigation did not use any commercially available finite volume package. The simulation is self-coded for problems with cylindrical geometry in Kylix and Delphi, a software development environment, based on object oriented Pascal. The graphical outputs are mostly generated by an evaluation (registered β-test) version of Teechart 7 used within the Delphi and Kylix environment. The simulation describes the time developing temperature difference maps around an aMRIi with cylindrical geometry for a constant total power loss P.
The entire simulation volume is assumed as a cylinder with length Lsim and radius Rsim. Adequate for such a volume are cylinder coordinates. Because of the rotational symmetry, the temperature distribution is only dependent on the axial position x and the distance r from the cylinder axis (Figures 2, 3). Because of the mirror symmetry to the plane orthogonal to the cylinder axis at center, the simulation is designed to calculate the temperature distribution only on one side of the center plane. (Figures 2, 3). Using both symmetries one can divide the entire cylindrical simulation volume with length Lsim and radius Rsim into the following cells, which are sufficient to obtain complete information of the entire simulation volume. The calculation volume consists of n cylindrical shaped pieces with length Δx and index x = 1, 2, ...., n denoting the position on the cylinder axis. Each sub cylinder of length Δx is divided in one cylinder with radius Δr and index r = 1 and m-1 shells with outer radius r* Δr, thickness Δr and index r = 2, 3, ...., m. The entire simulation volume is represented by a two dimensional array of cells C [r,x] (r = 1,2,3,....m; x = 1,2,....n). Most of the cells have contact to 4 adjacent cells with contact areas different from zero (Eqs (10a, 10b)). These contact areas as well as the volumes of the cells (Eq. (10c)) only vary with the index r.
Figure 2 Definition of the cell alignment within the cylindrical simulation volume. This figure shows the geometry of the simulation volume as example with 4 × 4 cells, typical simulation volumes are set up from 100 × 100 to 500 × 500 cells. Each simulation volume is just one half of the entire volume of the implant, because its mirror symmetry allows to restrict the simulation to half the volume. The plane of mirror symmetry is on the left side of the figure. (see also Fig. 3).
Figure 3 Definition of the cell alignment within the total examined cylindrical volume for the implant. Example of an entire volume with 8 × 4 cells. The part with complete positive indices corresponds to the simulation volume of Fig. 2. Because of the plane of mirror symmetry (dark line in the middle) only this part has to be calculated. (ΔT [r,x] = ΔT [r,-x])
Al[r] [m2] is the contact area in both cylinder axis directions (from index x to x-1 and to x+1) whereas Ar[r] [m2] is the contact area in radial direction from index r to r+1. The contact area in radial direction from index r to r-1 is identical to the area Ar[r-1]. A last "shell" with cells as heat (energy) sink is placed as one boundary of the simulation volume at r = m + 1 and x = n + 1. These cells always keep a constant temperature (ΔT = 0), even when receiving energy during one simulation step. Partly the simulations use a second heat sink representing a blood flow through the inner volume of the implant. For this second heat sink the temperature differences of all cell elements below a radius rflow*Δr are also kept zero independent of the energy transfer from the cell elements with higher index r. This approximation assumes that the energy applied to the inner cylinder volume with blood flow is completely transported away from the simulation volume during one time step Δt. At r = 1 and x = 1 as well as at r = m + 1 and x = n + 1 the simulated volume has its boundaries. The symmetry and the use of cylinder coordinates imply that cells with index r = 1 or x = 1 can not exchange energy with cells at a lower index. For x = 1 the symmetry plane defines an identical temperature at index x = -1, (Figure 3) with no energy exchange across the symmetry plane. For r = 1 no cells with lower index r exist and therefore also no energy exchange is possible.
For all heat generating cell elements (pcell[r,x]>0 Eq. (9)) the physical parameters λ, ρ and c can be set freely. For all other cell elements they are set as tissue. The presented simulations use different metal parameters for heat generating cells (Table 1). The different physical parameters of metal and tissue for different simulation cells results in discontinuities at the metal tissue interface. Assuming the heat conduction path is from the center of one finite volume to the center of a neighboring finite volume, it takes place over two different materials with only half the diffusion length for each of the materials. Because of the much lower thermal conductivity of tissue compared to that of metal it is a good approximation to use only the diffusion length through tissue.
Table 1 Physical constants of tissue, titanium, iron and tantalum
material density ρ specific heat c thermal conduct. λ electric conduct. σ
[kg/m3] [J/(kg K] [W/(m·K)] [S/m]
tissue 1000 [22] 3650 [23] 0.5 [24] 0.8–8.0 [22]
saline 1003 4185 1.45*
titanium 4510 523 21.9 2.56 106
iron 7870 449 80.2 10.4 106
tantalum 16680 140 57.5 7.63 106
* The electric conductivity of physiologic saline for the calculation of the SAR was determined to be 1.45 S/m at 22°C (295 K) and a frequency of 63.5 MHz. (Personal communication with Mr. Alexander Walkov at "Physikalisch-Technische Bundesanstalt" in Berlin, Germany)
The simulation starts with a temperature field T [r,x] = 0 for all r and x. For each iteration step with duration Δt the total energy change for each cell element is calculated according to Eq. (9a). This change ΔEtot[r,x] of one cell is the sum of five different energies. One is due to the power loss pcell = p[r,x] defined for a cell element at index r and x. The four others are energy exchanges (Eq. (8), depending on the contact areas (Eq. (10a, b)) and the temperature differences between adjacent cells with respect to the equivalent heat diffusion length through the material of the examined region. From ΔEtot [r,x] the temperature change ΔT*[r,x] is calculated according to Eq. (9b). This value is added to the prior value (Tnew [r,x] := Told[r,x] + ΔT*[r,x)) for each cell during the whole iteration process. The entire simulated time tsim consists of q iterations with time interval Δt (tsim = q * Δt).
Analytical model for a heat generating linear wire as test of the simulation
One test of the principal correctness of the self-coded simulation is possible by a comparison of the simulation results with those of an analytical solution. An easy straightforward analytical solution is available for an endless linear wire with radius rwire, which is heated by a constant power density P* [W/m], defined as power per wire length [W/m]. After reaching the thermal equilibrium, the power associated with P* penetrates through every cylinder surface surrounding the linear power source independent of the radius r (with r>rwire). The temperature difference ΔT between the wire surface and a point at radial position R can be calculated in a homogeneous medium from the equation for heat conduction.
The thermal capacity does not appear in Eq. (11), because of the assumed thermal equilibrium. A simulation for a wire with power loss P* and with energy exchange only in radial direction should approach, after sufficient time, similar temperature differences between the wire surface and a point at distance (R-rwire) (Figure 4).
Figure 4 Temperature difference ΔT from wire surface with radius rwire = 0.05 mm to radial position R for different power densities P* [W/m]. The graphs show the temperature differences from the wire surface at position rwire to the position at radius R (cylinder axis defines R = 0). The starting point at rwire (0.05 mm) is so close to R = 0 that it coincides with R = 0 in the drawing.
Control of simulation results
The implemented algorithm was controlled with different checks, beside the afore-mentioned comparison with an analytical model. First, the total energy uptake of the heat sink surface is added up during all iteration steps. This energy summed with the energy stored inside the simulation volume must equal the total applied energy. The energy inside the simulation volume is calculated independently from the finite volume algorithm by using the final temperature increase, the specific heat, the density and the volume of each cell.
Second, the algorithm was tested as to whether it provided similar results for identical geometries with different spatial resolution, different time resolution as well as a different size of the simulation volume surrounding the implant.
Simulations
The cylindrical simulation volume basically is assumed as homogeneous tissue with the physical parameters (ρ, λ, c) from Table 1 and length Lsim and radius Rsim. A cylindrical implant with sizes Limp and Rimp is placed at the center of this volume. The simulation is calculated for three different models.
1. Heat generation inside a cylinder of length Limp and radius Rimp.
The heat generation is assumed to be uniform over the entire volume of Vimp = π R2impLimp; the entire simulation volume is assumed to be tissue.
2. Heat generation in nr metallic rings with rectangular cross section of rwire times xwire at radius Rimp.
The rings are equally spaced over the length Limp. These power generating rings are used instead of a solenoid with nr turns to retain the cylindrical symmetry and they stand for a model where the power loss is equally distributed on and nearby the circuit paths forming the inductor. For the rings freely selectable physical parameters c, λ and ρ can be used (Table 1). Partly these simulations assume a flow inside the implant volume not allowing a temperature increase for the region of flow.
3. Heat generation in a cylinder shell of length Limp and thickness rwire at radius Rimp.
This is an approximation of a modified solenoid structure with a high density of circuit paths on the cylinder surface. In this case, despite the assumed dense wire, most of the shell volume is tissue. Therefore the physical properties of tissue are used for the entire simulation volume.
None of the three cases exactly match the real situation. However, consistent results for all cases, which also are checked to the order of magnitude with the linear wire model of Eq. (11), could certify a realistic estimation.
Results
Test of Simulations: Endless wire model
For a given MRI sequence and a given resonator with known geometry it is possible to calculate the power loss density P* with Eq. (7). Using this power density P* it is possible to calculate the temperature difference between the wire surface and the difference between the wire surface and a specific radial distance after reaching the thermal equilibrium according to the model of Eq.(11). The results are summarized in Figure 5. The simulation of a linear wire is compared with a theoretical curve with respect to the same wire radius. To simulate an endless wire the energy exchange is restricted to radial direction by setting the heat sink shell temperature at maximal index x = m + 1 after each iteration not to zero but to the same value as at index x = m for all r respectively. Power generating metallic cells are assumed within the radius rwire. In radial direction, just a few cells are sufficient to describe the heat generating wire; the radial resolution Δr is chosen in such a way that n Δr - Δr/2 is equal to rwire. This choice takes into consideration that the center of the last cell is the reference radius rwire for the theoretical description according to Eq. (11). For the comparison with theory, the temperature increase in a simulation cell is set to a radial position just half between the walls of a simulation cell. During the temporal development the simulated radial temperature differences increasingly approach the analytical solution for the thermal equilibrium.
Figure 5 By the simulation calculated temperature differences and comparison with the theory. The graphs show the calculated temperature differences from the surface of a wire with radius rwire = 0.025 mm to a radial position R after 50 s, 200 s and 1200 s and also allows a comparison with the analytical solution of Eq. (11) (P* = 1 W/m).
Test of Simulations: Temporal and spatial resolution
For sufficiently short time steps Δt, the simulations show the expected behavior that the temperature increases monotonically in each simulation cell. For overly large time steps the energy flow out of a cell during a time step can become larger than the energy loss inside it. This leads to a physical incorrect situation with decreasing temperatures inside a cell. For this case the simulation produces huge deviations and the temperature differences oscillate in an unpredictable fashion during the temporal development.
Increasing the simulation volume without changing the spatial and temporal resolution as well as the size of the implant generates only slight changes (see additional file 1) at the cells corresponding to our approach that the surface of the simulation volume acts as a heat sink forcing ΔT to be zero on the surface. The tests of temporal and spatial resolution were carried out only for the short simulation period of 10 s, because the largest changes occur during the first simulation steps. Increasing solely the temporal resolution for a stable simulation affected the simulation results only minimally (Figure 6c,d). Even a huge increase by a factor of 1000 in temporal resolution between c and d does not show significant deviations. This increase allows also a comparison of Figure 6d with a better spatial resolution, which needs an increased temporal resolution.
Figure 6 Graphic explanation and test of the simulation with respect to temporal and spatial resolution on resonator no 2 (Table 3). a) radial view (radial axis not perspective) of the entire calculated simulation volume. The calculated matrix as well as the size of the simulation cells and the calculation time is shown in the tables below b) partial view of a c) radial view of the total implant showing also the not calculated part at negative axial positions and also doubling the radial component to a non existing (negative) radial component. The result is a visualization of a cut of the entire implant through the cylinder containing the cylinder axis. d) same as a with a temporal resolution better by a factor of 1000 (see table below). e) same as a with a temporal resolution better by a factor of 117 as well as three fold better spatial resolution (see table below) f) same as a with a temporal resolution better by a factor of 1000 as well as six fold better spatial resolution (see table below)
An improved spatial resolution without changes of the heat generating volume modifies the results very little (Figures 6d,e,f; 10b,d). Because an increased temporal or spatial resolution only has marginal influence on the simulation results, it is allowed to use the lowest possible temporal and spatial resolutions, which do not generate oscillatory deviations.
Figure 10 Power loss for resonator no 1 with two rings. a) partial radial view for one of two power generating rings of tissue (see table below) b) partial radial view for one of two power generating rings of titanium c) partial radial view for one of two power generating rings of iron d) partial radial view for one of two power generating rings of titanium
Finite volume analysis
According to Eq. (7) the with Q normalized volumetric power loss density PV = P/(Q VImp) for resonators exposed to an MRI sequence with maximum SAR (Table 2) can be calculated to 4 mW/cm3. Realistic vascular implants are in the order of 0.1 cm3 (e.g. coronary stents; r = 1.5 mm; l = 14 mm) to 50 cm3, (e.g. aortic stent grafts, vena cava filters; r = 18 mm; l = 50 mm). The achievable quality factor is low in ionic surroundings, such as tissue, but increases with the insulation of the solenoid wire (unpublished diploma thesis, Karsten Behrens, University of applied sciences TFH Berlin, 2001, Department Mathematik-Physik-Chemie, Prof. Vollmann). For thin insulations, which are necessary for the retention of the mechanical properties of the vascular implants, the achievable quality factor is below 5. However, the simulation as a worst case scenario partly uses quality factors far above this value to test the safety of the implants also for these conditions. As examples, the resonators described in Table 3 were used with the physical parameters of Table 1. As simulated time, an MR sequence (Table 2) exposure of ten minutes was assumed.
Table 2 Data for calculation of power density PV according to Eq. (7) for an MRI sequence with an SAR of 4 W/kg (manufacturer declaration)
permeability of vacuum μ0 [V s / (A m)] 12.6E-7
magnitude of magnetic excitation field B1 [μT] 25
repetition time of MRI sequence TR [s] 2.23
duration of one excitation τ [ms] 0.80
number of identical excitations during TR N 246
duty cycle cdc = N τ/TR cdc 0.09
pulse waveform modulation factor cpwm cpwm 0.45
Larmor frequency = resonance frequency of LC circuit ν0 [MHz] 63.8
power loss density PV = P/(Q Vimp) (Eq. (7) PV [mW/(cm3Q)] 4.0
Table 3 Geometrical data and power losses for example resonators
Resonator No 1 2 3
Radius of ring resonator Rimp [mm] 18 5.55 1.5
Length of ring resonator Limp [mm] 50 9.9 14
Dimension of ring wire axial xwire [mm] 0.25 0.15 0.10
Dimension of ring wire radial rwire [mm] 0.25 0.15 0.10
Number of rings (turns) nr 2 6 12
Volume of implant Vimp [cm3] 50 1 0.1
Quality factor Q 5 12.5 80
Total power loss of resonator Ploss [mW] 1000 50 32
SAR inside resonator (Eq. (3b)) SARinternal [W/kg] 1.8 1.7 5.1
Power loss of resonator due to SAR PSAR [mW] 0.09 0.0016 0.0005
Power losses PSAR and Ploss are calculated for the MR sequence of Table 2. SARinternal is evaluated with Eq. (3b). PSAR can be calculated using SARinternal and the mass inside the inductor (mimp = ρtissue Vimp), whereas Ploss is calculated with the appropriate constants according to Eq. (7). Ploss for resonator no 2 was verified experimentally with temperature increase measurements after placing the resonator inside a phantom with known heat capacity.
Heat generation inside a cylinder
For the three differently shaped prototype examples (Table 3) the heat generation was assumed to be uniform over the whole volume of the resonator inductance. For one resonator (no 2) the power loss Ploss was verified experimentally. Figures 7,8,9 show the results of these simulations. In no simulation the temperature increases reach a value of 5 K. Literature [17-21] on hyperthermia permits an estimation of the temperature resistance of normal animal cells and human cancer cells, depending on exposure time and temperature increase. The value chosen as physiologically critical, 5 K for 10 min to 15 min exposure time, is a worst case assumption, as various cells tolerate higher temperature increases as well as longer exposure times.
Figure 7 Homogeneous power loss inside resonator no 1. a) total cross section through cylinder containing the cylinder axis with an axial view b) radial view of part of the simulation volume; the dotted auxiliary lines denote the dimension of the implant
Figure 8 Homogeneous power loss inside resonator no 2. a) axial view of a total cross section through cylinder containing the cylinder axis (see table below) b) radial view of part of the simulation volume; the dotted auxiliary lines denote the dimension of the implant
Figure 9 Homogeneous power loss inside the entire resonator no 3. a) axial view of a total cross section through cylinder containing the cylinder axis (see table below) b) radial view of part of the simulation volume; the dotted auxiliary lines denote the dimension of the implant
Heat generation in nr metallic rings with rectangular cross section
Two rings are the worst case for resonator no 1 from Table 3 setting up a saddle coil like design. Figure 10 shows the results of the simulation. For a quality factor of 5 the temperature increases inside tissue reach 6.9 K. This is in accordance with Eq. (11). For a value of 4.4 W/m the temperature difference between the wire surface and a point at distance r = 10 cm is 9.4 K for the thermal equilibrium. Setting up a second heat sink simulating blood flow through the implant reduces the maximum temperature difference. Increasing the radius exposed to flow towards the radius of the implant increases the reduction. For an implant with the total inner volume exposed to blood flow the maximum temperature increase in the vessel wall only reaches 3.5 K (Figure 11f).
Figure 11 Power loss for resonator no 1 with two rings and with simulated blood flow of varying cross sections. a) partial radial view without heatsink inside resonator b) partial radial view with flow inside central region of 5 mm radius c) partial radial view with flow inside central region of 10 mm radius d) partial radial view with flow inside central region of 15 mm radius e) partial radial view with flow inside central region of 17 mm radius f) partial radial view with flow inside central region of 17.5 mm radius
For resonator no 2 (Table 3) 6 rings are shown in Figure 12, whereas for resonator no3 the results for 12 windings are shown in Figure 13. The material of the power generating rings has only a negligible influence on the calculated temperature maps except during the steps at the beginning of the simulation (Figures 10, 12). This is an additional proof of a correctly implemented algorithm because of two reasons. Firstly, the thermal conductivity of metal is so much larger than that of tissue, that the algorithm uses only the heat conducting path through tissue at a metal tissue interface. Secondly, the thermal capacity and the density of metal or tissue have no influence on the temperature map in the case of thermal equilibrium, because in that case heat just passes the simulation cells without changing their temperatures anymore. This argument has been mentioned already in the context of Eq. (11), which is valid for thermal equilibrium.
Figure 12 Power loss on six rings for resonator no 2. a) partial axial view for 3 of 6 power generating rings of tissue (see table below) b) partial axial view for 3 of 6 power generating rings of titanium c) partial axial view for 3 of 6 power generating rings of iron d) partial axial view for 3 of 6 power generating rings of tantalum
Figure 13 Power loss on twelve rings for resonator no 3. a) partial axial view for 6 of 12 power generating rings of titanium (see table below) b) partial axial view for 6 of 12 power generating rings of iron
Heat generation in a cylinder shell
The temperature peaks at the power generating rings become negligible for six or more rings after ten minutes simulated time. Therefore it was reasonable to choose a cylinder shell as power generating source. The simulations with a homogenous power dissipating cylinder shell instead of rings are shown from Figure 14 to 15 for resonators 2 and 3 of Table 3. None of these simulations shows a critical temperature increase.
Figure 14 Power loss over a cylinder shell for resonator no 2. a) partial axial view for power generating cylinder shell b) partial radial view for power generating cylinder shell
Figure 15 Power loss over a cylinder shell for resonator no 3. a) partial axial view for power generating cylinder shell b) partial radial view for power generating cylinder shell
Discussion
Test of simulations: endless wire model
The simulation of an endless wire uses the first of the three above mentioned models. Especially, the radius Rimp is taken as equal to rwire. The condition that the wire is endless can be realized in two ways. The first way is to choose a very long Limp and to evaluate the temperature increases only at positions on the center plane (index x = 1) and at radial distances r small compared to Limp. The second way is to modify the simulation software in such a way that the energy transfer in axial direction is excluded. This can be done by considering only volume elements of the center plane (index x = 1) and setting the heat sink shell temperature (index x = 2) after each iteration not to zero, but to the same value as at index x = 1 for all r respectively.
The second way has been realized. The common and expected result was that during the temporal development the simulated radial temperature differences increasingly approach the analytical solution for the thermal equilibrium (Eq. (11)). One comparison between a simulated and the analytical result is shown in Figure 6. Although a test on agreement of both results is independent of the specific choice of the power loss density P* [W/m], we have chosen a realistic value of P* in order to obtain realistic temperature increases. The power loss density P* was calculated with Eq. (7) for the MRI sequence of Table 2 for the example resonators (Table 2) to be between 0.24 W/m and 4.4 W/m. The simulation uses value of 1 W/m.
Test of simulations
A correct simulation of an endless wire should deliver results similar to the analytical curve of Eq. (11) using an implementation of the algorithm with no energy exchange in longitudinal (x) direction. The physical model of such a linear wire claims that the temperature difference between the wire and a cell at a certain distance increases monotonically with that distance. The theoretically calculated curve is valid after reaching the thermal equilibrium. With increasing simulated time the calculated curves show an increasingly close coincidence with the prediction of Eq.(11) (Figure 5), indicating the correctness of the algorithm.
The analytical linear wire model can also indicate which power per unit of wire length P* is safe for active magnetic implants (Figure 4). The theoretical estimation according to Eq.(11) shows for a power density larger than 2 W/m a risk of reaching temperatures increases greater than 5 K. The results do not vary with the time resolution Δt as long as the value is small enough to prevent the simulation from oscillating (Figure 6c,d). Also the spatial resolution (Δr, Δx) does not affect the results very much as long as the temperature maps appears smooth (Figure 6d,e,f). On the other hand, the calculation time explodes with a finer resolution. An increase of the spatial resolution always requires a better temporal resolution. Especially if a metallic cell completely surrounded by tissue is divided into a few metallic cells, a superior temporal resolution is necessary. For a metallic cell completely surrounded by tissue the shortest diffusion length equals roughly half of the spatial resolution with the diffusion path through tissue. For adjacent metallic cells the diffusion length equals the spatial resolution, but the diffusion path is completely through metal with a much larger thermal conductivity. The energy transfer for an identical temperature difference through this path is at least an order of magnitude larger than the one between a metal cell and a tissue cell. This higher energy transfer requires the drastically reduced time step Δt for a stable temporal development of the simulation. The calculation time for an identical simulated time increases by several orders of magnitude with metallic rings and a high spatial resolution and can easily reach computing times of several days on a normal desk top computer. Some examples for the calculation time is given in the second Table below Figure 6.
Power loss of resonator
Achievable quality factors were derived from the construction of experimental solenoid resonator prototypes (Table 3). The quality factor for a given geometry is strongly dependent on the environment and the thickness of the wire insulation. It is reduced substantially after changing the environment from air or distilled water to saline solution. Also Q decreases with a thinner insulation of the wire. Using this information it is possible to conclude that the main power losses are electric losses. Losses due to eddy currents induced by the magnetic field would not change with the thickness of the electric insulation. The dominance of the electric losses is also verified by comparing the power loss from eddy currents inside the active implant (Eq. (3)) with the total power loss (Eq. (7)). For all example prototypes in Table 3, the power loss calculated with the SAR inside the resonator (Eq. (3)) is lower by several orders of magnitude than the total power loss according to Eq. (7) (Table 3).
Realistic values for the quality factor of a resonator placed in ionic surroundings such as tissue are below 5 for all example resonators of Table 3. The calculated examples are worst case assumptions as they assume the maximum achievable Q of 5 for resonator no 1 and therefore the maximum power loss. For resonator no 2 (Q = 12.5) and no 3 (Q = 80) the chosen value is far beyond the reachable inside an ionic surrounding. These high values for resonator 2 and 3 are chosen to pronounce the safety statement for smaller resonators with quality factor values well above achievable values. The unrealistic high quality factors do not show physiologically critical temperature increases. For resonator number 1 (with a large volume) a realistic value was chosen, because in this case the temperature increase reaches physiologically dangerous values.
Part of the simulation results could be confirmed on stents implanted inside the aorta of rabbits [9,10]. Small coated resonators with a volume below 0.5 cm3 and quality factors of 3 to 4 were used. After excision and histopathologic examination, the tissue did not show any indications of heating after several MRI investigations. This is in coincidence with the simulation results, because even for small resonators with a much higher quality factor no dangerous heating was calculated.
Ten minutes of simulated time is 5 minutes less than the critical time according to the FDA regulations for imaging of the trunk assuming a SAR of 4 W/kg. But all simulations show only small changes after 10 minutes. Therefore it is unnecessary to increase the simulation time further. Nevertheless, one example of 15 minutes simulated time is given in the movie (see additional file 1) and shows only small changes in the last five minutes.
The weak influence on the distance between power generating cells and heat sink (the outer most layer) is also shown in the movie (see additional file 1). The difference between two almost identical calculations is shown. Starting from a first calculation with simulation parameters Lsim and Rsim, the second calculation uses doubled values of Lsim and Rsim without changing the spatial resolution or the other simulation parameters. This approach shifts the heat sink to a larger distance from the heat generating cell elements. As expected the heat sink layer drags down the temperature increase in its vicinity, but the effect is small if the volume is adequately chosen. The differences for identical cells between both simulations are shown by alternating both views a few times at the end of the movie (see additional file 1).
Finite volume analysis
For all three example resonators (Table 3) the first simulation model assumes a uniform heat generation over the whole volume of the inductor. Figures 7,8,9 show the results of the simulations. All temperature increases are below the critical value of 5 K. The maximum temperature is at the cylinder axis, corresponding to the center of the vessel for a vessel implant. The uniform heat generation inside homogeneous tissue is – as discussed above – not the best approximation. The ring model is a more appropriate match for realistic conditions.
Therefore resonator 1 is assumed as a solenoid with 2 turns (Figure 10). The simulation assumes these turns as metallic rings, which should be similar to a solenoid for the question of temperature increases. For resonator no 2, six rings are calculated (Figure 12) and resonator 3 is simulated with twelve rings (Figure 13). Only the simulation for resonator no 1 shows temperature increases over 5 K, which are located on and directly adjacent to the power generating rings. For resonator 2 the temperature increase with 6 rings is slightly above 1 K (Figure 12). The same is true for resonator no 3 with twelve rings (Figure 13).
For 6 or 12 rings and after 600 s calculated simulation time, the peak values near the location of the power generating rings inside the tissue become more and more negligible. Therefore a model using a power generating cylinder shell is appropriate for resonators with a reasonable number of turns.
For resonators no 2 and no 3, the cylinder shell simulation did not show any unsafe heating and is similar to the ring calculations. Neglecting the peaks of the ring simulation the maximum temperature increase is nearly identical.
Only the simulation for resonator no 1 with two rings reaches a critical temperature value above 5 K, all others stay below this value. For a large number of rings or, stated more precisely, a high wire density at the cylinder surface, it is possible to use the cylinder shell model (compare Figures 12, 14 and Figures 13, 15). As expected, the rings cause higher maximum temperature increases according to the higher local power density. The difference between the shell generating the entire power (Figures 14, 15) and a uniform power generating volume (Figures 8, 9) is mainly a different temperature distribution inside the inductor. The former simulation first shows a moderate increase in radial direction up to the power generating shell inside the volume, followed by a decrease. The latter has a larger peak value at the axis of the cylinder and falls off continuously with greater distance from the axis. The ring model is the most preferable because of the dominance of the electric losses on and near the wire of the inductor. Using different metals without altering the other simulation parameters changes the results only marginally (Figures 10b,c; 11 b-d; 12 a, b).
The simulation worked on a "worst case" basis, neglecting in most cases all cooling effects except the energy transport due to the thermal conductivity. Especially for vascular implants in the "normal case", some blood flow and blood perfusion, causing a much faster energy transport, will reduce the temperature increases. The effect of blood flow is simulated for the example of resonator 1 with the two rings. This was the only simulation reaching physiologically critical temperature increases. Blood flow is simulated by keeping the temperature difference at zero for a central part of the cylindrical implant. Changing the dimension of the central part can simulate an implant, which may be more or less infiltrated to the vessel wall or covered by tissue (thrombosis, calcification or intimal hyperplasia). The reduced temperature increase for resonator no 1 and for various sizes of the cooling blood flow is shown in Figure 11. For an implant placed directly inside the wall, with none or only a small amount of tissue between wire and blood flow the temperature increases drop to a physiologically tolerable value (Figure 11). Nevertheless for implants with a tissue coverage of a few mm, which can be plaque or a thrombosis, critical temperatures can be reached.
With respect to the overall power absorption greater than 100 Watts inside the human body during MRI investigations with a maximum SAR, one additional Watt inside the entire body is negligible.
Conclusion
This investigation assumes a "worst-case scenario" in different ways. Firstly, the resonator is assumed being perfectly aligned within the plane of the excitation field B1. Secondly, to some extent no blood flow inside the inductor (vessel) of the implant is included. Thirdly, no blood perfusion inside the tissue around the resonator is taken into account. Lastly, too large quality factors are used.
For most "normal" cases, intact active implants will therefore be less critical, but it is not possible to exclude the worst case conditions. Pathologies (thrombosis, intimal hyperplasia, plaque) may alter blood flow and perfusion in the area directly adjacent to the current paths of the resonator and certainly the resonator can be perfectly aligned to the plane of the exciting rf-field B1.
Especially for resonators with a large volume (such as resonator no 1) and with a small number of rings it is possible to reach critical temperature increases above 5 K (Figures 10, 11). For peripheral and especially cardiac vessels, even with an unrealistically high quality factor, the power loss is too low for dangerous heating. On the other hand, stent grafts or vena cava filters built as active implants can reach volumes of a few ten cm3, which may be dangerous, if the tissue around the wires of the implant is not exposed to blood flow or sufficient blood perfusion. To reduce the risk for such active implants, the quality factor has to be low. This reduces also the amplification of the MR signal the resonator is made for and can thus be dispensed with altogether.
As the above simulations assume properly working resonance circuits without any failures on the electric paths of the system, one worst case scenario was not presented. Defects such as ruptures or partial ruptures may generate a relatively high resistance over very short distances. The current flow through this resistance can produce a large power loss inside an extremely small volume. This can generate a very high power density which, even for small implants, may induce physiologically critical temperature increases for a small volume. In fact the analysis of these "hot spots" is important, because ruptures of stent struts are likely and a high power density can occur also for smaller implants. An additional investigation estimating the maximum possible power loss inside such "hot spots" and the resulting temperature maps around them is necessary to check the safety of active implants under these circumstances. This is being prepared for future publication.
The study protocols for the cited animal experiments were approved by the responsible authority (Landesamt für Arbeitsschutz, Gesundheitsschutz und technische Sicherheit, Berlin, G 0142/99).
Authors' contributions
MB has drafted the investigation, written the manuscript, coded the simulation and contributed to the theory and part of the cited experimental work.
WV has drafted the theoretical part of the manuscript and reviewed the manuscript. Also part of the cited experimental tests for power loss, quality factor measurements and prototype constructions are from WV.
JS was responsible for the cited animal experiments and provided the in vivo images of prototypes in an animal model. He has also contributed to the medical background and has reviewed the manuscript.
DG has enabled and reviewed the manuscript, contributed to the medical background, and was helpful during the investigation with discussions.
Supplementary Material
Additional File 1
Movie with an example of a time developing temperature map This movie (animated GIF) shows the time development over a period of 900 s, which is the maximum permitted time for the imaging of the trunk with an SAR of 4 W/kg (manufacturer declaration, sequence of Table 2). At the end of the simulated time, the changes in the temperature distribution become small. For complete information, a 3D axial view as well as a 3D radial view is shown simultaneously. At the end of the movie two different simulations are shown alternately, which indicate the changes of the temperature map after 900 s due to a larger simulation volume shifting the heat sink more away from cells with power loss. One of the alternating results was calculated using a 250 × 250 matrix for a distance of 0 mm to 25 mm for r and x respectively. The second map was calculated for a 500 × 500 matrix for a distance of 50 mm for r and x respectively. Only the inner 250 × 250 points are plotted for a comparable size for both calculations. It can be seen that the temperature distribution is almost identical apart from the fact that, for x≈25 mm and r≈25 mm, the simulation with more cells shows a slight deviation from zero. The simulation with the smaller matrix shows a straight zero line, which is naturally because this is the boundary condition for this simulation. The small difference points out that the boundary condition with a heat sink works very well as long as the absolute value of the gradient at the boundary is low.
Click here for file
Acknowledgements
Many thanks to Peter van Leeuwen, a bilingual native English speaker who has reviewed and corrected the entire "German English" manuscript.
==== Refs
Bartels LW Smits HF Bakker CJ Viergever MA MR imaging of vascular stents: effects of susceptibility, flow, and radiofrequency eddy currents J Vasc Interv Radiol 2001 12 365 371 11287516
Hug J Nagel E Bornstedt A Schnackenburg B Oswald H Fleck E Coronary arterial stents: safety and artifacts during MR imaging Radiology 2000 216 781 787 10966711
Maintz D Kugel H Schellhammer F Landwehr P In vitro evaluation of intravascular stent artifacts in three-dimensional MR angiography Invest Radiol 2001 36 218 224 11283419 10.1097/00004424-200104000-00004
Meyer JM Buecker A Schuermann K Ruebben A Guenther RW MR evaluation of stent patency: in vitro test of 22 metallic stents and the possibility of determining their patency by MR angiography Invest Radiol 2000 35 739 746 11204801 10.1097/00004424-200012000-00008
Schnall MD Barlow C Hirihara Subrmaniam V Wireless implanted magnetic resonance probes for in vivo NMR J Magn Reson 1986 68 161 167
Farmer TH Cofer GP Johnson GA Maximizing contrast to noise with inductively coupled implanted coils Invest Radiol 1990 25 552 8 2345087
Patent of Simag GmbH, Teltowkanalstr.2, D-12247 Berlin, Germany, EP1092985A3
Quick HH Kuehl H Kaiser G Bosk S Debatin JF Ladd ME Inductively coupled stent antennas in MRI Magn Reson Med 2002 48 781 90 12417992 10.1002/mrm.10269
Kivelitz D Wagner S Schnorr J Wetzler R Busch M Melzer A Taupitz M Hamm B A Vascular Stent as an Active Component for Locally Enhanced Magnetic Resonance Imaging: Initial In Vivo Imaging Results After Catheter-guided Placement in Rabbits Invest Radiol 2003 38 147 152 12595794 10.1097/00004424-200303000-00002
Kivelitz D Wagner S Hansel J Schnorr J Wetzler R Busch M Melzer A Taupitz M Hamm B The active magnetic resonance imaging stent (AMRIS): initial experimental in vivo results with locally amplified MR angiography and flow measurements Invest Radiol 2001 36 625 631 11606839 10.1097/00004424-200111000-00001
Athey TW Ross RJ Ruggera PS Magnetic Fields Associated with Nuclear Magnetic Resonance Medical Imaging System Mag Res Imaging 1982 1 149 154 10.1016/0730-725X(82)90206-5
Margulis AR Higgins CB Kaufman L Crooks LE Clinical Magnetic Resonance Imaging Radiology Research and Education Foundation 1983 Library of Congress Catalog Card number 83-81823
Johnson CC Guy AW Nonionizing Electromagnetic Wave Effects in Biological Materials and Systems Proc of the IEEE 1972 60
Yeung CJ Susil RC Atalar R RF Safety of Wires in Interventional MRI: Using a Safety Index Magnetic Resonance in Medicine 2002 47 187 193 11754458 10.1002/mrm.10037
Hand JW Lau RW Lagendijk JJW Ling J Burl M Young IR Electromagnetic and Thermal Modeling of SAR and Temperature Fields in Tissue Due to an RF Decoupling Coil Magnetic Resonance in Medicine 1999 42 183 192 10398965 10.1002/(SICI)1522-2594(199907)42:1<183::AID-MRM24>3.0.CO;2-I
Liao SamuelY Microwave Devices and Circuits 1980 Prentice-Hall, Englewood Cliffs, New Jersey 134
Sminia P Haveman J Wondergem J van Dijk JD Lebesque JV Effects of 434 MHz microwave hyperthermia applied to the rat in the region of the cervical spinal cord Int J Hyperthermia 1987 3 441 52 3681044
Haveman J Van Der Zee J Wondergem J Hoogeveen JF Hulshof MC Effects of hyperthermia on the peripheral nervous system: a review Int J Hyperthermia 2004 20 371 91 15204519 10.1080/02656730310001637631
Khoei S Goliaei B Neshasteh-Riz A Deizadji A The role of heat shock protein 70 in the thermoresistance of prostate cancer cell line spheroids FEBS Lett 2004 561 144 8 15013766 10.1016/S0014-5793(04)00158-9
Poland J Schadendorf D Lage H Schnolzer M Celis JE Sinha P Study of therapy resistance in cancer cells with functional proteome analysis Clin Chem Lab Med 2002 40 221 34 12005211 10.1515/CCLM.2002.037
Sinha P Poland J Schnolzer M Celis JE Lage H Characterization of the differential protein expression associated with thermoresistance in human gastric carcinoma cell lines Electrophoresis 2001 22 2990 3000 11565793 10.1002/1522-2683(200108)22:14<2990::AID-ELPS2990>3.0.CO;2-F
Gowrishankar TR Stewart DA Martin GT Weaver JC Transport lattice models of heat transport in skin with spatially heterogeneous, temperature-dependent perfusion BioMedical Engineering OnLine 2004 3 42 15548324 10.1186/1475-925X-3-42
Geddes LA Baker LE The specific resistance of biological material – a compendium of data for the biomedical engineer and physiologist Med & Biol Eng 1967 5 271 293 6068939
Holmes KennethR
| 15819973 | PMC1087857 | CC BY | 2021-01-04 16:37:36 | no | Biomed Eng Online. 2005 Apr 8; 4:25 | utf-8 | Biomed Eng Online | 2,005 | 10.1186/1475-925X-4-25 | oa_comm |
==== Front
Biomed Eng OnlineBioMedical Engineering OnLine1475-925XBioMed Central London 1475-925X-4-251581997310.1186/1475-925X-4-25ResearchFinite volume analysis of temperature effects induced by active MRI implants with cylindrical symmetry: 1. Properly working devices Busch Martin HJ [email protected] Wolfgang [email protected] Jörg [email protected]önemeyer Dietrich HW [email protected] Research and Development Center for Microtherapy (EFMT), D-44799 Bochum, Germany2 TFH University of Applied Sciences, D-13353 Berlin, Germany3 Institut für Radiologie, Charité, Medizinische Fakultät, Humboldt-Universität zu Berlin, D-10117 Berlin, Germany4 Grönemeyer Institute for Microtherapy, University of Witten/Herdecke, D-44799 Bochum, Germany2005 8 4 2005 4 25 25 16 11 2004 8 4 2005 Copyright © 2005 Busch et al; licensee BioMed Central Ltd.2005Busch 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
Active Magnetic Resonance Imaging implants are constructed as resonators tuned to the Larmor frequency of a magnetic resonance system with a specific field strength. The resonating circuit may be embedded into or added to the normal metallic implant structure. The resonators build inductively coupled wireless transmit and receive coils and can amplify the signal, normally decreased by eddy currents, inside metallic structures without affecting the rest of the spin ensemble. During magnetic resonance imaging the resonators generate heat, which is additional to the usual one described by the specific absorption rate. This induces temperature increases of the tissue around the circuit paths and inside the lumen of an active implant and may negatively influence patient safety.
Methods
This investigation provides an overview of the supplementary power absorbed by active implants with a cylindrical geometry, corresponding to vessel implants such as stents, stent grafts or vena cava filters. The knowledge of the overall absorbed power is used in a finite volume analysis to estimate temperature maps around different implant structures inside homogeneous tissue under worst-case assumptions. The "worst-case scenario" assumes thermal heat conduction without blood perfusion inside the tissue around the implant and mostly without any cooling due to blood flow inside vessels.
Results
The additional power loss of a resonator is proportional to the volume and the quality factor, as well as the field strength of the MRI system and the specific absorption rate of the applied sequence. For properly working devices the finite volume analysis showed only tolerable heating during MRI investigations in most cases. Only resonators transforming a few hundred mW into heat may reach temperature increases over 5 K. This requires resonators with volumes of several ten cubic centimeters, short inductor circuit paths with only a few 10 cm and a quality factor above ten. Using MR sequences, for which the MRI system manufacturer declares the highest specific absorption rate of 4 W/kg, vascular implants with a realistic construction, size and quality factor do not show temperature increases over a critical value of 5 K.
Conclusion
The results show dangerous heating for the assumed "worst-case scenario" only for constructions not acceptable for vascular implants. Realistic devices are safe with respect to temperature increases. However, this investigation discusses only properly working devices. Ruptures or partial ruptures of the wires carrying the electric current of the resonance circuits or other defects can set up a power source inside an extremely small volume. The temperature maps around such possible "hot spots" should be analyzed in an additional investigation.
==== Body
Background
Metallic implants often cause distortions inside Magnetic Resonance (MR) images. These effects arise either from the different susceptibility of tissue and metal, generating a discontinuity of the local field strength at the interface, or from the Faraday cage effect, which is set up by induced eddy currents on the metallic implant structure [1-4]. An advantageous solution to overcome the eddy current shielding is to amplify the transmitted signal to and the detected signal from the spin ensemble appropriately for the Faraday cage. A local amplification of excitation and detection for a magnetic resonance signal can be achieved by using a resonator tuned to the resonance frequency of the MR system [5-8]. Such a resonator can be integrated into the metallic structure of the implant itself or can be added around the normal implant structure. The great advantage of such an active Magnetic Resonance Imaging implant (aMRIi) is the local amplification of the signal only where it is needed. The decreased excitation pulse can be amplified for the spin ensemble inside the faraday cage without affecting the signal and contrast behavior of the rest of the excited volume. The possible advantages of active MRI implants are demonstrated in Figure 1 showing the possible amplification inside a Faraday cage and the potential to get functional information out of such an implant.
Figure 1 Commercial Stent with additional resonator inside a water bath (a) and an implanted prototype in a rabbit aorta (b) with velocity curve over the cardiac cycle (c). a) shows an MR image of a commercially available nitinol stent with an additional resonator around a part of the metallic structure immersed in a water bath. The MR signal inside the stent lumen decreases outside the resonator (arrows) in contrast to the enhanced signal inside the resonator. b) shows one example of an in vivo image with a prototype implanted into a rabbit aorta [9, 10]. c) shows the measured velocity curve inside the prototype of image b and the spline approximation delivered by the the MRI System.
The amplification is adjustable by the quality factor Q of the resonance circuit and allows the imaging of the inner volume of such metallic implants as well as gathering functional information from this location, for example, velocity or flow. For these special investigations inside the implant volume the quality factor should be as high as possible to increase the signal to noise ratio for this region and to get as low signal as possible from spins outside the volume of interest. For an overall overview of the implant inside the anatomic surrounding it is only necessary to raise the decreased signal inside the Faraday cage to the normal level, which is possible with a lower quality factor. The active implant locally increases the absorbed power during MR investigations. The temperature increase induced by this extra power loss around vessel implants with a cylindrical geometry is calculated with a finite volume analysis for a "worst-case scenario". For most of the simulations only thermal heat conduction is assumed as "worst case" without cooling due to blood perfusion and without cooling due to blood flow inside the active implant. Partly also a blood flow is considered carrying energy out of the simulation volume. The calculations are carried out with respect to power loss, time of rf-exposure and size of the aMRIi. The aim of this investigation was to perform a detailed assessment of the safety for such implants under "worst case" assumptions.
Methods
Theory
Specific Absorption Rate (SAR)
The local specific absorption rate (SAR; [W/kg]) for electromagnetic radiation is defined as power absorbed per unit of mass of an object at location r [11-16],
where σ [S/m] is the electric conductivity, ρ [kg/m3] is the mass density, E [V/m] is the amplitude of a sinusoidal time dependent electric field and r is the position vector. Assuming that the amplitude of the electric field only arises by induction from a uniform, linearly polarized magnetic field with amplitude Bm [T] and angular frequency ω [rad/s] in a homogeneous body of rotational symmetry, the following equation holds [11]
with r as component of cylindrical coordinates and r = 0 on the rotational symmetry axis aligned parallel to the magnetic field Bm. Combining Eq. (1,2) the spatially averaged SAR over a cylinder with radius Rcyl and using dV = 2 π r h dr yields
For pulsed MR sequences two correction factors are necessary,
The duty cycle factor cdc equals N*τ/TR, where TR [s] is the MR-sequence repetition time, N is the number of identical excitations during TR and τ [s] is the duration of one excitation pulse. cdc equals to the time ratio of "rf-on" to "rf-on + rf-off" during an MR scan. cpwm corrects for rf-pulses which are not rectangular in shape and is the ratio between the energy of the MR excitation pulse and the energy of a rectangular shaped pulse with identical amplitude and identical duration.
Principles of B1-field amplification inside a resonator
A circularly polarized magnetic field with magnitude B1 or a linearly polarized magnetic field with amplitude Bm for excitation is amplified inside an aMRIi with respect to the used resonance circuit. Because of the shape of vascular implants and to simplify calculations, a solenoid is chosen as inductor of the resonator and is investigated as a magnetic antenna. For the theoretical estimation, the axis of the solenoid resonator is assumed to be parallel to a sinusoidal linearly polarized magnetic field with amplitude Bm, or equivalently to be in the plane of a circularly polarized magnetic field with magnitude B1. For the resonance case (2πν0 = ω0 = (LC)-1/2), where the overall resistance of the resonator is just R, the following equation can be derived from the law of induction, Ohms law and the definition of the quality factor Q of a resonator,
where Vind [V] and Vself [V] are the induced and self-induced voltage respectively, Z [Ω] is the impedance, R [Ω] is the resistance, L [H] is the inductance of the inductor and αres and α are the flip angles inside the resonator inductance and outside the resonator respectively. The total magnetic field inside the resonator arises from both components Bm and Bres. For large quality factors (Q>>1) Bres dominates Bm. With Eqs(3, 4) it is possible to calculate the magnetically induced SAR and with the SAR the corresponding power loss inside a resonator. But this takes into account only power losses due to eddy currents and not the total power loss of a resonator.
Total power loss of a resonator
The overall power loss P [W] also includes the electric losses on and around the resonator. It is given by the following basic equation [16],
where Wtotal [J] is the energy stored inside the resonator.
The total energy Wtotal can be expressed by the energy stored in the magnetic field of maximum amplitude using the following basic relations of a solenoid inductance with nr turns (),
where V, A, ℓ with index imp are the volume, cross section and length of the implant inductor. For a pulsed MR sequence with excitation magnitude B1 the total power loss can be calculated by combining Eqs. (4, 5, 6) to
A comparison of Eqs. (3b, 7) shows that for a certain resonator tuned to the resonance frequency of a specific MR field strength, the supplementary absorbed power is proportional to the SAR of the MR sequence due to the identical dependence on cdc, cpwm and B12. For a specific sequence on a MRI system with a definite resonance frequency, the extra power is proportional to the quality factor Q and the volume Vimp of the inductance (Eq. (7)). The proportionality to the quality factor Q may be surprising, because it states that for better quality factors the power loss is larger. This is due to the total energy of the resonator, which depends quadratically on the field strength inside the resonator. This field strength is proportional to the exciting field B1 (or Bm) as well as the quality factor Q. Combining Eq. (5) and Eq. (6) alters the inverse proportionality to Q (Eq. (5)) to a proportionality (Eq. (7)) with respect to the field established by the transmit coil. Examining the power loss with respect to the magnetic field inside the resonator confirms the inverse proportionality to Q.
Finite volume analysis
Principles
All temperature increases are calculated as temperature difference maps by using a simulation volume divided into many small simulation cells. The energy exchange ΔE [J] between two simulation cells with a specific contact area A [m2], temperature difference ΔT [K] and heat diffusion path length Δx [m] for a time interval Δt [s] is given by the equation for heat conduction as,
where λ [W/(m K)] is the thermal conductivity.
The total energy change ΔEtot of one cell during a time interval Δt is the sum of exchanges of this cell with all adjacent cells with non zero contact area and the energy change due to the power loss pcell inside the cell. From this total energy change the temperature difference ΔT* can be calculated,
where c [J/(kg K)] is the specific thermal capacity of the cell material and Vcell is the cell volume.
This investigation did not use any commercially available finite volume package. The simulation is self-coded for problems with cylindrical geometry in Kylix and Delphi, a software development environment, based on object oriented Pascal. The graphical outputs are mostly generated by an evaluation (registered β-test) version of Teechart 7 used within the Delphi and Kylix environment. The simulation describes the time developing temperature difference maps around an aMRIi with cylindrical geometry for a constant total power loss P.
The entire simulation volume is assumed as a cylinder with length Lsim and radius Rsim. Adequate for such a volume are cylinder coordinates. Because of the rotational symmetry, the temperature distribution is only dependent on the axial position x and the distance r from the cylinder axis (Figures 2, 3). Because of the mirror symmetry to the plane orthogonal to the cylinder axis at center, the simulation is designed to calculate the temperature distribution only on one side of the center plane. (Figures 2, 3). Using both symmetries one can divide the entire cylindrical simulation volume with length Lsim and radius Rsim into the following cells, which are sufficient to obtain complete information of the entire simulation volume. The calculation volume consists of n cylindrical shaped pieces with length Δx and index x = 1, 2, ...., n denoting the position on the cylinder axis. Each sub cylinder of length Δx is divided in one cylinder with radius Δr and index r = 1 and m-1 shells with outer radius r* Δr, thickness Δr and index r = 2, 3, ...., m. The entire simulation volume is represented by a two dimensional array of cells C [r,x] (r = 1,2,3,....m; x = 1,2,....n). Most of the cells have contact to 4 adjacent cells with contact areas different from zero (Eqs (10a, 10b)). These contact areas as well as the volumes of the cells (Eq. (10c)) only vary with the index r.
Figure 2 Definition of the cell alignment within the cylindrical simulation volume. This figure shows the geometry of the simulation volume as example with 4 × 4 cells, typical simulation volumes are set up from 100 × 100 to 500 × 500 cells. Each simulation volume is just one half of the entire volume of the implant, because its mirror symmetry allows to restrict the simulation to half the volume. The plane of mirror symmetry is on the left side of the figure. (see also Fig. 3).
Figure 3 Definition of the cell alignment within the total examined cylindrical volume for the implant. Example of an entire volume with 8 × 4 cells. The part with complete positive indices corresponds to the simulation volume of Fig. 2. Because of the plane of mirror symmetry (dark line in the middle) only this part has to be calculated. (ΔT [r,x] = ΔT [r,-x])
Al[r] [m2] is the contact area in both cylinder axis directions (from index x to x-1 and to x+1) whereas Ar[r] [m2] is the contact area in radial direction from index r to r+1. The contact area in radial direction from index r to r-1 is identical to the area Ar[r-1]. A last "shell" with cells as heat (energy) sink is placed as one boundary of the simulation volume at r = m + 1 and x = n + 1. These cells always keep a constant temperature (ΔT = 0), even when receiving energy during one simulation step. Partly the simulations use a second heat sink representing a blood flow through the inner volume of the implant. For this second heat sink the temperature differences of all cell elements below a radius rflow*Δr are also kept zero independent of the energy transfer from the cell elements with higher index r. This approximation assumes that the energy applied to the inner cylinder volume with blood flow is completely transported away from the simulation volume during one time step Δt. At r = 1 and x = 1 as well as at r = m + 1 and x = n + 1 the simulated volume has its boundaries. The symmetry and the use of cylinder coordinates imply that cells with index r = 1 or x = 1 can not exchange energy with cells at a lower index. For x = 1 the symmetry plane defines an identical temperature at index x = -1, (Figure 3) with no energy exchange across the symmetry plane. For r = 1 no cells with lower index r exist and therefore also no energy exchange is possible.
For all heat generating cell elements (pcell[r,x]>0 Eq. (9)) the physical parameters λ, ρ and c can be set freely. For all other cell elements they are set as tissue. The presented simulations use different metal parameters for heat generating cells (Table 1). The different physical parameters of metal and tissue for different simulation cells results in discontinuities at the metal tissue interface. Assuming the heat conduction path is from the center of one finite volume to the center of a neighboring finite volume, it takes place over two different materials with only half the diffusion length for each of the materials. Because of the much lower thermal conductivity of tissue compared to that of metal it is a good approximation to use only the diffusion length through tissue.
Table 1 Physical constants of tissue, titanium, iron and tantalum
material density ρ specific heat c thermal conduct. λ electric conduct. σ
[kg/m3] [J/(kg K] [W/(m·K)] [S/m]
tissue 1000 [22] 3650 [23] 0.5 [24] 0.8–8.0 [22]
saline 1003 4185 1.45*
titanium 4510 523 21.9 2.56 106
iron 7870 449 80.2 10.4 106
tantalum 16680 140 57.5 7.63 106
* The electric conductivity of physiologic saline for the calculation of the SAR was determined to be 1.45 S/m at 22°C (295 K) and a frequency of 63.5 MHz. (Personal communication with Mr. Alexander Walkov at "Physikalisch-Technische Bundesanstalt" in Berlin, Germany)
The simulation starts with a temperature field T [r,x] = 0 for all r and x. For each iteration step with duration Δt the total energy change for each cell element is calculated according to Eq. (9a). This change ΔEtot[r,x] of one cell is the sum of five different energies. One is due to the power loss pcell = p[r,x] defined for a cell element at index r and x. The four others are energy exchanges (Eq. (8), depending on the contact areas (Eq. (10a, b)) and the temperature differences between adjacent cells with respect to the equivalent heat diffusion length through the material of the examined region. From ΔEtot [r,x] the temperature change ΔT*[r,x] is calculated according to Eq. (9b). This value is added to the prior value (Tnew [r,x] := Told[r,x] + ΔT*[r,x)) for each cell during the whole iteration process. The entire simulated time tsim consists of q iterations with time interval Δt (tsim = q * Δt).
Analytical model for a heat generating linear wire as test of the simulation
One test of the principal correctness of the self-coded simulation is possible by a comparison of the simulation results with those of an analytical solution. An easy straightforward analytical solution is available for an endless linear wire with radius rwire, which is heated by a constant power density P* [W/m], defined as power per wire length [W/m]. After reaching the thermal equilibrium, the power associated with P* penetrates through every cylinder surface surrounding the linear power source independent of the radius r (with r>rwire). The temperature difference ΔT between the wire surface and a point at radial position R can be calculated in a homogeneous medium from the equation for heat conduction.
The thermal capacity does not appear in Eq. (11), because of the assumed thermal equilibrium. A simulation for a wire with power loss P* and with energy exchange only in radial direction should approach, after sufficient time, similar temperature differences between the wire surface and a point at distance (R-rwire) (Figure 4).
Figure 4 Temperature difference ΔT from wire surface with radius rwire = 0.05 mm to radial position R for different power densities P* [W/m]. The graphs show the temperature differences from the wire surface at position rwire to the position at radius R (cylinder axis defines R = 0). The starting point at rwire (0.05 mm) is so close to R = 0 that it coincides with R = 0 in the drawing.
Control of simulation results
The implemented algorithm was controlled with different checks, beside the afore-mentioned comparison with an analytical model. First, the total energy uptake of the heat sink surface is added up during all iteration steps. This energy summed with the energy stored inside the simulation volume must equal the total applied energy. The energy inside the simulation volume is calculated independently from the finite volume algorithm by using the final temperature increase, the specific heat, the density and the volume of each cell.
Second, the algorithm was tested as to whether it provided similar results for identical geometries with different spatial resolution, different time resolution as well as a different size of the simulation volume surrounding the implant.
Simulations
The cylindrical simulation volume basically is assumed as homogeneous tissue with the physical parameters (ρ, λ, c) from Table 1 and length Lsim and radius Rsim. A cylindrical implant with sizes Limp and Rimp is placed at the center of this volume. The simulation is calculated for three different models.
1. Heat generation inside a cylinder of length Limp and radius Rimp.
The heat generation is assumed to be uniform over the entire volume of Vimp = π R2impLimp; the entire simulation volume is assumed to be tissue.
2. Heat generation in nr metallic rings with rectangular cross section of rwire times xwire at radius Rimp.
The rings are equally spaced over the length Limp. These power generating rings are used instead of a solenoid with nr turns to retain the cylindrical symmetry and they stand for a model where the power loss is equally distributed on and nearby the circuit paths forming the inductor. For the rings freely selectable physical parameters c, λ and ρ can be used (Table 1). Partly these simulations assume a flow inside the implant volume not allowing a temperature increase for the region of flow.
3. Heat generation in a cylinder shell of length Limp and thickness rwire at radius Rimp.
This is an approximation of a modified solenoid structure with a high density of circuit paths on the cylinder surface. In this case, despite the assumed dense wire, most of the shell volume is tissue. Therefore the physical properties of tissue are used for the entire simulation volume.
None of the three cases exactly match the real situation. However, consistent results for all cases, which also are checked to the order of magnitude with the linear wire model of Eq. (11), could certify a realistic estimation.
Results
Test of Simulations: Endless wire model
For a given MRI sequence and a given resonator with known geometry it is possible to calculate the power loss density P* with Eq. (7). Using this power density P* it is possible to calculate the temperature difference between the wire surface and the difference between the wire surface and a specific radial distance after reaching the thermal equilibrium according to the model of Eq.(11). The results are summarized in Figure 5. The simulation of a linear wire is compared with a theoretical curve with respect to the same wire radius. To simulate an endless wire the energy exchange is restricted to radial direction by setting the heat sink shell temperature at maximal index x = m + 1 after each iteration not to zero but to the same value as at index x = m for all r respectively. Power generating metallic cells are assumed within the radius rwire. In radial direction, just a few cells are sufficient to describe the heat generating wire; the radial resolution Δr is chosen in such a way that n Δr - Δr/2 is equal to rwire. This choice takes into consideration that the center of the last cell is the reference radius rwire for the theoretical description according to Eq. (11). For the comparison with theory, the temperature increase in a simulation cell is set to a radial position just half between the walls of a simulation cell. During the temporal development the simulated radial temperature differences increasingly approach the analytical solution for the thermal equilibrium.
Figure 5 By the simulation calculated temperature differences and comparison with the theory. The graphs show the calculated temperature differences from the surface of a wire with radius rwire = 0.025 mm to a radial position R after 50 s, 200 s and 1200 s and also allows a comparison with the analytical solution of Eq. (11) (P* = 1 W/m).
Test of Simulations: Temporal and spatial resolution
For sufficiently short time steps Δt, the simulations show the expected behavior that the temperature increases monotonically in each simulation cell. For overly large time steps the energy flow out of a cell during a time step can become larger than the energy loss inside it. This leads to a physical incorrect situation with decreasing temperatures inside a cell. For this case the simulation produces huge deviations and the temperature differences oscillate in an unpredictable fashion during the temporal development.
Increasing the simulation volume without changing the spatial and temporal resolution as well as the size of the implant generates only slight changes (see additional file 1) at the cells corresponding to our approach that the surface of the simulation volume acts as a heat sink forcing ΔT to be zero on the surface. The tests of temporal and spatial resolution were carried out only for the short simulation period of 10 s, because the largest changes occur during the first simulation steps. Increasing solely the temporal resolution for a stable simulation affected the simulation results only minimally (Figure 6c,d). Even a huge increase by a factor of 1000 in temporal resolution between c and d does not show significant deviations. This increase allows also a comparison of Figure 6d with a better spatial resolution, which needs an increased temporal resolution.
Figure 6 Graphic explanation and test of the simulation with respect to temporal and spatial resolution on resonator no 2 (Table 3). a) radial view (radial axis not perspective) of the entire calculated simulation volume. The calculated matrix as well as the size of the simulation cells and the calculation time is shown in the tables below b) partial view of a c) radial view of the total implant showing also the not calculated part at negative axial positions and also doubling the radial component to a non existing (negative) radial component. The result is a visualization of a cut of the entire implant through the cylinder containing the cylinder axis. d) same as a with a temporal resolution better by a factor of 1000 (see table below). e) same as a with a temporal resolution better by a factor of 117 as well as three fold better spatial resolution (see table below) f) same as a with a temporal resolution better by a factor of 1000 as well as six fold better spatial resolution (see table below)
An improved spatial resolution without changes of the heat generating volume modifies the results very little (Figures 6d,e,f; 10b,d). Because an increased temporal or spatial resolution only has marginal influence on the simulation results, it is allowed to use the lowest possible temporal and spatial resolutions, which do not generate oscillatory deviations.
Figure 10 Power loss for resonator no 1 with two rings. a) partial radial view for one of two power generating rings of tissue (see table below) b) partial radial view for one of two power generating rings of titanium c) partial radial view for one of two power generating rings of iron d) partial radial view for one of two power generating rings of titanium
Finite volume analysis
According to Eq. (7) the with Q normalized volumetric power loss density PV = P/(Q VImp) for resonators exposed to an MRI sequence with maximum SAR (Table 2) can be calculated to 4 mW/cm3. Realistic vascular implants are in the order of 0.1 cm3 (e.g. coronary stents; r = 1.5 mm; l = 14 mm) to 50 cm3, (e.g. aortic stent grafts, vena cava filters; r = 18 mm; l = 50 mm). The achievable quality factor is low in ionic surroundings, such as tissue, but increases with the insulation of the solenoid wire (unpublished diploma thesis, Karsten Behrens, University of applied sciences TFH Berlin, 2001, Department Mathematik-Physik-Chemie, Prof. Vollmann). For thin insulations, which are necessary for the retention of the mechanical properties of the vascular implants, the achievable quality factor is below 5. However, the simulation as a worst case scenario partly uses quality factors far above this value to test the safety of the implants also for these conditions. As examples, the resonators described in Table 3 were used with the physical parameters of Table 1. As simulated time, an MR sequence (Table 2) exposure of ten minutes was assumed.
Table 2 Data for calculation of power density PV according to Eq. (7) for an MRI sequence with an SAR of 4 W/kg (manufacturer declaration)
permeability of vacuum μ0 [V s / (A m)] 12.6E-7
magnitude of magnetic excitation field B1 [μT] 25
repetition time of MRI sequence TR [s] 2.23
duration of one excitation τ [ms] 0.80
number of identical excitations during TR N 246
duty cycle cdc = N τ/TR cdc 0.09
pulse waveform modulation factor cpwm cpwm 0.45
Larmor frequency = resonance frequency of LC circuit ν0 [MHz] 63.8
power loss density PV = P/(Q Vimp) (Eq. (7) PV [mW/(cm3Q)] 4.0
Table 3 Geometrical data and power losses for example resonators
Resonator No 1 2 3
Radius of ring resonator Rimp [mm] 18 5.55 1.5
Length of ring resonator Limp [mm] 50 9.9 14
Dimension of ring wire axial xwire [mm] 0.25 0.15 0.10
Dimension of ring wire radial rwire [mm] 0.25 0.15 0.10
Number of rings (turns) nr 2 6 12
Volume of implant Vimp [cm3] 50 1 0.1
Quality factor Q 5 12.5 80
Total power loss of resonator Ploss [mW] 1000 50 32
SAR inside resonator (Eq. (3b)) SARinternal [W/kg] 1.8 1.7 5.1
Power loss of resonator due to SAR PSAR [mW] 0.09 0.0016 0.0005
Power losses PSAR and Ploss are calculated for the MR sequence of Table 2. SARinternal is evaluated with Eq. (3b). PSAR can be calculated using SARinternal and the mass inside the inductor (mimp = ρtissue Vimp), whereas Ploss is calculated with the appropriate constants according to Eq. (7). Ploss for resonator no 2 was verified experimentally with temperature increase measurements after placing the resonator inside a phantom with known heat capacity.
Heat generation inside a cylinder
For the three differently shaped prototype examples (Table 3) the heat generation was assumed to be uniform over the whole volume of the resonator inductance. For one resonator (no 2) the power loss Ploss was verified experimentally. Figures 7,8,9 show the results of these simulations. In no simulation the temperature increases reach a value of 5 K. Literature [17-21] on hyperthermia permits an estimation of the temperature resistance of normal animal cells and human cancer cells, depending on exposure time and temperature increase. The value chosen as physiologically critical, 5 K for 10 min to 15 min exposure time, is a worst case assumption, as various cells tolerate higher temperature increases as well as longer exposure times.
Figure 7 Homogeneous power loss inside resonator no 1. a) total cross section through cylinder containing the cylinder axis with an axial view b) radial view of part of the simulation volume; the dotted auxiliary lines denote the dimension of the implant
Figure 8 Homogeneous power loss inside resonator no 2. a) axial view of a total cross section through cylinder containing the cylinder axis (see table below) b) radial view of part of the simulation volume; the dotted auxiliary lines denote the dimension of the implant
Figure 9 Homogeneous power loss inside the entire resonator no 3. a) axial view of a total cross section through cylinder containing the cylinder axis (see table below) b) radial view of part of the simulation volume; the dotted auxiliary lines denote the dimension of the implant
Heat generation in nr metallic rings with rectangular cross section
Two rings are the worst case for resonator no 1 from Table 3 setting up a saddle coil like design. Figure 10 shows the results of the simulation. For a quality factor of 5 the temperature increases inside tissue reach 6.9 K. This is in accordance with Eq. (11). For a value of 4.4 W/m the temperature difference between the wire surface and a point at distance r = 10 cm is 9.4 K for the thermal equilibrium. Setting up a second heat sink simulating blood flow through the implant reduces the maximum temperature difference. Increasing the radius exposed to flow towards the radius of the implant increases the reduction. For an implant with the total inner volume exposed to blood flow the maximum temperature increase in the vessel wall only reaches 3.5 K (Figure 11f).
Figure 11 Power loss for resonator no 1 with two rings and with simulated blood flow of varying cross sections. a) partial radial view without heatsink inside resonator b) partial radial view with flow inside central region of 5 mm radius c) partial radial view with flow inside central region of 10 mm radius d) partial radial view with flow inside central region of 15 mm radius e) partial radial view with flow inside central region of 17 mm radius f) partial radial view with flow inside central region of 17.5 mm radius
For resonator no 2 (Table 3) 6 rings are shown in Figure 12, whereas for resonator no3 the results for 12 windings are shown in Figure 13. The material of the power generating rings has only a negligible influence on the calculated temperature maps except during the steps at the beginning of the simulation (Figures 10, 12). This is an additional proof of a correctly implemented algorithm because of two reasons. Firstly, the thermal conductivity of metal is so much larger than that of tissue, that the algorithm uses only the heat conducting path through tissue at a metal tissue interface. Secondly, the thermal capacity and the density of metal or tissue have no influence on the temperature map in the case of thermal equilibrium, because in that case heat just passes the simulation cells without changing their temperatures anymore. This argument has been mentioned already in the context of Eq. (11), which is valid for thermal equilibrium.
Figure 12 Power loss on six rings for resonator no 2. a) partial axial view for 3 of 6 power generating rings of tissue (see table below) b) partial axial view for 3 of 6 power generating rings of titanium c) partial axial view for 3 of 6 power generating rings of iron d) partial axial view for 3 of 6 power generating rings of tantalum
Figure 13 Power loss on twelve rings for resonator no 3. a) partial axial view for 6 of 12 power generating rings of titanium (see table below) b) partial axial view for 6 of 12 power generating rings of iron
Heat generation in a cylinder shell
The temperature peaks at the power generating rings become negligible for six or more rings after ten minutes simulated time. Therefore it was reasonable to choose a cylinder shell as power generating source. The simulations with a homogenous power dissipating cylinder shell instead of rings are shown from Figure 14 to 15 for resonators 2 and 3 of Table 3. None of these simulations shows a critical temperature increase.
Figure 14 Power loss over a cylinder shell for resonator no 2. a) partial axial view for power generating cylinder shell b) partial radial view for power generating cylinder shell
Figure 15 Power loss over a cylinder shell for resonator no 3. a) partial axial view for power generating cylinder shell b) partial radial view for power generating cylinder shell
Discussion
Test of simulations: endless wire model
The simulation of an endless wire uses the first of the three above mentioned models. Especially, the radius Rimp is taken as equal to rwire. The condition that the wire is endless can be realized in two ways. The first way is to choose a very long Limp and to evaluate the temperature increases only at positions on the center plane (index x = 1) and at radial distances r small compared to Limp. The second way is to modify the simulation software in such a way that the energy transfer in axial direction is excluded. This can be done by considering only volume elements of the center plane (index x = 1) and setting the heat sink shell temperature (index x = 2) after each iteration not to zero, but to the same value as at index x = 1 for all r respectively.
The second way has been realized. The common and expected result was that during the temporal development the simulated radial temperature differences increasingly approach the analytical solution for the thermal equilibrium (Eq. (11)). One comparison between a simulated and the analytical result is shown in Figure 6. Although a test on agreement of both results is independent of the specific choice of the power loss density P* [W/m], we have chosen a realistic value of P* in order to obtain realistic temperature increases. The power loss density P* was calculated with Eq. (7) for the MRI sequence of Table 2 for the example resonators (Table 2) to be between 0.24 W/m and 4.4 W/m. The simulation uses value of 1 W/m.
Test of simulations
A correct simulation of an endless wire should deliver results similar to the analytical curve of Eq. (11) using an implementation of the algorithm with no energy exchange in longitudinal (x) direction. The physical model of such a linear wire claims that the temperature difference between the wire and a cell at a certain distance increases monotonically with that distance. The theoretically calculated curve is valid after reaching the thermal equilibrium. With increasing simulated time the calculated curves show an increasingly close coincidence with the prediction of Eq.(11) (Figure 5), indicating the correctness of the algorithm.
The analytical linear wire model can also indicate which power per unit of wire length P* is safe for active magnetic implants (Figure 4). The theoretical estimation according to Eq.(11) shows for a power density larger than 2 W/m a risk of reaching temperatures increases greater than 5 K. The results do not vary with the time resolution Δt as long as the value is small enough to prevent the simulation from oscillating (Figure 6c,d). Also the spatial resolution (Δr, Δx) does not affect the results very much as long as the temperature maps appears smooth (Figure 6d,e,f). On the other hand, the calculation time explodes with a finer resolution. An increase of the spatial resolution always requires a better temporal resolution. Especially if a metallic cell completely surrounded by tissue is divided into a few metallic cells, a superior temporal resolution is necessary. For a metallic cell completely surrounded by tissue the shortest diffusion length equals roughly half of the spatial resolution with the diffusion path through tissue. For adjacent metallic cells the diffusion length equals the spatial resolution, but the diffusion path is completely through metal with a much larger thermal conductivity. The energy transfer for an identical temperature difference through this path is at least an order of magnitude larger than the one between a metal cell and a tissue cell. This higher energy transfer requires the drastically reduced time step Δt for a stable temporal development of the simulation. The calculation time for an identical simulated time increases by several orders of magnitude with metallic rings and a high spatial resolution and can easily reach computing times of several days on a normal desk top computer. Some examples for the calculation time is given in the second Table below Figure 6.
Power loss of resonator
Achievable quality factors were derived from the construction of experimental solenoid resonator prototypes (Table 3). The quality factor for a given geometry is strongly dependent on the environment and the thickness of the wire insulation. It is reduced substantially after changing the environment from air or distilled water to saline solution. Also Q decreases with a thinner insulation of the wire. Using this information it is possible to conclude that the main power losses are electric losses. Losses due to eddy currents induced by the magnetic field would not change with the thickness of the electric insulation. The dominance of the electric losses is also verified by comparing the power loss from eddy currents inside the active implant (Eq. (3)) with the total power loss (Eq. (7)). For all example prototypes in Table 3, the power loss calculated with the SAR inside the resonator (Eq. (3)) is lower by several orders of magnitude than the total power loss according to Eq. (7) (Table 3).
Realistic values for the quality factor of a resonator placed in ionic surroundings such as tissue are below 5 for all example resonators of Table 3. The calculated examples are worst case assumptions as they assume the maximum achievable Q of 5 for resonator no 1 and therefore the maximum power loss. For resonator no 2 (Q = 12.5) and no 3 (Q = 80) the chosen value is far beyond the reachable inside an ionic surrounding. These high values for resonator 2 and 3 are chosen to pronounce the safety statement for smaller resonators with quality factor values well above achievable values. The unrealistic high quality factors do not show physiologically critical temperature increases. For resonator number 1 (with a large volume) a realistic value was chosen, because in this case the temperature increase reaches physiologically dangerous values.
Part of the simulation results could be confirmed on stents implanted inside the aorta of rabbits [9,10]. Small coated resonators with a volume below 0.5 cm3 and quality factors of 3 to 4 were used. After excision and histopathologic examination, the tissue did not show any indications of heating after several MRI investigations. This is in coincidence with the simulation results, because even for small resonators with a much higher quality factor no dangerous heating was calculated.
Ten minutes of simulated time is 5 minutes less than the critical time according to the FDA regulations for imaging of the trunk assuming a SAR of 4 W/kg. But all simulations show only small changes after 10 minutes. Therefore it is unnecessary to increase the simulation time further. Nevertheless, one example of 15 minutes simulated time is given in the movie (see additional file 1) and shows only small changes in the last five minutes.
The weak influence on the distance between power generating cells and heat sink (the outer most layer) is also shown in the movie (see additional file 1). The difference between two almost identical calculations is shown. Starting from a first calculation with simulation parameters Lsim and Rsim, the second calculation uses doubled values of Lsim and Rsim without changing the spatial resolution or the other simulation parameters. This approach shifts the heat sink to a larger distance from the heat generating cell elements. As expected the heat sink layer drags down the temperature increase in its vicinity, but the effect is small if the volume is adequately chosen. The differences for identical cells between both simulations are shown by alternating both views a few times at the end of the movie (see additional file 1).
Finite volume analysis
For all three example resonators (Table 3) the first simulation model assumes a uniform heat generation over the whole volume of the inductor. Figures 7,8,9 show the results of the simulations. All temperature increases are below the critical value of 5 K. The maximum temperature is at the cylinder axis, corresponding to the center of the vessel for a vessel implant. The uniform heat generation inside homogeneous tissue is – as discussed above – not the best approximation. The ring model is a more appropriate match for realistic conditions.
Therefore resonator 1 is assumed as a solenoid with 2 turns (Figure 10). The simulation assumes these turns as metallic rings, which should be similar to a solenoid for the question of temperature increases. For resonator no 2, six rings are calculated (Figure 12) and resonator 3 is simulated with twelve rings (Figure 13). Only the simulation for resonator no 1 shows temperature increases over 5 K, which are located on and directly adjacent to the power generating rings. For resonator 2 the temperature increase with 6 rings is slightly above 1 K (Figure 12). The same is true for resonator no 3 with twelve rings (Figure 13).
For 6 or 12 rings and after 600 s calculated simulation time, the peak values near the location of the power generating rings inside the tissue become more and more negligible. Therefore a model using a power generating cylinder shell is appropriate for resonators with a reasonable number of turns.
For resonators no 2 and no 3, the cylinder shell simulation did not show any unsafe heating and is similar to the ring calculations. Neglecting the peaks of the ring simulation the maximum temperature increase is nearly identical.
Only the simulation for resonator no 1 with two rings reaches a critical temperature value above 5 K, all others stay below this value. For a large number of rings or, stated more precisely, a high wire density at the cylinder surface, it is possible to use the cylinder shell model (compare Figures 12, 14 and Figures 13, 15). As expected, the rings cause higher maximum temperature increases according to the higher local power density. The difference between the shell generating the entire power (Figures 14, 15) and a uniform power generating volume (Figures 8, 9) is mainly a different temperature distribution inside the inductor. The former simulation first shows a moderate increase in radial direction up to the power generating shell inside the volume, followed by a decrease. The latter has a larger peak value at the axis of the cylinder and falls off continuously with greater distance from the axis. The ring model is the most preferable because of the dominance of the electric losses on and near the wire of the inductor. Using different metals without altering the other simulation parameters changes the results only marginally (Figures 10b,c; 11 b-d; 12 a, b).
The simulation worked on a "worst case" basis, neglecting in most cases all cooling effects except the energy transport due to the thermal conductivity. Especially for vascular implants in the "normal case", some blood flow and blood perfusion, causing a much faster energy transport, will reduce the temperature increases. The effect of blood flow is simulated for the example of resonator 1 with the two rings. This was the only simulation reaching physiologically critical temperature increases. Blood flow is simulated by keeping the temperature difference at zero for a central part of the cylindrical implant. Changing the dimension of the central part can simulate an implant, which may be more or less infiltrated to the vessel wall or covered by tissue (thrombosis, calcification or intimal hyperplasia). The reduced temperature increase for resonator no 1 and for various sizes of the cooling blood flow is shown in Figure 11. For an implant placed directly inside the wall, with none or only a small amount of tissue between wire and blood flow the temperature increases drop to a physiologically tolerable value (Figure 11). Nevertheless for implants with a tissue coverage of a few mm, which can be plaque or a thrombosis, critical temperatures can be reached.
With respect to the overall power absorption greater than 100 Watts inside the human body during MRI investigations with a maximum SAR, one additional Watt inside the entire body is negligible.
Conclusion
This investigation assumes a "worst-case scenario" in different ways. Firstly, the resonator is assumed being perfectly aligned within the plane of the excitation field B1. Secondly, to some extent no blood flow inside the inductor (vessel) of the implant is included. Thirdly, no blood perfusion inside the tissue around the resonator is taken into account. Lastly, too large quality factors are used.
For most "normal" cases, intact active implants will therefore be less critical, but it is not possible to exclude the worst case conditions. Pathologies (thrombosis, intimal hyperplasia, plaque) may alter blood flow and perfusion in the area directly adjacent to the current paths of the resonator and certainly the resonator can be perfectly aligned to the plane of the exciting rf-field B1.
Especially for resonators with a large volume (such as resonator no 1) and with a small number of rings it is possible to reach critical temperature increases above 5 K (Figures 10, 11). For peripheral and especially cardiac vessels, even with an unrealistically high quality factor, the power loss is too low for dangerous heating. On the other hand, stent grafts or vena cava filters built as active implants can reach volumes of a few ten cm3, which may be dangerous, if the tissue around the wires of the implant is not exposed to blood flow or sufficient blood perfusion. To reduce the risk for such active implants, the quality factor has to be low. This reduces also the amplification of the MR signal the resonator is made for and can thus be dispensed with altogether.
As the above simulations assume properly working resonance circuits without any failures on the electric paths of the system, one worst case scenario was not presented. Defects such as ruptures or partial ruptures may generate a relatively high resistance over very short distances. The current flow through this resistance can produce a large power loss inside an extremely small volume. This can generate a very high power density which, even for small implants, may induce physiologically critical temperature increases for a small volume. In fact the analysis of these "hot spots" is important, because ruptures of stent struts are likely and a high power density can occur also for smaller implants. An additional investigation estimating the maximum possible power loss inside such "hot spots" and the resulting temperature maps around them is necessary to check the safety of active implants under these circumstances. This is being prepared for future publication.
The study protocols for the cited animal experiments were approved by the responsible authority (Landesamt für Arbeitsschutz, Gesundheitsschutz und technische Sicherheit, Berlin, G 0142/99).
Authors' contributions
MB has drafted the investigation, written the manuscript, coded the simulation and contributed to the theory and part of the cited experimental work.
WV has drafted the theoretical part of the manuscript and reviewed the manuscript. Also part of the cited experimental tests for power loss, quality factor measurements and prototype constructions are from WV.
JS was responsible for the cited animal experiments and provided the in vivo images of prototypes in an animal model. He has also contributed to the medical background and has reviewed the manuscript.
DG has enabled and reviewed the manuscript, contributed to the medical background, and was helpful during the investigation with discussions.
Supplementary Material
Additional File 1
Movie with an example of a time developing temperature map This movie (animated GIF) shows the time development over a period of 900 s, which is the maximum permitted time for the imaging of the trunk with an SAR of 4 W/kg (manufacturer declaration, sequence of Table 2). At the end of the simulated time, the changes in the temperature distribution become small. For complete information, a 3D axial view as well as a 3D radial view is shown simultaneously. At the end of the movie two different simulations are shown alternately, which indicate the changes of the temperature map after 900 s due to a larger simulation volume shifting the heat sink more away from cells with power loss. One of the alternating results was calculated using a 250 × 250 matrix for a distance of 0 mm to 25 mm for r and x respectively. The second map was calculated for a 500 × 500 matrix for a distance of 50 mm for r and x respectively. Only the inner 250 × 250 points are plotted for a comparable size for both calculations. It can be seen that the temperature distribution is almost identical apart from the fact that, for x≈25 mm and r≈25 mm, the simulation with more cells shows a slight deviation from zero. The simulation with the smaller matrix shows a straight zero line, which is naturally because this is the boundary condition for this simulation. The small difference points out that the boundary condition with a heat sink works very well as long as the absolute value of the gradient at the boundary is low.
Click here for file
Acknowledgements
Many thanks to Peter van Leeuwen, a bilingual native English speaker who has reviewed and corrected the entire "German English" manuscript.
==== Refs
Bartels LW Smits HF Bakker CJ Viergever MA MR imaging of vascular stents: effects of susceptibility, flow, and radiofrequency eddy currents J Vasc Interv Radiol 2001 12 365 371 11287516
Hug J Nagel E Bornstedt A Schnackenburg B Oswald H Fleck E Coronary arterial stents: safety and artifacts during MR imaging Radiology 2000 216 781 787 10966711
Maintz D Kugel H Schellhammer F Landwehr P In vitro evaluation of intravascular stent artifacts in three-dimensional MR angiography Invest Radiol 2001 36 218 224 11283419 10.1097/00004424-200104000-00004
Meyer JM Buecker A Schuermann K Ruebben A Guenther RW MR evaluation of stent patency: in vitro test of 22 metallic stents and the possibility of determining their patency by MR angiography Invest Radiol 2000 35 739 746 11204801 10.1097/00004424-200012000-00008
Schnall MD Barlow C Hirihara Subrmaniam V Wireless implanted magnetic resonance probes for in vivo NMR J Magn Reson 1986 68 161 167
Farmer TH Cofer GP Johnson GA Maximizing contrast to noise with inductively coupled implanted coils Invest Radiol 1990 25 552 8 2345087
Patent of Simag GmbH, Teltowkanalstr.2, D-12247 Berlin, Germany, EP1092985A3
Quick HH Kuehl H Kaiser G Bosk S Debatin JF Ladd ME Inductively coupled stent antennas in MRI Magn Reson Med 2002 48 781 90 12417992 10.1002/mrm.10269
Kivelitz D Wagner S Schnorr J Wetzler R Busch M Melzer A Taupitz M Hamm B A Vascular Stent as an Active Component for Locally Enhanced Magnetic Resonance Imaging: Initial In Vivo Imaging Results After Catheter-guided Placement in Rabbits Invest Radiol 2003 38 147 152 12595794 10.1097/00004424-200303000-00002
Kivelitz D Wagner S Hansel J Schnorr J Wetzler R Busch M Melzer A Taupitz M Hamm B The active magnetic resonance imaging stent (AMRIS): initial experimental in vivo results with locally amplified MR angiography and flow measurements Invest Radiol 2001 36 625 631 11606839 10.1097/00004424-200111000-00001
Athey TW Ross RJ Ruggera PS Magnetic Fields Associated with Nuclear Magnetic Resonance Medical Imaging System Mag Res Imaging 1982 1 149 154 10.1016/0730-725X(82)90206-5
Margulis AR Higgins CB Kaufman L Crooks LE Clinical Magnetic Resonance Imaging Radiology Research and Education Foundation 1983 Library of Congress Catalog Card number 83-81823
Johnson CC Guy AW Nonionizing Electromagnetic Wave Effects in Biological Materials and Systems Proc of the IEEE 1972 60
Yeung CJ Susil RC Atalar R RF Safety of Wires in Interventional MRI: Using a Safety Index Magnetic Resonance in Medicine 2002 47 187 193 11754458 10.1002/mrm.10037
Hand JW Lau RW Lagendijk JJW Ling J Burl M Young IR Electromagnetic and Thermal Modeling of SAR and Temperature Fields in Tissue Due to an RF Decoupling Coil Magnetic Resonance in Medicine 1999 42 183 192 10398965 10.1002/(SICI)1522-2594(199907)42:1<183::AID-MRM24>3.0.CO;2-I
Liao SamuelY Microwave Devices and Circuits 1980 Prentice-Hall, Englewood Cliffs, New Jersey 134
Sminia P Haveman J Wondergem J van Dijk JD Lebesque JV Effects of 434 MHz microwave hyperthermia applied to the rat in the region of the cervical spinal cord Int J Hyperthermia 1987 3 441 52 3681044
Haveman J Van Der Zee J Wondergem J Hoogeveen JF Hulshof MC Effects of hyperthermia on the peripheral nervous system: a review Int J Hyperthermia 2004 20 371 91 15204519 10.1080/02656730310001637631
Khoei S Goliaei B Neshasteh-Riz A Deizadji A The role of heat shock protein 70 in the thermoresistance of prostate cancer cell line spheroids FEBS Lett 2004 561 144 8 15013766 10.1016/S0014-5793(04)00158-9
Poland J Schadendorf D Lage H Schnolzer M Celis JE Sinha P Study of therapy resistance in cancer cells with functional proteome analysis Clin Chem Lab Med 2002 40 221 34 12005211 10.1515/CCLM.2002.037
Sinha P Poland J Schnolzer M Celis JE Lage H Characterization of the differential protein expression associated with thermoresistance in human gastric carcinoma cell lines Electrophoresis 2001 22 2990 3000 11565793 10.1002/1522-2683(200108)22:14<2990::AID-ELPS2990>3.0.CO;2-F
Gowrishankar TR Stewart DA Martin GT Weaver JC Transport lattice models of heat transport in skin with spatially heterogeneous, temperature-dependent perfusion BioMedical Engineering OnLine 2004 3 42 15548324 10.1186/1475-925X-3-42
Geddes LA Baker LE The specific resistance of biological material – a compendium of data for the biomedical engineer and physiologist Med & Biol Eng 1967 5 271 293 6068939
Holmes KennethR
| 0 | PMC1087858 | CC BY | 2021-01-04 16:37:36 | no | Biomed Eng Online. 2005 Apr 12; 4:26 | latin-1 | Biomed Eng Online | 2,005 | 10.1186/1475-925X-4-26 | oa_comm |
==== Front
Cancer Cell IntCancer Cell International1475-2867BioMed Central London 1475-2867-5-71580198610.1186/1475-2867-5-7Primary ResearchDownregulation of calcineurin activity in cervical carcinoma Padma S [email protected] A Pavani [email protected] Usha Rani [email protected] Meenakshi [email protected] BN [email protected] Gayatri [email protected] Centre for DNA Fingerprinting and Diagnostics, Nacharam, Hyderabad, A.P, India2 M.N.J Institute of Oncology and Regional cancer centre, Hyderabad, A.P, India3 Mediciti Hospital, Medchal, A.P, India2005 1 4 2005 5 7 7 3 2 2005 1 4 2005 Copyright © 2005 Padma et al; licensee BioMed Central Ltd.2005Padma 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
Calcineurin (CaN) is an important serine-threonine phosphatase (PP2B), which plays a crucial role in calcium-calmodulin mediated signal transduction events. Calcineurin has been implicated in pathogenesis of various diseases cardiac hypertrophy, diabetic neuropathy and Alzheimer's, however its role in neoplasia remains unclear.
Results
In view of this we evaluated the calcineurin activity in serum and biopsy samples collected from women diagnosed with invasive squamous cell carcinoma of cervix. A significant reduction was observed in the calcineurin activity in cancer cervix patients compared to the control group. However the calcineurin activity remained unaltered in the cervical scrapes obtained from patients diagnosed with low-grade squamous intra epithelial lesions (LSIL). Interestingly the downregulation of calcineurin activity in squamous cell carcinomas was not accompanied by any significant change in DNA-binding affinity of the transcriptional factor NFAT (Nuclear Factor of Activated T-cells). All the squamous cell carcinoma samples used in the present study were positive for high-risk human papillomavirus (HPV) types.
Conclusion
The present study demonstrates the downregulation of calcineurin activity in squamous cell carcinoma of cervix with high risk HPV infection. We conclude that perturbations in calcineurin-mediated pathway may be involved in development of cervical neoplasia.
cervical neoplasiasquamous cell carcinoma (SCC)calcineurin (PP2B/CaN)calmodulin (CaM)
==== Body
Introduction
Cervical cancer is one of the most prevalent neoplastic diseases affecting women world wide, especially in the developing countries [1,2]. In India, cervical neoplasia is not only the most common malignancy but also a major cause of death among middle-aged women particularly in the rural areas [3,4]. It is well established now that infection with human papilloma virus (HPV) is a major etiological risk factor for development of cervical cancer, however other host/environmental factors contribute to the progression of the disease [5-7]. Infection with HPV can modulate the host factors involved in signaling cascades and deregulation of signaling pathways is a hallmark of cancer. An important feature of the signal transduction pathway is reversible serine/threonine phosphorylation of proteins and any aberrant activity in the kinases or phosphatases can disturb the phosphorylation/dephosphorylation cycle leading to uncontrolled proliferation. Therefore, identification of key-signaling molecules (kinases and phosphatases) responsible for neoplastic development can have potential clinical relevance allowing for diagnosis, early detection and intervention.
There is a plethora of information existing on the involvement of various kinases such MAPK [8], PI3K [9], JNK [10], cdk4 [11,12], cdc2 [13], Her2 [14], p38 [15] etc. in cervical carcinoma cells. While the phosphatases such as cdc25A/C and PTEN have been shown to be associated with cervical cancer [12,16,17], the existing information falls short on the involvement of serine/threonine protein phosphatase family members viz., PP1, PP2A, PP2B (calcineurin) and PP2C. There is only one report, which demonstrates downregulation of protein phosphatase 1, regulatory subunit in cervical cancer cells [18]. In view of the limited information available on involvement of serine/threonine phosphatase, an attempt was made in the present study to investigate if there is any association between calcineurin and cervical neoplasia.
Calcineurin is the only serine/threonine phosphatase whose activity is modulated by both calcium and calmodulin and it is a crucial mediator of the calcium mediated signal transduction pathways. Calcineurin is a heterodimeric protein consisting of a catalytic subunit (calcineurin A), with an active metal binding centre, and a regulatory subunit (calcineurin B), which binds to calcium [19]. One of the major functions of calcineurin is T-cell activation by dephosphorylation of the transcriptional factor NFAT (Nuclear Factor of Activated T-cells) [20]. Calcineurin is involved in a wide variety of biological responses including neuronal and muscle development, neurite outgrowth, morphogenesis of vertebrate heart valves, memory development, behavioral response etc [21-26]. In addition, calcineurin has also been implicated in pathogenesis of diseases such as cardiac hypertrophy, Alzheimer's, schizophrenia, diabetic nephropathy and Down's syndrome [27-32]. Calcineurin plays a role in apoptosis via inducing the cytochrome c/caspase dependent pathway and by dephosphorylating Bad, a proapototic member of Bcl-2 family [33,34]. Recent studies have also demonstrated the involvement of calcineurin in the cell cycle by regulation of the cdk4 (cyclin dependent kinase 4), a G0/G1 checkpoint element [35]. The immunosuppressant cyclosporin A (CsA), which specifically inhibits calcineurin activity, has proved to be a useful tool in understanding the role of calcineurin in various cellular processes. CsA has been shown to exert an antiproliferative effect not only in T cell but also in a wide variety of cells including lymphomas, keratinocytes, fibroblasts and smooth muscle cells indicating that calcineurin activity is important for cell cycle progression [36]. In contrast to the antiproliferative action of CsA in cell culture system, it has been shown that treatment with CsA in kidney transplant recipients is associated with increased incidence of renal cancer [37,38]. However it is still unclear if the genesis of cancer in CsA administered persons is due to direct inhibition of calcineurin-mediated pathways or its other indirect effects. This warrants a study on involvement of calcineurin in cancer progression and therefore the present study was undertaken to ascertain the importance of calcineurin in cervical neoplasia.
Materials and methods
Specimen collection
For the present study patient's consent was taken and the study was approved by the institutional bioethical committee. Cervical biopsy samples were collected (N = 45) from women diagnosed with invasive cervical cancer. The biopsy samples collected from patients were immediately snap frozen in liquid nitrogen. Cervical tissue specimens (N = 30) collected from women undergoing hysterectomy for non-neoplastic conditions formed the control group. Blood samples (3 ml) were collected in plain bottles [without anticoagulant] from both the normal and cancer patients. Blood was allowed to clot and serum was separated by centrifugation at 3000 rpm for 10 min at room temperature. The serum samples were stored at -20°C till further use.
Cervical scrapes were also collected using Ayer's spatula from women attending the cervical screening outpatient clinic (N = 30). Four of these women were found to have low-grade intraepithelial lesion (LSIL) and the rest with chronic inflammation only on cytology. The cervical scrape was collected in chilled phosphate buffer saline (pH 7.2) and the cells were collected by centrifugation at 13000 rpm for 10 min at 4°C. The supernatant was discarded and the cells were immediately frozen in liquid nitrogen till further use.
Sample preparation
Tissues (100 mg) and cervical scrapes were homogenized in 500 μl or 250 μl of buffer A respectively, containing 25 mM HEPES (pH 7.2), 150 mM NaCl, 1% NP-40, 10 mM MgCl2, 10% glycerol, 10 μg/ml PMSF and 10 μg/ml aprotinin for 3 min. The homogenates were cleared of the debris by centrifugation at 13,000 rpm for 10 min at 4°C. The supernatants were aliquoted and stored at -70°C till further use.
Assay for calcineurin
An aliquot of the homogenized sample (100 μl) was passed through sephadex G-25 column (Amersham Pharmacia) to eliminate the free phosphate and further used for calcineurin activity. The protein estimation was done using the BCA Kit [Pierce company] as per the manufacturers protocol. Calcineurin activity was assayed in a total volume of 50 μl containing 25 μl of 2X assay buffer (200 mM NaCl, 100 mM Tris [pH 7.5], 12 mM MgCl2, 1 mM CaCl2) and 5 μl of the cell/tissue homogenate. The samples were incubated with 2X-assay buffer for 10 min at 30°C. The total volume was adjusted to 50 μl by the addition of water. Reactions were initiated by adding RII phosphopeptide [5 μM] and incubated for a further period of 10 min at 30°C. Reactions were terminated by addition of 100 μl of Malachite green reagent (3 vol. of 0.045% Malachite green and 1 vol. of 4.2% ammonium molybdate in 4N HCl). The color was allowed to develop for 30 min and the absorbance read at 660 nm. The calcineurin activity was calculated as nmoles of inorganic phosphate released/min/mg total protein in case of tissues and for serum samples as nmoles of inorganic phosphate released/min/dl of serum.
Determination of calmodulin and calcineurin contents in tissue lysates
The calmodulin and calcineurin contents were determined by competitive ELISA methods as described earlier [39]. The calcineurin antibody against the catalytic subunit A was used in this study.
Electrophoretic mobility shift assay for NFATc-DNA binding
Electrophoretic mobility shift assay (EMSA) was performed using oligonucleotide with consensus binding site for NFATc (5'CGCCCAAAGAAAATTTGTTTCATA3'). Briefly, biopsy tissues were homogenized and the nuclear pellet was collected by centrifugation (600 × g: 10 min, 4°C). The nuclear lysates (100 μg) were prepared and incubated in 25 μl of reaction mixture containing polydI/dC, 10X binding buffer (50 μM ZnCl 2,0.25 mM DTT, 20 mM Tris [pH 7.5], 60 mM KCl, 1 mM MgCl2 0.1 mM EDTA, 10% glycerol) and the labeled oligo for 30 min at room temperature. The reactions were terminated using 25 μl of the loading dye and loaded onto a 7% gel retardation assay gel. The run was done at 250 V for 3 hrs at 4°C. At the end of the run, the gel was removed, dried and exposed in a phosphoimager overnight at room temperature.
HPV-testing and Typing
The DNA isolated from the squamous cell carcinoma samples were tested for presence of HPV by a PCR based reverse line blot assay as described earlier (40)
Statistical analysis
Comparisons between the normal and the cancer samples were done using the Student's t-test for paired data.
Results
1. Calcineurin activity in serum and cervical tissue (normal vs squamous cell carcinoma)
Calcineurin activity was assayed in both sera and tissue lysate by measuring the dephosphorylation of RII peptide, a specific substrate of calcineurin. The biopsy samples collected from the cancer patients were classified as moderately differentiated large cell non-keratinising squamous cell carcinomas of the cervix. We observed a significant reduction in the calcineurin activity both in the cervical tissues (p = 0.0001) and sera (p = 0.0037) of patients with invasive cervical cancer in comparison to the control group (Fig. 1). It has to be noted that we have calculated the calcineurin activity in serum as nmoles of phosphate released/min/deciliter. However if we consider the total protein content per deciliter of serum for each sample and accordingly calculate the calcineurin activity per milligram of serum protein, the calcineurin activity in serum will be much lower than that of tissue. Since the international standard for representing the units for enzyme activity for serum/plasma is represented in units per deciliter/litre the same has been represented in Fig. 1a.
Figure 1 Calcineurin activity in serum (a) and biopsy samples (b) of control and cervical cancer patients diagnosed with squamous cell carcinoma (SCC). The calcineurin activity is significantly reduced in serum samples [1a, p = 0.0001] and in tissue lysates [1b, p = 0.0037] obtained from cancer patients compared to the normal group. The calcineurin activity was calculated as nmoles of inorganic phosphate released per mg protein of tissues or per deciliter of serum.
2. Calmodulin and calcineurin A content in cervical tissues (normal vs squamous cell carcinoma)
The calmodulin and calcineurin contents were also analyzed in the biopsies, to evaluate if the decrease in the calcineurin activity could be correlated with their respective contents. The content for both calcineurin A subunit and calmodulin was measured using competitive ELISA method in the tissue samples. There was no change in the level of calcineurin A between the control and the cancer samples (Fig. 2). Interestingly, there was a significant increase in the calmodulin level in cervical carcinoma (Fig. 3).
Figure 2 Total calcineurin content in normal and cervical cancer tissue lysates [SCC]. The calcineurin content was assayed by using antibody specific to calcineurin [Sigma monoclonal anti-CaNα, 1:7500 dilution using a sandwich ELISA method.
Figure 3 Total calmodulin content in normal and cervical cancer tissue lysates (SCC) were determined by using antibody specific to calmodulin. Note a significant increase [p = 0.0263] in calmodulin content in cancer patients.
3. Calcineurin activity in Low-grade squamous cell intraepithelial lesions
To correlate the decrease in calcineurin activity with the early dysplastic changes, we collected cervical scrapes from thirty women attending the outpatient department in the hospital. Of these four were diagnosed as Low-grade squamous intraepithelial lesions (LSIL) by cytology and the rest were cytologically normal but with chronic inflammation. We didn't observe any significant change in the calcineurin activity between the LSIL and the controls in the cervical scrapes (Fig. 4).
Figure 4 Calcineurin activity as determined in the cervical scrapes obtained from control (N = 25) and those diagnosed with low-grade intraepithelial squamous lesion (LSIL) or CIN1. Note, no significant change in calcineurin activity in control and LSIL group.
4. NFAT activity
In view of our observation on downregulation of calcineurin activity in cervical carcinoma, we wanted to investigate if there is any alteration in the DNA-binding activity of NFAT, one of the major substrates of calcineurin. We analyzed the nuclear extracts from 5 squamous cell carcinoma and 5 controls by incubating it with the radiolabelled oligonucleotide probes specific for NFAT binding. Interestingly we didn't find any significant change in the DNA binding affinity of NFAT between cancer and the control samples (Fig. 5).
Figure 5 NFAT-DNA-binding activity in normal and cervical cancer nuclear extracts. Briefly 100 μg of nuclear extracts was incubated with 32P-labelled NFATc oligonucleotide and Electrophoretic mobility shift assay was performed as described in Material and methods. Lane one contains free probe without nuclear extract and lane 2 contains 100-fold excess of unlabelleled NFAT oligonucleotide as a specific competitor. Specific NFAT- DNA complexes are indicated by arrow.
5. HPV typing and testing
A PCR-based reverse line blot assay (40) was used to check the presence of HPV types in the squamous cell carcinomas. We could detect oncogenic HPV-types (16,18,31,45 etc) in all of the cervical cancer but none in the control samples, however we could not correlate the presence of HPV-types with any variation in calcineurin activity (data not shown).
Discussion
The present study shows downregulation of calcineurin activity in invasive cervical cancer. High calcineurin activity can lead to apoptosis [33], thereby indicating that the downregulation in calcineurin activity can help in cellular survival thereby promoting neoplastic progression. In a previous study [41] a decrease in calcineurin activity in MCF-7 cells following retinoic acid treatment has been linked to downregulation in the expression level of catalytic subunit (A) of calcineurin, however in our study we didn't find any alterations in the level of calcineurin A.
The basis for the decrease in calcineurin activity observed in cervical neoplasia in absence of a change in Calcineurin A content needs to be identified and we speculate that a change in redox state of the cancer cell due to oxidative stress contributes to the alteration in the activity of calcineurin without altering its content. Calcineurin by virtue of its Fe2+-Zn2+ binuclear center in its active site is oxidized to Fe3+-Zn2+ by superoxide and hydrogen peroxide ions [42]. Superoxide dismutase (SOD), a free radical scavenging enzyme, protects calcineurin from inactivation by superoxide and hydrogen peroxide ions [43,44]. Low levels of the antioxidants such as glutathione, Vitamin E and C as well as superoxide dismutase have been reported to be lower in cervical cancer [45,46], which in turn can modulate the calcineurin activity. Interestingly the lower SOD in circulation reported in cervical cancer [46] may have a role in downregulation of the serum calcineurin activity. There is clearly strong evidence on the involvement of reactive oxygen species (ROS) in cervical cancer [47] and it can most likely alter the calcineurin activity atleast in the tissues without affecting its expression. This argument is also supported by an observation that activity of creatine kinase B, a ROS sensitive protein, is downregulated in cervical cancer [48]. Another possible mechanism of inhibition in calcineurin activity is through upregulation of naturally occurring endogenous inhibitors of calcineurin such as calcipressin 1/DSCR1 and forskolin binding protein [31,49]. The role of these negative regulators of calcineurin in cervical cancer is not known. However it is interesting to note that oxidative stress modulates the calcipressin expression and phosphorylation status [50], which in turn can affect the calcineurin activity.
The natural history of invasive cervical cancer involves sequential progression of persistent HPV infection to reversible LSIL (CIN1) and then to high-grade intraepithelial lesions (HGIL/CIN2). In the presence of other precipitating factors HSIL is a direct precursor of invasive cancer [51]. To evaluate if calcineurin activity is also altered in early precancerous lesions we examined the calcineurin activity in cervical scrapes collected from LSIL cases. Interestingly, there was no significant alteration in the calcineurin activity in the early precursor lesions (LSIL) compared to the control group. None of the samples collected in this study were of HSIL. It has to be however noted that unlike the biopsy specimens of squamous cell carcinoma consisting mostly of the clonally derived malignant cells, the cervical scrapes collected is a mixed population of both the normal and the dysplastic cells. It is possible that the overwhelming number of normal cells in the cervical scrapes diagnosed as LSIL resulted in no significant difference in the calcineurin activity between the control and LSIL group. On the other hand more samples of LSIL and HSIL precursor lesions have to be analyzed before we rule out the possibility of calcineurin involvement in the early stages of cervical cancer progression. Nonetheless, the significant decline in calcineurin activity in invasive squamous cell carcinoma is suggestive of its role in cervical neoplasia.
The regulation of gene expression in response to calcium stimuli is one of the most explored functions of calcineurin. Importantly, the critical target of calcineurin is the NFAT family of transcription factors. Calcineurin dephosphorylates NFAT allowing translocation of this protein into the nucleus [19]. Studies conducted on preadipocyte cell line (3T3-L1) have demonstrated that constitutive expression of NFATc1 can induce cellular transformation providing a direct evidence for the oncogenic potential of the NFATc1 transcription factor [52]. Further, NFAT has also been linked to tumor metastasis [53]. In view of the above evidence, it was of interest to investigate the activation of NFAT in cervical cancer. There was no major difference in DNA binding affinity of NFAT from nuclear extracts obtained from cancer and the control samples. The results on the unaltered DNA binding activity of NFAT seem at odds in view of the reduced calcineurin activity in the cervical neoplasia. However, it is possible that regulation of NFAT by calcineurin is cell type specific and is not affected by calcineurin in cervical cells. Another possibility could be, that calcineurin is known to exist in different isoforms (α,β, ↓) and it is probable that the isoforms existing in cervix has a different substrate specificity. Besides NFAT, calcineurin can also modulate the phosphorylation status of other key regulatory proteins such as Bad, IkB and NFkB, which in turn can promote neoplasia by disturbing the balance existing between cell death and proliferation [30,54]. Already the NFkB mediated pathway is known to be upregulated in cervical cancer progression [55]. Decrease in calcineurin activity could lead to cellular survival by deregulating these major players thereby maintaining the malignant phenotype of cervical neoplasia.
Conclusion
In conclusion, our pilot study is indicative of a possible role of calcineurin-mediated pathway in pathogenesis of cervical neoplasia. While NFAT is a major target of calcineurin, our study rules out the involvement of this transcriptional factor indicating presence of a NFAT independent calcineurin pathway atleast in cervical cancer. The present investigation has opened up new avenues to evaluate further the role of calcineurin in cervical cancer progression.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
SP carried out the calcineurin assays and data compilation. HPV testing and typing in cancer cervix samples was carried out by APS. Both SP and APS performed NFAT assay. The samples were provided by URP, MJ and BNR. The study was conceived and coordinated by GR. The manuscript was finalized by GR and approved by all the investigators.
Acknowledgements
We thank all the women who participated in this study. SP was supported by a fellowship from the Department of Biotechnology. We thank Patti Gravitt for helping us with PCR line blot analysis. The PCR line blot reagents were a kind gift from Janet Kornegay, Roche diagnostics, USA. We thank Sangita Mukhopadhyay with helpful suggestions on EMSA.
==== Refs
Franco EL Schlecht NF Saslow D The epidemiology of cervical cancer Cancer J 2003 9 348 59 14690309
Bosch FX de Sanjose S Human papillomavirus and cervical cancer-burden and assessment of causality J Natl Cancer Inst Monogr 2003 31 3 13 12807939
Shanta V Krishnamurthi S Gajalakshmi CK Swaminathan R Ravichandran K Epidemiology of cancer of the cervix: global and national perspective J Indian Med Assoc 2000 98 49 52 11016150
Pal SK Mittal B Improving cancer care in India: prospects and challenges Asian Pac J Cancer Prev 2004 5 226 8 15244530
Castellsague X Bosch FX Munoz N Environmental co-factors in HPV carcinogenesis Virus Res 2002 89 191 9 12445659 10.1016/S0168-1702(02)00188-0
Wang SS Hildesheim A Viral and host factors in human papillomavirus persistence and progression J Natl Cancer Inst Monogr 2003 31 35 40 12807943
zur Hausen H Papillomaviruses in human cancers Proc Assoc Am Physicians 1999 111 581 7 10591087 10.1046/j.1525-1381.1999.99723.x
Chakrabarti O Veeraraghavalu K Tergaonkar V Liu Y Androphy EJ Stanley MA Krishna S Human papillomavirus type 16 E6 amino acid 83 variants enhance E6-mediated MAPK signaling and differentially regulate tumorigenesis by notch signaling and oncogenic Ras J Virol 2004 78 5934 45 15140991 10.1128/JVI.78.11.5934-5945.2004
Ma YY Wei SJ Lin YC Lung JC Chang TC Whang-Peng J Liu JM Yang DM Yang WK Shen CY PIK3CA as an oncogene in cervical cancer Oncogene 2000 19 2739 44 10851074 10.1038/sj.onc.1203597
Liu B Fang M Lu Y Lu Y Mills GB Fan Z Involvement of JNK-mediated pathway in EGF-mediated protection against paclitaxel-induced apoptosis in SiHa human cervical cancer cells Br J Cancer 2001 85 303 11 11461094 10.1054/bjoc.2001.1910
Cheung TH Yu MM Lo KW Yim SF Chung TK Wong YF Alteration of cyclin D1 and CDK4 gene in carcinoma of uterine cervix Cancer Lett 2001 166 199 206 11311493 10.1016/S0304-3835(01)00457-8
Skomedal H Kristensen GB Lie AK Holm R Aberrant expression of the cell cycle associated proteins TP53, MDM2, p21, p27, cdk4, cyclin D1, RB, and EGFR in cervical carcinomas Gynecol Oncol 1999 73 223 8 10329038 10.1006/gyno.1999.5346
Hashiguchi Y Tsuda H Nishimura S Inoue T Kawamura N Yamamoto K Relationship between HPV typing and the status of G2 cell cycle regulators in cervical neoplasia Oncol Rep 2004 12 587 91 15289842
Chavez-Blanco A Perez-Sanchez V Gonzalez-Fierro A Vela-Chavez T Candelaria M Cetina L Vidal S Duenas-Gonzalez A HER2 expression in cervical cancer as a potential therapeutic target BMC Cancer 2004 4 59 15341668 10.1186/1471-2407-4-59
Shen MR Chou CY Hsu KF Ellory JC Osmotic shrinkage of human cervical cancer cells induces an extracellular Cl-dependent nonselective cation channel, which requires p38 MAPK J Biol Chem 2002 277 45776 84 12226098 10.1074/jbc.M207993200
Naeger LK Goodwin EC Hwang ES DeFilippis RA Zhang H DiMaio D Bovine papillomavirus E2 protein activates a complex growth-inhibitory program in p53-negative HT-3 cervical carcinoma cells that includes repression of cyclin A and cdc25A phosphatase genes and accumulation of hypophosphorylated retinoblastoma protein Cell Growth Differ 1999 10 413 22 10392903
Cheung TH Lo KW Yim SF Chan LK Heung MS Chan CS Cheung AY Chung TK Wong YF Epigenetic and genetic alternation of PTEN in cervical neoplasm Gynecol Oncol 2004 93 621 7 15196854 10.1016/j.ygyno.2004.03.013
Narayan G Pulido HA Koul S Lu XY Harris CP Yeh YA Vargas H Posso H Terry MB Gissmann L Schneider A Mansukhani M Rao PH Murty VV Genetic analysis identifies putative tumor suppressor sites at 2q35-q36.1 and 2q36.3-q37.1 involved in cervical cancer progression Oncogene 2003 22 3489 99 12776201 10.1038/sj.onc.1206432
Rusnak F Mertz P Calcineurin form and function Physiol Rev 2000 80 1483 521 11015619
Im SH Rao A Activation and deactivation of gene expression by Ca2+/calcineurin-NFAT-mediated signaling Mol Cells 2004 18 1 9 15359117
Graef IA Chen F Chen L Kuo A Crabtree GR Signals transduced by Ca[2+]/calcineurin and NFATc3/c4 pattern the developing vasculature Cell 2001 105 863 75 11439183 10.1016/S0092-8674(01)00396-8
Lambrechts D Carmeliet P Sculpting heart valves with NFATc and VEGF Cell 2004 118 532 4 15339657 10.1016/j.cell.2004.08.022
Schulz RA Yutzey KE Calcineurin signaling and NFAT activation in cardiovascular and skeletal muscle development Dev Biol 2004 266 1 16 14729474 10.1016/j.ydbio.2003.10.008
Lee JI Ahnn J Calcineurin in animal behavior Mol Cells 2004 17 390 6 15232211
Sugiura R Sio SO Shuntoh H Kuno T Molecular genetic analysis of the calcineurin signaling pathways Cell Mol Life Sci 2001 58 278 88 11289309
Mansuy IM Calcineurin in memory and bidirectional plasticity Biochem Biophys Res Commun 2003 311 1195 208 14623305 10.1016/j.bbrc.2003.10.046
Molkentin JD Lu JR Antos CL Markham B Richardson J Robbins J Grant SR Olson EN A calcineurin-dependent transcriptional pathway for cardiac hypertrophy Cell 1998 93 215 28 9568714 10.1016/S0092-8674(00)81573-1
Ladner CJ Czech J Maurice J Lorens SA Lee JM Reduction of calcineurin enzymatic activity in Alzheimer's disease: correlation with neuropathologic changes J Neuropathol Exp Neurol 1996 55 924 31 8759782
Miyakawa T Leiter LM Gerber DJ Gainetdinov RR Sotnikova TD Zeng H Caron MG Tonegawa S Conditional calcineurin knockout mice exhibit multiple abnormal behaviors related to schizophrenia Proc Natl Acad Sci U S A 2003 100 8987 92 12851457 10.1073/pnas.1432926100
Gooch JL Pergola PE Guler RL Abboud HE Barnes JL Differential expression of calcineurin A isoforms in the diabetic kidney J Am Soc Nephrol 2004 15 1421 9 15153553 10.1097/01.ASN.0000128076.91545.BB
Fuentes JJ Genesca L Kingsbury TJ Cunningham KW Perez-Riba M Estivill X de la Luna S DCR1, overexpressed in Down syndrome, is an inhibitor of calcineurin-mediated signaling pathways Hum Mol Genet 2000 9 1681 90 10861295 10.1093/hmg/9.11.1681
Lian Q Ladner CJ Magnuson D Lee JM Selective changes of calcineurin (protein phosphatase 2B) activity in Alzheimer's disease cerebral cortex Exp Neuro 2001 167 158 65 10.1006/exnr.2000.7534
Asai A Qiu J Narita Y Chi S Saito N Shinoura N Hamada H Kuchino Y Kirino T High level calcineurin activity predisposes neuronal cells to apoptosis J Biol Chem 1999 274 34450 8 10567426 10.1074/jbc.274.48.34450
Wang HG Pathan N Ethell IM Krajewski S Yamaguchi Y Shibasaki F McKeon F Bobo T Franke TF Reed JC Ca2+-induced apoptosis through calcineurin dephosphorylation of BAD Science 1999 284 339 43 10195903 10.1126/science.284.5412.339
Baksh S DeCaprio JA Burakoff SJ Calcineurin regulation of the mammalian G0/G1 checkpoint element, cyclin dependent kinase 4 Oncogene 2000 19 2820 7 10851085 10.1038/sj.onc.1203585
Kahl CR Means AR Regulation of cell cycle progression by calcium/calmodulin dependent pathways Endocr Rev 2003 24 719 36 14671000 10.1210/er.2003-0008
Andre N Roquelaure B Conrath J Molecular effects of cyclosporine and oncogenesis: a new model Med Hypotheses 2004 63 647 52 15325009 10.1016/j.mehy.2004.03.030
Hojo M Morimoto T Maluccio M Asano T Morimoto K Lagman M Shimbo T Suthanthiran M Cyclosporine induces cancer progression by a cell-autonomous mechanism Nature 1999 397 530 4 10028970 10.1038/17401
Padma S Subramanyam C Clinical significance of serum calcineurin in acute leukemia Clin Chim Acta 2002 321 17 21 12031587 10.1016/S0009-8981(02)00094-3
Gravitt PE Peyton CL Apple RJ Wheeler CM Genotyping of 27 human papillomavirus types by using L1 consensus PCR products by a single-hybridization, reverse line blot detection method J Clin Microbiol 1998 36 3020 7 9738060
Sanli UA Uslu R Karabulut B Sezgin C Selvi N Aydin HH Saydam G Goker E Omay SB Alterations in the activity and expression of serine/threonine protein phosphatases during all trans retinoic acid-induced apoptosis in breast cancer cells Oncol Rep 2003 10 2083 8 14534748
Namgaladze D Hofer HW Ullrich V Redox control of calcineurin by targeting the binuclear Fe2+-Zn2+ center at the enzyme active site J Biol Chem 2002 277 5962 9 11741966 10.1074/jbc.M111268200
Ullrich V Namgaladze D Frein D Superoxide as inhibitor of calcineurin and mediator of redox regulation Toxicol Lett 2003 139 107 10 12628745 10.1016/S0378-4274(02)00424-1
Wang X Culotta VC Klee CB Superoxide dismutase protects calcineurin from inactivation Nature 1996 383 434 7 8837775 10.1038/383434a0
Ahmed MI Fayed ST Hossein H Tash FM Lipid peroxidation and antioxidant status in human cervical carcinoma Dis Markers 1999 15 283 91 10689551
Manju V Balasubramanian V Nalini N Oxidative stress and tumor markers in cervical cancer patients J Biochem Mol Biol Biophys 2002 6 387 90 14972792 10.1080/1025814021000036115
Giuliano A Cervical carcinogenesis: the role of co-factors and generation of reactive oxygen species Salud Publica Mex 2003 45 S354 60 14746028
Choi H Park CS Kim BG Cho JW Park JB Bae YS Bae DS Creatine kinase B is a target molecule of reactive oxygen species in cervical cancer Mol Cells 2001 12 412 7 11804344
Shirane M Nakayama KI Inherent calcineurin inhibitor FKBP38 targets Bcl-2 to mitochondria and inhibits apoptosis Nat Cell Biol 2003 5 28 37 12510191 10.1038/ncb894
Lin HY Michtalik HJ Zhang S Andersen TT Van Riper DA Davies KK Ermak G Petti LM Nachod S Narayan AV Bhatt N Crawford DR Oxidative and calcium stress regulate DSCR1 (Adapt78/MCIP1) protein Free Radic Biol Med 2003 35 528 39 12927602 10.1016/S0891-5849(03)00358-7
Stoler MH Rohan TE and Shah KV The pathology of cervical neoplasia Cervical Cancer: From etiology to prevention 2004 Kluwer Academic Publishers. USA 3 60
Neal JW Clipstone NA A constitutively active NFATc1 mutant induces a transformed phenotype in 3T3-L1 fibroblasts J Biol Chem 2003 278 17246 54 12598522 10.1074/jbc.M300528200
Jauliac S Lopez-Rodriguez C Shaw LM Brown LF Rao A Toker A The role of NFAT transcription factors in integrin-mediated carcinoma invasion Nat Cell Biol 2002 4 540 4 12080349 10.1038/ncb816
Steffan NM Bren GD Frantz B Tocci MJ O'Neill EA Paya CV Regulation of IkB alpha phosphorylation by PKC- and Ca2+-dependent signal transduction pathways J Immunol 1995 155 4685 91 7594468
Nair A Venkatraman M Maliekal TT Nair B Karunagaran D NF-kappaB is constitutively activated in high-grade squamous intraepithelial lesions and squamous cell carcinomas of the human uterine cervix Oncogene 2003 22 50 8 12527907 10.1038/sj.onc.1206043
| 15801986 | PMC1087859 | CC BY | 2021-01-04 16:40:10 | no | Cancer Cell Int. 2005 Apr 1; 5:7 | utf-8 | Cancer Cell Int | 2,005 | 10.1186/1475-2867-5-7 | oa_comm |
==== Front
J CarcinogJournal of Carcinogenesis1477-3163BioMed Central London 1477-3163-4-71584273310.1186/1477-3163-4-7ResearchSYK expression in human breast cancer Elkak AE [email protected] Sarakbi W [email protected] K [email protected] The Breast Unit, St George's Hospital and Medical School, Blackshaw Road London, SW17 0QT, UK2005 20 4 2005 4 7 7 18 3 2005 20 4 2005 Copyright © 2005 Elkak et al; licensee BioMed Central Ltd.2005Elkak 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
Syk (Splenic Tyrosine Kinase) is an intracellular receptor protein kinase involved in cell proliferation, differentiation and phagocytosis. It has been studied in T and B lymphocytes, NK cells and platelets. The strong expression of Syk in mammary gland prompted research into its potential role in mammary carcinogenesis. There have been very few studies about its role in breast cancer with conflicting results. This study aims to investigate the hypothesis that Syk expression is down-regulated in breast cancer compared with ANCT and the association between its expression and clinicopathological parameters.
Materials and methods
mRNA was extracted from 48 breast cancer specimens. Relative Syk to ribosomal RNA expression was determined by RT-PCR and Taqman methodology. Mann-Whitney U test was used to examine the association between Syk expression in cancer and ANCT. Spearman's rank correlation test was used to examine the association between Syk expression in tumours and patients' age, tumour size, tumour grade, estrogen and progesterone receptor status, lymph node metastasis, vascular invasion and clinical outcome.
Results
The median for the relative value of Syk expression was 0.17 and 0.18 (range: 0.12 – 0.56 and 0.0 – 1.77) for tumours and ANCT respectively. There was no significant association between Syk expression in cancers and ANCT (p= 0.598) nor between Syk expression in tumours and patients' age, tumour size, tumour grade, estrogen and progesterone receptor status, lymph node metastasis, vascular invasion or prognosis.
Conclusion
This study shows that Syk mRNA expression does not seem to vary between breast tumours and ANCT. Furthermore, we observed no significant association between Syk expression and clinicopathological parameters.
==== Body
Introduction
Breast cancer development and progression are thought to occur through a multistep process, including oncogene activation and mutation or loss of tumour suppresser genes [1]. Protein-tyrosine kinases (PTKs) have been classified into two families: trans-membrane receptors family and non- receptor family [2-5]. More than 30 receptor type PTKs have been identified in mammals and divided into 11 subfamiliies [6], of which we mention Tec, JAK, Sck, Fes, Abl, Fak, Scr and Syk [6,7]. Syk PTKs family comprises Syk and ZAP-70 [6]. Syk (named after Spleen Tyrosine Kinase) is associated with cell surface receptors to amplify receptor-activated signals inside the cells [6]. Syk family contains tandem Scr homology 2 (SH2) and c-terminal kinase domains interrupted by two interdomains A and B [8-12]. Syk assembles into signalling complexes at the plasma membrane via interaction between its tandem SH2 domains and a tyrosine-phosphorylated immunoreceptor tyrosine-based activation motif (ITAM), and this binding activates Syk [6,11,13,14] which is then tyrosine phosphorylated by autophosphorylation [5,6]. ITAM becomes phosphorylated on both tyrosine residues creating high affinity binding sites for the tandem SH2 domains [15]. Receptors containing these ITAMs act to initiate an activation cascade [16]. For example, stimulation of the B cell receptors (BCR) initiates a biochemical cascade, in which PTKs activation is the earliest known event [16]. Thus, Syk is recruited, activated by binding of SH2 to ITAM and then subsequently phosphorylated by autophosphorylation. Activated Syk phosphorylates its own substrate, which mediates further functions in the cells [16]. Syk is involved in mediating diverse cellular responses such as proliferation, differentiation and phagocytosis [12,17,18]. This includes T cell receptor (TCR) function and thymocyte development [7,12], B cell antigen receptor signalling [19] and natural killer (NK) cell differentiation (although not essential) and platelet activation by collagen [13]. Syk is not tyrosine phosphorylated in resting cells, possibly because it is dephosphorylated by specific protein tyrosine phosphatases [6].
Syk expression was found in a variety of organs such as spleen, heart, mammary gland [20] and in hepatocytes [21], haemtopoietic cells [22] and colonic carcinoma cells [23]. In an article published in Nature, Coopman et al [18] have reported that Syk was commonly expressed in normal human breast tissue, benign breast lesions and low-tumorigenic breast cancer cell lines whereas Syk mRNA and protein were low or undetectable in invasive breast carcinoma tissue and cell lines. In contrast with that, Fluck et al [20] demonstrated that aggressive metastasising mammary gland tumours expressed higher level of Syk compared with well differentiated non-metastasising tumours in a murine model of breast cancer. More recently, Yuan and colleagues [24] demonstrated that Syk was frequently inactivated through an epigenetic pathway in breast cancer. Using breast cancer cell lines, and human breast cancer tissues and PCR methodology, the authors observed that Syk was hypermethylated and therefore inactive in 12 (32%) of 37 breast tumours whereas all of the matched neighbouring normal breast tissue exhibited unmethylated DNA status (i.e. active Syk). Such observations are consistent with the hypothesis that Syk functions as a tumour suppresser gene and DNA methylation of this gene seems to play an important role in mammary carcinogenesis [25,26]. Other authors [22,27] have demonstrated that Syk is a prognostic factor in breast cancer. However, Wang et al, using immunoblotting technique, have found that breast cancer tissues expressed variable levels of Syk compared with normal breast tissues [5].
The present study aims to determine Syk mRNA expression in both breast cancer and adjacent normal breast tissue (ANCT) using RT-PCR methodology and correlate the expression in cancer with patients' age, tumour size, tumour grade, estrogen and progesterone receptor status, lymph node metastasis, vascular invasion and clinical outcome.
Materials and methods
Institutional guidelines including ethical approval and informed consent were followed. Patients were treated with wide local excision or mastectomy and axillary node dissection. Patients with estrogen and /or progesterone positive tumours received tamoxifen. Adjuvant radiotherapy was administered to all patients who had breast conserving surgery and to patients undergoing mastectomy for node positive disease. Chemotherapy was given to fit patients (<65 years) with lymph node involvement or high grade tumours.
A 50 mg tumour specimen and an other 50 mg of macroscopically normal adjacent normal breast tissue (ANCT) were obtained from each patient. They were frozen in liquid nitrogen within 30 minutes of excision and stored at -70°C until use.
RNA extraction
mRNA was extracted from 25 tumours and matching 23 ANCT using Absolutely RNA, RT-PCR Miniprep (Stratagene) Kit according to the manufacturer's protocol. Extraction of RNA was carried out in a laboratory separate to that used for RT-PCR.
RT Step
cDNA was synthesised. Conditions for reverse transcriptase were 25°C for 10 minutes, 48°C for 30 minutes and 95°C for 5 minutes. Genomic contamination was tested for and was virtually nil.
PCR
Taqman RT-PCR was performed for each sample using ABI PRISM 7700 Sensor Detection System (Applied Biosystems). The primers were designed using the Primer Selection Software. The primers used were ATT AGC AGA AGC AAA TGT CAT GCA (forward) and CCT CGC ATA TCC CGA TCA TC (reverse). PCR was carried out for 40 cycles of 15 seconds at 95°C and 1 minute at 60°C.
The relative expression of Syk to that of ribosomal RNA was determined using the Sequence Detection System Software. This experiment was carried out three times and the average values were calculated and used for statistical analysis. The positive and negative controls were ribosomal RNA and sterile water respectively.
Statistical analysis
Mann-Whitney U test was used to examine the association between Syk expression in cancer and ANCT. Spearman's rank correlation test was used to examine the association between Syk expression in tumours and patients' age, tumour size, tumour grade, estrogen and progesterone receptor status, lymph node metastasis, vascular invasion and clinical outcome.
Results
25 breast tumours and matching 23 ANCT were included in this study with patients age range 25–75 years (median 56.5, mean 56.80).
Syk expression in tumours and ANCT
The median for the relative value of Syk expression was 0.17 and 0.18 (range: 0.12 – 0.56 and 0.0 – 1.77) for tumours and ANCT respectively. Mann-Whitney U test showed no significant association between Syk expression in cancers and ANCT (p= 0.598). (Figures 1 and 2)
Figure 1 Scatter plot to show the relationship between Syk expression in cancers and normal breast tissues.
Figure 2 Box plot to show the relationship between Syk expression in cancers and normal breast tissues.
Correlation between Syk expression in tumours and clinicopathological parameters
The expression of Syk in tumours was correlated with patients' age, tumour size, tumour grade (well, moderately or poorly differntiated), estrogen and progesterone receptor status (positive, weak positive or negative), lymph node metastasis, vascular invasion, recurrence and clinical outcome (distant metastasis and death with a median follow up of 1.95 years). Spearman's rank correlation test showed no significant association between Syk expression in tumours and patients' age, tumour size, tumour grade, estrogen and progesterone receptor status, lymph node metastasis, vascular invasion or prognosis survival curves. Table 1 demonstrates the correlation coefficient and p values for the variables studied.
Table 1 the correlation coefficient and p values for the variables studied.
Age Size Hist Grade LN Vasc ER PR Rec OC
Corr coef -0.104 0.102 0.01 0.137 -0.007 0.141 0.089 -0.294 0.130 0.339
P value 0.653 0.679 0.963 0.626 0.977 0.603 0.708 0.208 0.595 0.143
Discussion
Our report in the present article represents the largest study of Syk expression in human breast cancer and ANCT. It has that Syk mRNA is expressed in both human breast cancer and ANCT specimens. This observation is consistent with the established role of Syk in the control of cell proliferation and intracellular signalling in epithelial cells and immunocytes [12,17,18]. The lack of statistically significant difference between Syk mRNA expression in the whole tumour and ANCT specimens observed in this study may represent a true observation and therefore, Syk should not be considered a tumour suppressor gene. It does not, however, exclude the presence of different levels of Syk expression in malignant and benign mammary epithelial cells. Breast tumours are heterogenous lesions [5] that contain various components including fibroblasts and lymphocytes. The latter can make up to 45% of the whole breast tumour volume [28]. Lymphocytes are known to express significant levels of Syk and this could lead to a false overestimation of Syk mRNA in epithelial cancer cells. A more accurate estimation of Syk mRNA in epithelial elements can be achieved by using Laser capture microdissection (LCM) in order to separate epithelial cells from other components in both tumours and ANCT. Alternatively, in situ hybridization and immunohistochemistry techniques can be used to identify the cellular source of the enzyme (Coopman et al). The ANCT in our study was obtained from macroscopically normal breast tissue more than 10 mm away from the primary tumour. However, one cannot exclude the presence of malignant epithelial cells in these specimens, as histopathological verification of ANCT was not carried out. Another limitation of the present study is the fact that we measured mRNA transcript rather than the protein, which is the effective end product of the gene. However, experimental data from breast cancer cell lines suggest that there is a good correlation between mRNA and protein expression (Coopman et al). Furthermore, the presence of splice variants of Syk mRNA (Fluck et al) represents another limitation to our RT-PCR methodology. Our results also show that Syk mRNA expression in breast tumours as determined by RT-PCR methodology is independent of tumour stage and therefore it is unlikely to play a prognostic role in patients with breast cancer.
Abbreviations
Corr Coef: Correlation coefficient
Hist: Histology
LN: Lymph node metastasis
Vasc: Vascular invasion
ER: Estrogen receptors
PR: Progesterone receptors
Rec: Recurrence
OC: Clinical outcome
==== Refs
Stewart ZA Pietinpol JA Syk: a new player in the field of breast cancer Breast Cancer Res 2001 3 5 7 11250739 10.1186/bcr261
Taniguchi T Kobayachi T Kondo J Molecular cloning of a porcine gene Syk that encodes a 72-kDa protein-tyrosine kinase showing high susceptibility to proteolysis J Biol Chem 1991 266 15790 6 1874735
Cantley LC Auger KR Carpenter C Oncogenes and signal transduction Cell 1991 64 281 302 1846320 10.1016/0092-8674(91)90639-G
Hanks SK Quinn AM Hunter T The protein kinase family: conserved features and deduced phylogeny of the catalytic domains Science 1988 241 42 52 3291115
Wang L Duke L Zhang PS Alternative splicing disrupts a nuclear localization signal in spleen tyrosine kinase that is required for invasion suppression in breast cancer Cancer Res 2003 63 4724 30 12907655
Sada K Takano T Yanagi S Structure and function of protein- tyrosine kinase J Biochem 2001 130 177 86 11481033
Cheng AM Chan AC Protein tyrosine kinase in thymocyte development Curr Opin Immunol 1997 9 528 33 9287188 10.1016/S0952-7915(97)80106-9
Hanks SK Quinn AM Protein kinase catatlytic domain sequence database: identification of conserved features of primary structures and classification of family members Meth Enzymol 1991 200 38 62 1956325
Hatada MH Lu X Laird ER Molecular basis for interaction of the protein tyrosine kinase ZAP-70 with the T-cell receptor Nature 1995 377 32 8 7659156 10.1038/377032a0
Koch CA Anderson D Moran M SH2 and SH3 domains: elements that control interaction of cytoplasmic signalling proteins Science 1991 252 668 74 1708916
Futterer K Wong J Grucza RA Structural basis for Syk tyrosine kinase ubiquity in signal transduction pathways revealed by the crystal structure of its regulatory SH2 domains bound to a dually phosphorylated ITAM peptide J Mol Biol 1998 281 523 37 9698567 10.1006/jmbi.1998.1964
Chu DH Morita CT Weiss A The Syk family of protein tyrosine kinase in T- cell activation and development Immunol Reviews 1998 165 167 80
Poole A Gibbins JM Turner M The Fc receptor γ-chain and tyrosine kinase Syk are essential for activation of mouse platelets by collagen EMBO J 1997 16 2333 41 9171347 10.1093/emboj/16.9.2333
Katz E Lareef MH Rassa JC MMTV Env encodes an ITAM responsible for transformation of mammary epithelial cells in three-dimentional culture J Exp Med 2005 201 431 9 15684322 10.1084/jem.20041471
Vely F Nunes JA Malisson B Analysis of immunoreceptor tyrosine- based activation motif (ITAM) binding to ZAP-70 by surface plasma resonance Eur J Immunol 1997 27 3010 4 9394831
Kurosaki T Johnson SA Pao L Role of the Syk autophosphorylation site and SH2 domains in B cell antigen receptor signalling J exp Med 1995 182 1815 23 7500027 10.1084/jem.182.6.1815
Indik ZK Park JG Pan XQ Induction of phagocytosis by a protein tyrosine kinase Blood 1995 1175 80 7858249
Coopman PJP Do MTH Barth M The Syk tyrosine kinase suppresses malignant growth of human breast cancer cells Nature 2000 406 742 7 10963601 10.1038/35021086
Kurosaki T Molecular mechanisms in B cell antigen receptor signalling Curr Opin Immunol 1997 9 309 18 9203415 10.1016/S0952-7915(97)80075-1
Fluck M Zurche G Andres A Molecular characterisation of the murine Syk protein tyrosine kinase cDNA transcripts and protein Biochem Biophys Res Commun 1995 213 273 81 7639745 10.1006/bbrc.1995.2126
Tsuchida S Yanagi S Inatome R Purification of a 72-kDA protein-tyrosine kinase from rat liver and its identification as Syk: involvement of Syk in signalling events of hepatocytes J Biochem 2000 127 321 7 10731700
Dejmek J Leandersson K Manjer J Expression and signalling of Wnt05a/discoidin domain receptor-1 and Syk plays distinct but decisive roles in breast cancer patient survival Clin Cancer Res 2005 11 520 8 15701836
Okamura S Ng CC Koyama K Identification of seven genes regulated by wild type p53 in a colon cancer cell line carrying a well controlled wild type p53 expression system Oncol Res 1999 11 281 5 10691030
Yuan Y Mendez R Sahin A Hypermethylation leads to silencing of the SYK gene in human breast cancer Cancer Res 2001 61 5558 61 11454707
Yuan Y Liu H Sahin A Reactivation of Syk expression by inhibition of DNA methylation suppresses breast cancer cell invasiveness Int J Cancer 2005 113 654 9 15455373 10.1002/ijc.20628
Moroni M Soldatenkov V Zhang L Progressive loss of Syk and abnormal proliferation in breast cancer cells Cancer Res 2004 64 7346 54 15492255
Toyama T Iwase H Yamashita H Reduced expression of the Syk gene is correlated with poor prognosis in human breast cancer Cancer Lett 2003 189 97 102 12445683 10.1016/S0304-3835(02)00463-9
Balch CM Riley LB Bae YJ Patterns of human tumour-infiltrating lymphocytes in 120 human cancers Arch Surg 1990 125 200 5 1689143
| 15842733 | PMC1087860 | CC BY | 2021-01-04 16:39:20 | no | J Carcinog. 2005 Apr 20; 4:7 | utf-8 | J Carcinog | 2,005 | 10.1186/1477-3163-4-7 | oa_comm |
==== Front
Cardiovasc UltrasoundCardiovascular Ultrasound1476-7120BioMed Central London 1476-7120-3-91580788710.1186/1476-7120-3-9ReviewDiastolic dysfunction and diastolic heart failure: diagnostic, prognostic and therapeutic aspects Galderisi Maurizio [email protected] Division of Cardioangiology with CCU Department of Clinical and Experimental Medicine "Federico II" University, Medical School Napoli, Italy2005 4 4 2005 3 9 9 5 1 2005 4 4 2005 Copyright © 2005 Galderisi; licensee BioMed Central Ltd.2005Galderisi; 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.
Left ventricular (LV) diastolic dysfunction (DD) and diastolic heart failure (HF), that is symptomatic DD, are due to alterations of myocardial diastolic properties. These alterations involve relaxation and/or filling and/or distensibility. Arterial hypertension associated to LV concentric remodelling is the main determinant of DD but several other cardiac diseases, including myocardial ischemia, and extra-cardiac pathologies involving the heart are other possible causes. In the majority of the studies, isolated diastolic HF has been made equal to HF with preserved systolic function (= normal ejection fraction) but the true definition of this condition needs a quantitative estimation of LV diastolic properties. According to the position of the European Society of Cardiology and subsequent research refinements the use of Doppler echocardiography (transmitral inflow and pulmonary venous flow) and the new ultrasound tools has to be encouraged for diagnosis of DD. In relation to uncertain definitions, both prevalence and prognosis of diastolic heart failure are very variable. Despite an apparent lower death rate in comparison with LV systolic HF, long-term follow-up (more than 5 years) show similar mortality between the two kinds of HF. Recent studies performed by Doppler diastolic indexes have identified the prognostic power of both transmitral E/A ratio < 1 (pattern of abnormal relaxation) and > 1.5 (restrictive patterns). The therapy of LV DD and HF is not well established but ACE-inhibitors, angiotensin inhibitors, aldosterone antagonists and β-blockers show potential beneficial effect on diastolic properties. Several trials, completed or ongoing, have been planned to treat DD and diastolic HF.
Diastolic dysfunctionDiastolic heart failureLeft ventricleCardiac catheterizationDoppler echocardiography
==== Body
Introduction
Heart failure (HF) is a clinical syndrome whose symptoms and signs are due to increased extravascular water and decreased tissue / organ perfusion. The definition of the mechanisms inducing HF needs the measurement of both left ventricular (LV) systolic and diastolic function since HF may occur in patients with either normal or abnormal LV ejection fraction (EF) [1].
Arterial hypertension is the most common risk factor for HF in the general population and myocardial infarction, LV hypertrophy (LVH) and valve heart disease represent predictors of subsequent HF in hypertensive patients of both genders [2]. The progression of hypertensive cardiomyopathy towards HF includes serial LV changes – LV concentric remodelling and LVH – whose prognostic role is recognized [3-5]. In presence of these LV geometric abnormalities, deep modifications of LV diastolic properties occur. These modification are globally defined as LV diastolic dysfunction (DD) and include alterations of both relaxation and filling [6,7] which can precede alterations of LV systolic function and be per se main determinants of symptoms and signs of HF. Several other cardiac pathologies as well as extra-cardiac diseases involving secondarily the left ventricle can also affect myocardial diastolic properties and determine LV DD.
LV DD and diastolic HF, that is the symptomatic DD, represent clinical entities which can be described at different levels, from the hystologic and ultrastructural features to the clinic manifestations and diagnostic instrumental findings, until the prognostic and therapeutic aspects. The growing interest for DD and for diastolic HF has been developed gradually in the last 10–15 years. It rises mainly from the advancement of non invasive imaging tools, above all Doppler echocardiography, which, to date, allows easy and repeatable identification of LV diastolic abnormalities, and by the growing impulse of pharmaceutical industry, at constant search of new therapeutic applications. In relation to the increase of the average life and the future projections which suggest HF as the most important pathology of the new millennium, particularly in the elderly population, it has to be understood how diagnosis, prognosis and therapeutic management of DD represent very attractive perspectives.
Physiology of diastole
Although in normal hearts the transition from contraction to relaxation begins much more before LV end-systole, i.e., at 16% to 20% of the ejection period [8,9] and even prior to aortic valve opening when LV contractility is severely impaired (9), the traditional definition of diastole (in ancient Greek language the term διαστολε means "expansion"), includes the part of the cardiac cycle starting at the aortic valve closure – when LV pressure falls below aortic pressure – and finishing at the mitral valve closure. A normal LV diastolic function may be clinically defined as the capacity of the left ventricle to receive a LV filling volume able in its turn to guarantee an adequate stroke volume, operating at a low pressure regimen.
In merely descriptive terms, diastole can be divided in 4 phases [10]:
1. Isovolumetric relaxation, period occurring between the end of LV systolic ejection (= aortic valve closure) and the opening of the mitral valve, when LV pressure keeps going its rapid fall while LV volume remains constant. This period Is mainly attributed to the active LV relaxation, with a lower, variable contribution of elastic recoil of the contracted fibers;
2. LV rapid filling, which begins when LV pressure falls below left atrial pressure and the mitral valve opens. During this period the blood has an acceleration which achieves a maximal velocity, direct related to the magnitude of atrio-ventricular pressure, and stops when this gradient ends. This period represents a complex interaction between LV suction (= active relaxation) and visco-elastic properties of the myocardium (= compliance);
3. diastasis, when left atrial and LV pressures are almost equal and LV filling is essentially maintained by the flow coming from pulmonary veins – with left atrium representing a passive conduit – with an amount depending of LV pressure, function of LV "compliance".
4. atrial systole, which corresponds to left atrial contraction and ends at the mitral valve closure. This period is mainly influenced by LV compliance, but depends also by the pericardial resistance, by the atrial force and by the atrio-ventricular synchronicity (= ECG PR interval).
Cardiac catheterization allows to assess the pressure-volume relation along the overall cardiac cycle. Among the various hemodynamic measurements, τ (= time constant of the isovolumic-pressure decline) and DP/DV ratio, expression of LV end-diastolic myocardial stiffness, are the main invasive measurements of LV diastolic function [10]. On the other hand, Doppler recording of transmitral and pulmonary venous flow measure flow velocities and time intervals, whose variations occur in relation to analogous variations of left atrial and LV pressures [11,12]. Thus, Doppler parameters provide important information about dynamics of LV filling and LV diastolic properties during disease evolution or improvement [13].
Ultrastructural features of diastolic dysfunction
The extracellular matrix (ECM), corresponding to fibrillar collagen, is an important structure for processes of both myocardial contraction and relaxation. It facilities the arrangement of the cardiomyocites into the most suitable allocation for the development of force and shortening, giving a substantial support to the maintenance of an effective myocardial performance [14]. The myocardial remodelling is accompanied by changes of myocardial cell factors but also of the ECM where fibroblast proliferation, alteration of the collagen network and increase in interstitial and perivascular collagen are strongly promoted by renin-angiotensin-aldosterone system [15]. ECM has, therefore, to be considered a dynamic entity playing a fundamental role into the myocardial adaptation to physiologic and pathologic stress [14]. ECM undergoes an intense turnover, due to balanced action of metalloproteases, proteolytic enzymes activated by several factors including also BNP, and tissue inhibitors counterbalancing the activity of metalloproteases [16]. Thus, if the collagen destruction alters both geometry and function of contractile myocardium throughout an up-regulation of metalloproteases, on the other hand myocardial fibrosis occurs because of an imbalance where collagen deposition prevails over its degradation. According to the ultrastructural view, we can hypothesize two opposite pathologic conditions: the first one, when the collagen loss, e.g, after acute myocardial infarction, deprives myocardium of its indispensable support structure, thus inducing a reduction of myocardial systolic function; the second one, when the accumulation of the same collagen, main component of myocardial fibrosis, determines both systolic and diastolic myocardial dysfunction. In this context not only the total amount of collagen is main determinant of LV diastolic stiffness but also distribution, configuration, disorganization of collagen fibers (cross-hatching), and ratio of collagen type I to collagen type III play an important role [14].
Clinical, hemodynamic and diagnostic and aspects of diastolic dysfunction
In the clinical setting the coexistence of systolic and diastolic dysfunction in patients with symptomatic HF occurs very often. In fact, LV stiffness (or compliance) is related to the length of myocardial fibers, reflecting in its turn on LV end-diastolic dimensions. LV diastolic function, through the influence on left atrial and capillary wedge pressures, determines the onset of symptom in patients with prevalent LV systolic dysfunction too.
In parallel to the ultra-structural level, the clinical progression of HF may follow two different routes. In the first one, as it happens after acute myocardial infarction, post-infarction LV dilation (= remodelling) leads to systolic dysfunction and/or systolic heart failure. In the second one, LV structural abnormalities (= LV concentric geometry) induce functional alterations of DD. When diastolic dysfunction becomes symptomatic – that is, when dyspnoea occurs – diastolic heart failure rises.
The majority of patients affected by isolated diastolic HF show symptoms not at rest but in relation to stress conditions (II NYHA class). Symptoms can be induced or worsened by, firstly, physical exercise but also by events as anaemia, fever, tachycardia and some systemic pathologies. In particular, tachycardia reduces the time needed for global LV filling, thus inducing an increase of left atrial pressure and consequent appearance of dyspnoea, because of accumulation of pulmonary extravascular water.
The diagnosis of HF can be performed obviously by the simple clinical examination but the identification of the diastolic origin needs an instrumental assessment. In fact, the objective examination of patients with diastolic HF allows to notice the same signs occurring for systolic HF and even the thoracic X-ray can not be useful to distinguish the two entities. ECG can show signs of LVH, due to hypertensive cardiomyopathy or other causes. DD may be asymptomatic and, therefore, identified occasionally during a Doppler echocardiographic examination (Figure 1). The diagnostic importance of this tool rises from the high feasibility of transmitral Doppler indexes of diastolic function, shown even in studies on population [17], such to be suitable and accurate also for serial evaluations over time. To date, standard Doppler indexes may be efficaciously supported by the evaluation of pulmonary venous flow [18] (Figure 2) and by new ultrasound technologies as Tissue Doppler [19] (Figure 3) and color M-mode derived flow propagation rate [20]. The application of maneuvers (Valsalva, leg lifting) [21,22] to Doppler transmitral pattern and/or different combination of standard transmitral Doppler with the new tools (ratio between atrial reverse velocity duration and transmitral A velocity duration, ratio between transmitral E peak velocity and Tissue Doppler derived Em of the mitral annulus or flow propagation velocity [Vp]) are sufficiently reliable to predict capillary wedge pressure and to distinguish accurately variations of LV end-diastolic pressure [23,24]. Some of these tools are effective even in particular situations as sinus tachycardia [25] and atrial fibrillation [26] while the pulmonary venous flow or the Valsalva maneuver applied to transmitral inflow has to be preferred in the case of mitral valve prosthesis and aortic valve regurgitation [27]. In addition, Tissue Doppler is also able to "read" the percentage of myocardial fibrosis [28], primum movens of DD. Alone or, better, combined together, these tools permits to recognize normal diastole as well as to diagnose and follow the progression of DD from the pattern of abnormal relaxation (grade I of DD) until pseudonormal (grade II) and restrictive (grade III-IV) patterns (Table 1).
Figure 1 In the left screen, methodological outline for the measurement of Doppler transmitral indexes of diastolic function. In the right screen, normal diastolic pattern (upper part) and pattern of abnormal relaxation (lower part). A = atrial velocity (m/s), DT = deceleration time of E velocity (ms), E = early diastolic velocity (cm/s), IVRT = isovolumic relaxation time (ms)
Figure 2 In the left screen, methodological outline for the measurement of pulmonary veins flow. In the right screen, normal pulmonary veins flow pattern (upper part) and pattern of abnormal relaxation (lower part).
Figure 3 In the left screen, methodological outline for the measurement of Tissue Doppler indexes. In the right screen, normal myocardial diastolic pattern (upper part) and pattern of abnormal myocardial relaxation (lower part). Am = myocardial atrial velocity (cm/s), CTm = myocardial contraction time (ms), DTm = myocardial deceleration time of Em(ms), Em = myocardial early-diastolic velocity (cm/s), PCTm = myocardial pre-contraction time (ms), RTm = myocardial relaxation time (ms).
Table 1 Doppler echocardiographic patterns of current echocardiographic tools in relation to the grading of LV diastolic dysfunction
Parameter Normal pattern Pattern of abnormal relaxation (Grade I) Pseudonormal pattern (Grade II) Restrictive patterns (Grades III-IV)
E/A >1 <1 1 – 2 ≥2
DT (ms) 160 – 210 >220 150 – 200 <150
IVRT (ms) 70 – 90 >95 60 – 95 <60
S/D 1.3 – 1.5 1.6 – 2.0 <1 0.40 – 0.60
AR (m/sec) 0.22 – 0.32 0.21 – 0.28 ≥0.35 ≥0.25
Em (cm/sec) >8 <8 <8 <5
Vp (cm/sec) >55 <45 <45 <35
E/Em < 8 > 16
E/Vp > 2.5
AR = Atrial retrograde velocity, DT = Deceleration time, E/A = Transmitral E/A ratio, E/Em = Transmitral early diastolic velocity to myocardial early diastolic velocity of lateral mitral annulus by Tissue Doppler, Em = myocardial early diastolic velocity by Tissue Doppler at lateral mitral annulus, IVRT = Isovolumic relaxation time, = Systolic velocity to diastolic velocity ratio by pulmonary veins assessment, Vp = Velocity propagation
By the hemodynamic point of view, the differences between diastolic and systolic HF are expressed by the pressure-volume loop (Figure 4) [29]. When systolic HF occurs, increased LV filling pressures correspond to increased LV volumes, with a displacement of the loop upon and at right. In the case of diastolic HF, the increase of LV filling pressures occur in the presence of normal or even reduced LV volumes, thus moving the loop up and to the left. It is obvious that in the more advanced stages of HF, diastolic and systolic dysfunction coexist.
Figure 4 Pressure-volume loops in systolic HF (upper part) and diastolic HF (lower part). The continuous black line refers to normal, the interrupted red line to the pathologic condition.
Determinants of diastolic dysfunction
LV DD develops in several cardiac diseases [30] as well as in extra-cardiac pathologies involving the heart (accumulation diseases as amyloidosis, thyroid disorders, acromegaly and others) [31,32] and in myocardial ischemia due to coronary artery stenosis or even to isolated dysfunction of coronary microcirculation [33]. However, the main cause of DD is arterial hypertension [5-7]. Overweight and obesity, often coexisting with the same hypertension, deeply affects LV diastolic function, forcing the left ventricle to a working overload [34]. In this view, DD represents one of the cardiac consequences of pluri-metabolic syndrome, where arterial hypertension, obesity, glucose intolerance and hypertrygliceridemia cohabit in the same subject, having their common matrix in the insulin resistance. High levels of insulin resistance, often evident in arterial hypertension [35], are positively associated with the prolongation of isovolumic relaxation time, independent of LV geometric changes and of increased afterload [36]. The alteration of diastolic isovolumic relaxation is probably due to an increment of intracellular calcium, which has been observed in insulin resistant hypertensives and is induced in its turn by an abnormal re-uptake of calcium by sarcoplasmic reticulum [37]. Also the hormones produced by adipose tissue, as leptin – involved into the control of body weight throughout food absorption and energy-giving cost – negatively affects LV diastolic function [38]. The association of arterial hypertension and diabetes mellitus worsens furher Doppler indexes of LV diastolic function as shown into the population of the Strong Heart Study [39].
It is controversial whether LV DD is necessarily accompanied on the development of LVH or rises up independent of it [5-7,40-43]. It is true that DD is a direct sequence of pressure overload, associated to elevated 24-hour blood pressure [40] and even more to the increment of night-time diastolic blood pressure [43]. Recent studies point out that the diastolic abnormalities of hypertensive patients are related to inappropriately high levels of LV mass, disproportionate to the hemodynamic load predicted by the individual body size and cardiac load, more than to the values of LV mass which traditionally define LVH [44]. Inappropriately high LV mass is a potent predictor of cardiovascular risk in hypertensive patients, in presence as in absence of clear cut LVH [45]. The concept of DD onset preceding the appearance of LVH is consistent with the observation that BNP, whose levels grow gradually with the progression of DD grading (from abnormal relaxation until restrictive Doppler patterns) [46], are increased in patients with diastolic HF independent of the magnitude of LV mass [47]. Even a new ultrasound technology as Tissue Doppler supports the hypothesis of an early evidence of DD in hypertensive heart: myocardial DD (= Em/Am ratio < 1 at the level of multiple LV walls in the apical views) is detectable before the appearance of the abnormalities involving LV transmitral inflow and is uniform in non hypertrophic patients while it becomes prominent at the septum in presence of overt LVH [48]. Table 2 reports the differential characteristics involving in the meantime the myocardial ultra-structure and LV geometry in systolic and diastolic HF: it is clear that diastolic HF is associated to both increase of collagen amount and LV concentric geometry [49]. This concept is further supported by the HyperGEN study where delayed LV relaxation is independently associated with concentric LV geometry in 1384 hypertensive participants including obese and diabetic patients [50].
Table 2 Differential characteristics of LV geometry between systolic and diastolic HF (modified from Zile MR [49]).
Characteristic Systolic HF Diastolic HF
LV volume ↑
↑↑ N (o ↓)
LV mass ↑ ↑
LV geometry Eccentric Concentric
Cardiomyocites ↑ Length ↑ Diameter
Extracellular Matrix ↓(o ↑ o N) Collagen ↑↑ Collagen
HF = heart failure, LV = left ventricular, N = normal
Definition and classification criteria for diastolic HF
The evidence of acute HF in absence of overt LV systolic dysfunction rises by the experience of Gandhi and coworkers [51]: thirty-height hypertensive patients affected by pulmonary oedema, undergoing echocardiographic examination during the acute episode and after clinical stabilization respectively (1–3 day after), did not show significant variations of LV EF (50 ± 15% and 50 ± 13% respectively, NS) and of wall motion score index between the two examinations. This clinical condition, defined as heart failure with preserved systolic function or, better, with normal EF, has been made equal to isolated diastolic heart failure. A truly correct definition of this clinical entity should, however, be done on the grounds of direct estimation of LV diastolic function and establishment of reference normal values. Strong controversy has been developed in the previous years about this issue, with opposite scientific positions. The American point of view, corresponding to the Framingham Heart Study investigators, has sustained the concept that diastolic HF is "definite" only when an invasive hemodynamic assessment shows diastolic alterations in the temporal proximity of the acute episode [52]. On the other hand, the European point of view (European Group on Diastolic Heart Failure) has defined diastolic HF according to criteria including clinical examination, echocardiographic assessment (normal EF) and Doppler indexes (derived by both transmitral inflow e pulmonary veins flow), whose normal partition values are referred for age ranges [53] (Table 3). Despite the obvious superiority of the invasive technique [54], it has to be taken into account that the need of cardiac catheterization for establishing a definite diagnosis of diastolic HF raises practical and even ethical issues. Practical issues are related to the low priority such examination would have in a cath-lab overloaded by coronary procedures and to the poor interest of hemodynamists in the assessment of indexes of LV diastolic function. Ethical concern lies upon the fact that the present reliance on echo-Doppler examination of LV diastolic function makes cardiac catheterization an useless invasive procedure to this end, except very particular cases. Moreover, if it is true that the prevalence of abnormal Doppler indexes (from 38% of isovolumic relaxation time to 64% for deceleration time) is much lower to that showed by the more reliable invasive measurements (92 % for LV end-diastolic pressure and 79 % for τ) [55], is also true that this can be, at least partially, due to the confounding influence of physiologic variables as age [56] and heart rate [57]. In this view, reference normal values of Doppler indexes of LV diastolic function should be done considering ranges of both age and heart rate. It is now current opinion that the diagnosis of diastolic HF can be made even without measurement of diastolic function if three criteria are present: 1) symptoms and signs of HF (Framingham criteria), 2) LV EF> 50%, and 3) ability to rule out mitral stenosis, pericardial disease, and non cardiac causes of dyspnoea, oedema and fatigue [58]. Recent evidences further sustain the definite role of Doppler echocardiography to diagnose diastolic HF [59,60].
Table 3 Criteria for diastolic HF according to the European Society of Cardiology (53).
Signs and symptoms of HF Effort dyspnoea, Hortopnoea, III-IV tones, Pulmonary rales
Normal or mildly reduced LV systolic function EF ≥ 45 % e LVIDDi > 3.2 cm·m -2
Evidence of abnormalities LV of relaxation/filling and/or distensibility IVRT <30 years > 92 ms, 30–50 years > 100 ms, >50 years > 105 msc
E/A<1 + DT>220 ms + S/D<1.5 <50 years
E/A<0.5 + DT>280 msec + S/S>2.5 >-50 years
DT = deceleration time of E velocity, E/A = ratio of early diastolic velocity to atrial velocity, EF = ejection fraction, HF = heart failure, IVRT = isovolumic relaxation time, LV = left ventricular, LVIDDi = left ventricular internal diastolic diameter index, S/D = ratio of systolic to diastolic velocity of venous pulmonary veins
To date, however, no certain definition of diastolic HF exists and the recognition of its existence is not unanimously accepted [49]. Studies performed by both standard Doppler echocardiography [61] and Tissue Doppler [62,63] demonstrated how sub-clinic alterations of myocardial systolic function are already overt in diastolic HF. Because of the use of LV EF is a rather insensitive indicator of true LV myocardial contractility, the assessment of LV long-axis function by the simple M-mode of the mitral lateral annulus could help to identify initial LV systolic dysfunction [64]. Finally, it has also to be taken into account how concomitant variables, including obesity, chronic obstructive lung disease and even myocardial ischemia, can be confounding factors leading to "false" diagnosis of diastolic HF, particularly in the elderly population [65].
Prevalence of diastolic HF
The studies performed until now have assessed above all the prevalence of HF with normal EF, using standard echocardiography without Doppler. In a first meta-analysis of 1995, the investigators of the Framingham Heart Study [66] showed wide variability in the prevalence of this kind of HF (range = 13–74%) while a subsequent study involving the Framingham offspring cohort pointed out a 51% prevalence of overall HF [67]. Very recently, Hogg et al collected ten "cross-sectional" studies on population, in the United States as in several European countries, and found very high variability of HF with normal EF. The explanation of this variability is related mostly to different age and gender of participants. It has to be considered that this kind of HF is particularly frequent in the elderly population, occurs more often in the female gender and is associated much more with arterial hypertension and atrial fibrillation than to coronary heart disease [68]. The data collected between 1995 and 1999 from Italian Network on Congestive Heart Failure (IN-CHF) are strongly consistent with these results [69]. The choice of different cut-off points for normal LV EF can be an additional reason of variability for the prevalence of diastolic HF in the above mentioned studies.
Prognosis of diastolic heart failure
Great heterogeneity exists also for results in prognosis of diastolic HF. By the Framingham meta-analysis the annual mortality varies from 1.3% to 17.5% [66]. This wide variability depends by several factors including first of all, the modality used to classify this kind of HF – mostly according to the evidence of normal EF – but also age and follow-up duration. In a study by registry on 1291 hospitalized patients (70) the mortality was lower in patients with EF ≥ 50% than in those with EF ≤ 39% (OR = 0.69 95% CI 0.49–0.98, p = 0.04) . The Framingham offspring cohort informed that the rate of death after 5 years is 68% in patients with HF and normal EF in comparison with 82% of systolic HF, with a mortality, however, four times greater than that presented by healthy subjects [67]. Although Senni et al (71) did not find difference of mortality between the two kinds of HF in a 4-year follow-up of a population with mean age of 78 years, the analysis of Hogg and coworkers, assembling the results of recent cohort studies performed on patients hospitalized for HF, noticed how the percentage of mortality for patients with HF and normal EF, mild during the first year and half, becomes similar to that of systolic HF after 5–6 years of follow-up (68). It is worthy of note the recent study of Badano and coworkers who, using the ESC criteria to identify diastolic HF in 179 patients hospitalized with HF, do not observe significant difference in 6-month mortality in comparison with patients having prevalent LV systolic dysfunction [72].
Two important studied have finally pointed out the prognostic value of Doppler indexes of LV diastolic function and in particular of transmitral E/A ratio [73,74]. The first one, the PIUMA study [73], evidenced that the pattern of abnormal relaxation (= E/A ratio lower than that predicted individually by age and heart rate) increases the risk of cardiovascular events (odds ratio 1.57, 95% CI 1.1-2,18, p < 0.01) in a population of 1839 hypertensive patients during a 11 years follow-up. This prognostic value is independent of the effect of LV mass and even of ambulatory 24-hour blood pressure. In the second one, the Strong Heart Study [74], by a 3-year follow-up on a population of 3008 American Indians, a transmitral E/A ratio < 0.6 (= likewise pattern of abnormal relaxation) is associated to a doubled increase of mortality risk – despite not independent of other covariates – and an E/A ratio > 1.5 (= likewise pattern pseudonormal / restrictive) is associated to an threefold increase of cardiac mortality, which is also independent of several confounders including LVH. This result is consistent with the findings of the Framingham Heart Study, where an "U" relation between transmitral A velocity and risk of atrial fibrillation is detectable, and the arrhythmia appears independently associated with both A velocity increase (= abnormal relaxation) and E/A ratio increase (pattern pseudonormal / restrictive) [75]. These two studies, in particular the Strong Heart Study by data on mortality, are very consistent with the physiopathologic point of view of the Mayo Clinic investigators, who created an ingenious classification of Doppler-.derived DD some years ago [21]. In this classification, the pattern of abnormal relaxation (grade I of DD) and both reversible and non reversible restrictive patterns (grade III and IV respectively) are at opposite sides in the clinical progression towards the end stages of HF while the pseudo-normal pattern has an intermediate, but clinically crucial, position (Figure 5). In view of these findings and combining the value of the prognostic studies, we can suppose that the relatively long time (5–6 years) needed to assimilate the prognosis of diastolic HF to that of systolic HF depends mainly by the transition from the initial grade of DD, when the pattern of abnormal relaxation prevails and dyspnoea is overt only during exercise, to the more advanced stages, when the high LV end-diastolic pressure is associated to "end-stage" HF.
Figure 5 Results of overall and cardiac mortality in relation to transmitral E/A ratio in the Strong Heart Study [74] (upper panel)) and classification of DD grades (I-IV) according to Mayo Clinic suggestions [21] (lower panel). It can be observed a parallel behaviour between clinic progression and prognostic value of different grade of DD: the increment of mortality in Strong Heart Study has an "U" behavior, where E/A ratio <0.6 (grade I of DD) and >1.5 (grades II, III, IV) are both main predictors of mortality DD = diastolic dysfunction, NYHA = New York Heart Association, MAP = mean atrial pressure
Therapy of DD and diastolic HF
The objectives of the therapy for LV DD include the improvement of hemodynamic conditions, concerning both preload and afterload. The volume overload, such to induce episodes of acute HF, can be prevented or reduced by hypo-saline diet or also by a moderate diuretic administration.
Conceptually, both ACE-inhibitors and angiotensin-inhibitors can exert a beneficial effect on DD, since they reduce both afterload and preload, induce regression of LVH and decrease of myocardial interstitial fibrosis [70]. Also the aldosterone antagonists, as sprironalattone [76] and canrenone [77], able to reduce the myocardial fibrosis, can be suitable to this aim.
When DD is overt, it is also important to control heart rate and avoid tachycardia. β-blockers, and, with a lower extent, calcium antagonist verapamile, can be particularly useful. Lower heart rate induces prolongation of LV filling time, allowing to counterbalance the resistance to the diastolic inflow of a stiffened left ventricle. Last generation β-blockers (carvedilol, nebivolol), provided of vasodilation activity, could be particularly indicated for the management of DD. A recent study has tested the ability of nebivolol on 26 patients affected by HF and normal EF, in comparison with the traditional atenolol, combining both invasive hemodynamic and Doppler echocardiographic assessment [78]. After six-month therapy, nebivolol much more than atenolol induced increase of both E/A ratio (from 0.79 ± 0.13 a 0.91 ± 0.11) from 0.84 ± 0.12 a 0.89 ± 0.15) (p < 0.004) and cardiac index and reduction of "wedge" pressures, both at rest and during exercise.
On these grounds, pharmaceutical industry has planned clinical trias to evaluate the prognostic impact of several drugs on diastolic HF. Indeed, the trials completed to date have been disappointing. The CHARM-2 (= Candesartan in Heart Failure – Assessment of Reduction in Mortality) [79] did not evidenced significant improvement of all-cause mortality, of cardiovascular mortality and of hospitalization rate for HF in the sub-set of patients with preserved systolic function, but the follow-up (37.7 months) was probably too short to verify the effects. Into SWEDIC (= Swedish Doppler-Echocardiographic study) [80], carvedilol, inducing a positive influence on transmitral E/A ratio in patients with heart rate > 71 bpm but not in those with HR < 71 bpm, did not exert any effects on the events. Among ongoing trials, the analysis of SENIORS (= Study of the Effects of Nebivolol Intervention on Outcomes and Rehospitalisation in Seniors with heart failure) [81] has not yet performed in the sub-set of patients with normal EF. PEP-CHF (perindopril versus placebo), I-Preserve (Irbesartan versus placebo) study and Hong Kong (rampiril, irbesartan, placebo) study have not completed to date [82].
New therapeutic fields for HF will be opened in view of the associations observed between the state of coronary microcirculation and LV diastolic function [33]. The beneficial effect of ACE-inhibitors on coronary flow reserve, reliable marker of coronary microcirculation function when stenosis of epicardial coronary arteries are not detectable , has been documented in relation to both blood pressure fall and reduction of LV mass. It has recently shown an improvement of transthoracic Doppler-derived coronary flow reserve after only 4-week anti-hypertensive nebivolol therapy, in relation to its endothelium-mediated vasodilation activity [85]. It has to hypothesize that the a restored function of the coronary microcirculation could be useful even for the improvement of DD in hypertensive heart [83,84].
Conclusive implications
DD and diastolic HF are common entities in the clinical practice, particularly in hypertensive population. The diagnosis of diastolic HF can be considered in the presence of the signs of HF and normal EF (50% or more) but it should be usually supported by a Doppler examination. Diastolic HF is associated to four-fold increase mortality. If it is true that the mortality expectation is lower than in patients with systolic HF, it is even true that this difference has a trend to be blunted during long-term follow-up, with possible overlapping after 5.5 years or more. The therapeutic management of diastolic HF is, at least partially, empirical and several studies, ongoing or completed, have been planned to test the effects of ACE-inhibitors, angiotensin-inhibitors and β-blockers. The prevention of diastolic HF may be obtained by a better control of blood pressure values and of concomitant risk factors in hypertensive patients.
Competing interests
Financial competing interests
None.
Non-financial competing interests
None.
==== Refs
Jessup M Brozena S Heart failure N Engl J Med 2003 348 2007 2018 12748317 10.1056/NEJMra021498
Levy D Larson MG Vasan RS Kannel WB Ho KK The progression from hypertension to congestive heart failure JAMA 1996 275 1557 1762 8622246 10.1001/jama.275.20.1557
Levy D Garrison RJ Savage DD Kannel WB Castelli WP Prognostic implications of echocardiographically determined left ventricular mass in the Framingham Heart Study N Engl J Med 1990 322 1561 1566 2139921
Koren MJ Devereux RB Casale PN Savage DD Laragh JH Relation of left ventricular mass and geometry to morbidity and mortality in uncomplicated essential hypertension Ann Intern Med 1991 114 345 352 1825164
Verdecchia P Schillaci G Borgioni C Ciucci A Battistelli M Bartoccini C Santucci A Santucci C Reboldi G Porcellati C Adverse prognostic significance of concentric remodeling of the left ventricle in hypertensive patients with normal left ventricular mass J Am Coll Cardiol 1995 25 871 878 7884090 10.1016/0735-1097(94)00424-O
Inouye I Massie B Loge D Simpson P Tubau JF Abnormal left ventricular filling: an early finding in mild to moderate systemic hypertension Am J Cardiol 1984 53 120 126 6691248 10.1016/0002-9149(84)90695-7
Fouad FM Slominsky JM Tarazi RC Left ventricular diastolic function in hypertension: relation to left ventricular mass and systolic function J Am Coll Cardiol 1984 3 1500 1506 6232306
Solomon SB Nikolic SD Fraser RWM Yellin EL Contraction-relaxation coupling: determination of the onset of diastole Am J Physiol 1999 277 H23 H27 10409177
Gillebert TC Leite-Moreira AF De Hert SG The hemodynamic manifestations of normal myocardial relaxation. A framework from experimental and clinical evaluation Acta Cardiol 1997 52 223 246 9217915
Zile MR Diastolic dysfunction: Detection, consequences and treatment. Part I: Definition and determinants of diastolic function Mod Concepts Cardiovasc Dis 1989 58 67 71
Rokey R Kuo LC Zoghbi WA Limacher MC Quinones MA Determination of parameters of left ventricular diastolic filling with pulsed Doppler echocardiography: comparison with cineangiography Circulation 1985 71 543 550 3971525
Stoddard MF Pearson AC Kern MJ Ratcliff J Mrosek DG Labovitz AJ Left ventricular diastolic function: comparison of pulsed Doppler echocardiographic and hemodynamic indexes in subjects with and without coronary artery disease J Am Coll Cardiol 1989 13 327 336 2913110
Appleton CP Galloway JM Gonzalez MS Gaballa M Basnight MA Estimation of left ventricular filling pressures using two-dimensional and Doppler echocardiography in adult patients with cardiac disease: additional value of analyzing left atrial size, left atrial ejection fraction and the difference in duration of pulmonary venous and mitral flow velocity at atrial contraction J Am Coll Cardiol 1993 22 1972 1982 8245357
Spinale FG Bioactive peptide signaling within the myocardial interstitium and the matrix metalloproteinases Circ Res 2002 91 1082 1084 12480806 10.1161/01.RES.0000047874.80576.5A
Unger T Li J The role of the renin-angiotensin-aldosterone system in heart failure J Renin Angiotensin Aldosterone Syst 2004 5 S7 10 15526242
MacKenna D Summerour SR Villarreal FJ Role of mechanical factors in modulating cardiac fibroblast function and extracellular matrix synthesis Cardiovasc Res 2000 46 257 263 10773229 10.1016/S0008-6363(00)00030-4
Galderisi M Benjamin EJ Evans JC D'Agostino RB Fuller DL Lehman B Wolf PA Levy D Intra- and inter-observer reproducibility of Doppler assessed indexes of left ventricular diastolic function in a population based study (the Framingham Heart Study) Am J Cardiol 1992 70 1341 1346 1442588 10.1016/0002-9149(92)90772-Q
Masuyama T Nagano R Nariyama K Lee JM Yamamoto K Naito J Mano T Kondo H Hori M Kamada T Transthoracic Doppler echocardiographic measurements of pulmonary venous flow patterns: comparison with transesophageal echocardiography J Am Soc Echocardiogr 1995 8 61 69 7710752
Nagueh SF Middleton KJ Kopelen HA Zoghbi WA Quinones MA Doppler tissue imaging: a noninvasive technique for evaluation of left ventricular relaxation and estimation of filling pressures J Am Coll Cardiol 1997 30 1527 1533 9362412 10.1016/S0735-1097(97)00344-6
Garcia MJ Smedira NG Greenberg NL Main M Firstenberg MS Odabashian J Thomas JD Color M-mode Doppler flow propagation velocity is a preload insensitive index of left ventricular relaxation: animal and human validation J Am Coll Cardiol 2000 35 201 208 10636281 10.1016/S0735-1097(99)00503-3
Nishimura RA Tajik J Evaluation of diastolic filling of left ventricle in health and disease: Doppler echocardiography is the clinician's Rosetta Stone J Am Coll Cardiol 1997 30 8 18 9207615 10.1016/S0735-1097(97)00144-7
Pozzoli M Traversi S Cioffi G Stenner R Sanarico M Tavazzi L Loading manipulations improve the prognostic value of Doppler evaluation of mitral flow in patients with chronic heart failure Circulation 1997 5 1222 1230 9054853
Ommen SR Nishimura RA Appleton CP Miller FA Oh JK Redfield MM Tajik AJ Clinical utility of Doppler echocardiography and tissue Doppler imaging in the estimation of left ventricular filling pressures: a comparative simultaneous Doppler catheterization study Circulation 2000 102 1788 1794 11023933
Garcia MJ Ares MA Asher C Rodriguez L Vandervoort P Thomas JD An index of early left ventricular filling that combined with pulsed Doppler peak E velocity may estimate capillary wedge pressure J Am Coll Cardiol 1997 29 448 454 9015003 10.1016/S0735-1097(96)00496-2
Nagueh SF Mikati I Kopelen HA Middleton KJ Quinones MA Zoghbi WA Doppler estimation of left ventricular filling pressure in sinus tachycardia. A new application of Tissue Doppler Imaging Circulation 1998 98 1644 1650 9778330
Nagueh SF Kopelen HA Quinones MA Assessment of left ventricular filling pressures by Doppler in presence of atrial fibrillation Circulation 1996 94 2138 2145 8901664
Vilacosta I San Roman JA Castillo JA Arganda L Rollan MJ Peral V Sanchez-Harguindey L Zarco P Retrograde atrial kick in acute aortic regurgitation. Study of mitral and pulmonary venous flow velocities by transthoracic and transesophageal echocardiography Clin Cardiol 1997 20 35 40 8994736
Shan K Bick RJ Poindexter BJ Shimoni S Letsou GV Reardon MJ Howell JF Zoghbi WA Nagueh SF Relation of tissue Doppler derived myocardial velocities to myocardial structure and beta-adrenergic receptor density in humans J Am Coll Cardiol 2000 36 891 896 10987616 10.1016/S0735-1097(00)00786-5
Zile MR Brutsaert DI New concepts in diastolic dysfunction and diastolic heart failure Circulation 2002 105 1387 1393 11901053 10.1161/hc1102.105289
Spirito P Maron BJ Relation between extent of left ventricular hypertrophy and diastolic filling abnormalities in hypertrophic cardiomyopathy J Am Coll Cardiol 1990 15 808 813 2307791
Klein AL Hatle LK Taliercio CP Taylor CL Kyle RA Bailey KR Seward JB Tajik AJ Serial Doppler echocardiographic follow-up of left ventricular diastolic function in cardiac amyloidosis J Am Coll Cardiol 1990 16 1135 1141 2229760
Biondi B Fazio S Palmieri EA Carella C Panza N Cittadini A Bone F Lombardi G Sacca Left ventricular diastolic dysfunction in patients with subclinical hypothyroidism J Clin Endocrinol Metab 1999 84 2064 2067 10372711 10.1210/jc.84.6.2064
Galderisi M Cicala S Caso P De Simone L D'Errico A Petrocelli A de Divitiis O Coronary flow reserve and myocardial diastolic dysfunction in arterial hypertension Am J Cardiol 2002 90 860 864 12372574 10.1016/S0002-9149(02)02708-X
Mureddu GF de Simone G Greco R Rosato GF Contaldo F Left ventricular filling in arterial hypertension. Influence of obesity and hemodynamic and structural confounders Hypertension 1997 29 544 550 9040436
Ferrannini E Buzzigoli G Bonadonna R Giorico MA Oleggini M Graziadei L Pedrinelli R Brandi L Bevilacqua S Insulin resistance in essential hypertension N Engl J Med 1987 317 350 357 3299096
Galderisi M Paolisso G Tagliamonte MR Alfieri A Petrocelli A de Divitiis M Varricchio M de Divitiis O Is insulin action a determinant of left ventricular relaxation in uncomplicated essential hypertension? J Hypertens 1997 15 745 50 9222942 10.1097/00004872-199715070-00006
Draznin B Sussman KE Eckel RH Kao M Yost T Sherman NA Possible role of cytosolic free calcium concentrations in mediating insulin resistance of obesity and hyperinsulinemia J Clin invest 1988 82 1848 1852 3143744
Galderisi M Tagliamone MR D'Errico A Carella C Varricchio G Mondillo S de Divitiis O Paolisso G Independent association of plasma leptin levels and left ventricular isovolumic relaxation in uncomplicated hypertension Am J Hypertens 2001 14 1019 1024 11710780 10.1016/S0895-7061(01)02137-9
Liu JE Palmieri V Roman MJ Bella JN Fabsitz R Howard BV Welty TK Lee ET Devereux RB The impact of diabetes on left ventricular filling pattern in normotensive and hypertensive adults: the Strong Heart Study J Am Coll Cardiol 2001 37 1943 1949 11401136 10.1016/S0735-1097(01)01230-X
White WB Schulman P Dey HM Katz AM Effects of age and 24-hour ambulatory blood pressure on rapid left ventricular filling Am J Cardiol 1989 63 1343 1347 2729106 10.1016/0002-9149(89)91046-1
Verdecchia P Schillaci G Guerrieri M Boldrini F Gatteschi C Benemio G Porcellati C Prevalence and determinants of left ventricular diastolic filling abnormalities in an unselected hypertensive population Eur Heart J 1990 11 679 691 2144489
Aeschbacher BC Hutter D Fuhrer J Weidmann P Delacretaz E Allemann Y Diastolic dysfunction precedes myocardial hypertrophy in the development of hypertension Am J Hypertens 2001 14 106 113 11243300 10.1016/S0895-7061(00)01245-0
Galderisi M Petrocelli A Alfieri A Garofalo M de Divitiis O Impact of ambulatory blood pressure on left ventricular diastolic dysfunction in uncomplicated arterial systemic hypertension Am J Cardiol 1996 77 597 601 8610609 10.1016/S0002-9149(97)89313-7
Palmieri V Wachtell K Gerdts E Bella JN Papademetriou V Tuxen C Nieminen MS Dahlof B de Simone G Devereux RB Left ventricular function and hemodynamic features of inappropriate left ventricular hypertrophy in patients with systemic hypertension: the LIFE study Am Heart J 2001 141 784 791 11320367 10.1067/mhj.2001.114803
de Simone G Verdecchia P Pede S Gorini M Maggioni AP Prognosis of inappropriate left ventricular mass in hypertension: the MAVI Study Hypertension 2002 40 470 476 12364349 10.1161/01.HYP.0000034740.99323.8A
Lubien E DeMaria A Krishnaswamy P Clopton P Koon J Kazanegra R Gardetto N Wanner E Maisel AS Utility of B-natriuretic peptide in detecting diastolic dysfunction: comparison with Doppler velocity recordings Circulation 2002 105 595 601 11827925 10.1161/hc0502.103010
Yamaguchi H Yoshida J Yamamoto K Sakata Y Mano T Akehi N Hori M Lim YJ Mishima M Masuyama T Elevation of plasma brain natriuretic peptide is a hallmark of diastolic heart failure independent of ventricular hypertrophy J Am Coll Cardiol 2004 43 55 60 14715184 10.1016/j.jacc.2003.07.037
Galderisi M Caso P Severino S Petrocelli A De Simone L Izzo A Mininni N de Divitiis O Myocardial diastolic impairment caused by left ventricular hypertrophy involves basal septum more than other walls: analysis by pulsed Doppler tissue imaging J Hypertens 1999 17 685 693 10403613 10.1097/00004872-199917050-00013
Zile MR Heart failure with preserved ejection fraction: Is this diastolic heart failure? J Am Coll Cardiol 2003 41 1519 1522 12742292 10.1016/S0735-1097(03)00186-4
de Simone G Kitzman DW Chinali M Oberman A Hopkins PN Rao DC Arnett DK Devereux RB Left ventricular concentric geometry is associated with impaired relaxation in hypertension: the HyperGEN study Eur J Heart 2004 30 1 7
Gandhi SK Powers JC Nomeir AM Fowle K Kitzman DW Rankin KM Little WC The pathogenesis of acute pulmonary edema associated with hypertension N Engl J Med 2001 344 17 22 11136955 10.1056/NEJM200101043440103
Vasan RS Levy D Defining diastolic heart failure. A call for standardized diagnostic criteria Circulation 2000 101 2118 2121 10790356
European Study Group on Diastolic Heart Failure How to diagnose diastolic heart failure Eur Heart J 1998 19 990 1003 9717033 10.1053/euhj.1998.1057
Zile MR Baicu CF Gaasch WH Diastolic heart failure. Abnormalities in acrive relaxation and passive stiffness of the left ventricle N Engl J Med 2004 350 1953 1959 15128895 10.1056/NEJMoa032566
Zile MR Gaasch WH Carroll JD Feldman MD Aurigemma GP Schaer GL Ghali JK Liebson PR Heart failure with a normal ejection fraction: is measurement of diastolic function necessary to make the diagnosis of diastolic heart failure? Circulation 2001 104 779 782 11502702
Benjamin EJ Levy D Anderson KM Wolf PA Plehn JF Evans JC Comai K Fuller DL Sutton MS Determinants of Doppler indexes of left ventricular diastolic function in normal subjects (the Framingham Heart Study) Am J Cardiol 1992 70 508 515 1642190 10.1016/0002-9149(92)91199-E
Galderisi M Benjamin EJ Evans JC D'Agostino RB Fuller DL Lehman B Levy D Impact of heart rate and PR interval on Doppler indexes of left ventricular diastolic filling in an elderly cohort (the Framingham Heart Study) Am J Cardiol 1993 72 1183 1187 8237811 10.1016/0002-9149(93)90991-K
Zile MR Baicu CF Mann D Alterations in ventricular function: diastolic heart failure "Heart Failure, A Companion to Braunwald's Heart Disease" 2004 Saunders
Redfield MM Jacobsen SJ Burnett JC JrMahoney DW Bailey KR Rodeheffer RJ Burden of systolic and diastolic ventricular dysfunction in the community: appreciating the scope of the heart failure epidemic JAMA 2003 289 194 202 12517230 10.1001/jama.289.2.194
Khouri SJ Maly GT Suh DD Walsh TE A practical approach to the echocardiographic evaluation of diastolic function J Am Soc Echocardiogr 2004 17 290 297 14981433
de Simone G Greco R Mureddu GF Romano C Guida R Celentano A Contaldo F Relation of left ventricular diastolic properties to systolic function in arterial hypertension Circulation 2001 101 152 157 10637202
Yp G Wang M Zhang Y Fung JW Ho PY Sanderson JE Left ventricular long axis function in diastolic heart failure is reduced in both diastole and systole: time for a redefinition? Heart 2002 87 121 125 11796546 10.1136/heart.87.2.121
Yu CM Lin H Yang H Kong SL Zhang Q Lee SW Progression of systolic abnormalities in patients with "isolated" diastolic heart failure and diastolic dysfunction Circulation 2002 105 1195 1201 11889013 10.1161/hc1002.105185
Chen QM Li W O'Sullivan C Francis DP Gibson D Henein MY Clinical in vivo calibration of pulse wave tissue Doppler velocities in the assessment of ventricular wall motion. A comparison study with M-mode echocardiography Int J Cardiol 2004 97 289 295 15458697 10.1016/j.ijcard.2004.03.048
Caruana L Petrie MC Davie AP McMurray JJ Do patients with suspected heart failure and preserved left ventricular systolic function suffer from "diastolic heart failure" or from misdiagnosis? A prospective descriptive study BMJ 2000 321 215 218 10903655 10.1136/bmj.321.7255.215
Vasan RS Benjamin EJ Levy D Prevalence, clinical features and prognosis of diastolic heart failure: an epidemiologic perspective J Am Coll Cardiol 1995 26 1565 1574 7594087 10.1016/0735-1097(95)00381-9
Vasan RS Larson MG Benjamin EJ Evans JC Reiss CK Levy D Congestive heart failure in subjects with normal versus reduced left ventricular ejection fraction: prevalence and mortality in a population-based cohort J Am Coll Cardiol 1999 33 1948 1955 10362198 10.1016/S0735-1097(99)00118-7
Hogg K Swedberg K McMurray J Heart failure with preserved left ventricular systolic function; epidemiology, clinical characteristics, and prognosis J Am Coll Cardiol 2004 43 317 327 15013109 10.1016/j.jacc.2003.07.046
Tarantini L Faggiano P Senni M Lucci D Bertoli D Porcu M Opasich C Tavazzi L Maggioni AP Clinical features and prognosis associated with a preserved left ventricular systolic function in a large cohort of congestive heart failure outpatients managed by cardiologists. Data from the Italian Network on Congestive Heart Failure Ital Heart J 2002 3 656 664 12506524
Philbin EF Rocco TA JrLindenmuth NW Ulrich K Jenkins PL Systolic versus diastolic heart failure in community practice: clinical features, outcomes, and the use of angiotensin-converting enzyme inhibitors Am J Med 2000 109 605 613 11099679 10.1016/S0002-9343(00)00601-X
Senni M Tribouilloy CM Rodeheffer RJ Jacobsen SJ Evans JM Bailey KR Redfield MM Congestive heart failure in the community: a study of all incident cases in Olmsted County, Minnesota, in 1991 Circulation 1998 98 2282 2289 9826315
Badano LP Albanese MC De Biaggio P Rozbowsky P Miani D Fresco C Fioretti PM Prevalence, clinical characteristics, quality of life, and prognosis of patients with congestive heart failure and isolated left ventricular diastolic dysfunction J Am Soc Echocardiogr 2004 17 253 261 14981424
Schillaci G Pasqualini L Verdecchia P Vaudo G Marchesi S Porcellati C de Simone G Mannarino E Prognostic significance of left ventricular diastolic dysfunction in essential hypertension J Am Coll Cardiol 2002 39 2005 2011 12084601 10.1016/S0735-1097(02)01896-X
Bella JN Palmieri V Roman MJ Liu JE Welty TK Lee ET Fabsitz RR Howard BV Devereux RB Mitral ratio of peak early to late diastolic filling velocity as a predictor of mortality in middle-aged and elderly adults. The Strong Heart Study Circulation 2002 105 1928 1933 11997279 10.1161/01.CIR.0000015076.37047.D9
Vasan RS Larson MG Levy D Galderisi M Wolf PA Benjamin EJ Doppler transmitral flow indexes and risk of atrial fibrillation (The Framingham Heart Study) Am J Cardiol 2003 91 1079 1083 12714150 10.1016/S0002-9149(03)00152-8
Brilla CG Matsubara LS Weber KT Antifibrotic effects of spironolactone in preventing myocardial fibrosis in systemic arterial hypertension Am J Cardiol 1993 71 12A 16A 8421998 10.1016/0002-9149(93)90239-9
Grandi AM Imperiale D Santillo R Barlocco E Bertolini A Guasti L Venco A Aldosterone antagonist improves diastolic function in essential hypertension Hypertension 2002 40 647 652 12411457 10.1161/01.HYP.0000036399.80194.D8
Nodari S Metra M Dei Cas L Beta-blocker treatment of patients with diastolic heart failure and arterial hypertension. A prospective, randomized, comparison of the long-term effects of atenolol vs. nebivolol Eur J Heart Fail 2003 5 621 627 14607201 10.1016/S1388-9842(03)00054-0
Solomon SD Wang D Finn P Skali H Zornoff L McMurray JJ Swedberg K Yusuf S Granger CB Michelson EL Pocock S Pfeffer MA Effect of candesartan on cause-specific mortality in heart failure patients: the Candesartan in Heart failure Assessment of Reduction in Mortality and morbidity (CHARM) program Circulation 2004 110 2180 2183 15466644 10.1161/01.CIR.0000144474.65922.AA
Bergstrom A Andersson B Edner M Nylander E Persson H Dahlstrom U Effect of carvedilol on diastolic function in patients with diastolic heart failure and preserved systolic function. Results of the Swedish Doppler-echocardiographic study (SWEDIC) Eur J Heart Fail 2004 6 453 461 15182771 10.1016/j.ejheart.2004.02.003
Shibata MC Flather MD Bohm M Borbola J Cohen-Solal A Dumitrascu D Ferrari R Lechat P Parkhomenko A Soler-Soler J Tavazzi L Toman J Van Veldhuisen DJ Coats AJ Poole-Wilson P Study of the Effects of Nebivolol Intervention on Outcomes and Rehospitalisation in Seniors with Heart Failure. Rehospitalisation in Seniors with Heart Failure (SENIORS). Rationale and design Int J Cardiol 2002 86 77 85 12243852 10.1016/S0167-5273(02)00321-2
Banerjee P Banerjee T Khand A C lark AL Cleland JG Diastolic heart failure: neglected or misdiagnosed? J Am Coll Cardiol 2002 39 138 141 11755299 10.1016/S0735-1097(01)01704-1
Motz W Strauer BE Improvement of coronary flow reserve after long-term therapy with enalapril Hypertension 1996 27 1031 1038 8621193
Schwartzkopff B Brehm M Mundhenke M Strauer BE Repair of coronary arterioles after treatment with perindopril in hypertensive heart disease Hypertension 2000 36 220 225 10948081
Galderisi M Cicala S D'Errico A de Divitiis O de Simone G Nebivolol improves coronary flow reserve in hypertensive patients without coronary heart disease J Hypertens 2004 22 2201 2208 15480106 10.1097/00004872-200411000-00024
| 15807887 | PMC1087861 | CC BY | 2021-01-04 16:38:31 | no | Cardiovasc Ultrasound. 2005 Apr 4; 3:9 | utf-8 | Cardiovasc Ultrasound | 2,005 | 10.1186/1476-7120-3-9 | oa_comm |
==== Front
Curr Control Trials Cardiovasc MedCurrent Controlled Trials in Cardiovascular Medicine1468-67081468-6694BioMed Central 1468-6708-6-61581397510.1186/1468-6708-6-6ReviewBeyond the Evidence of the New Hypertension Guidelines. Blood pressure measurement – is it good enough for accurate diagnosis of hypertension? Time might be in, for a paradigm shift (I) Pater Cornel [email protected] Chippenham, UK2005 6 4 2005 6 1 6 6 25 1 2005 6 4 2005 Copyright © 2005 Pater; licensee BioMed Central Ltd.2005Pater; 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.
Despite widespread availability of a large body of evidence in the area of hypertension, the translation of that evidence into viable recommendations aimed at improving the quality of health care is very difficult, sometimes to the point of questionable acceptability and overall credibility of the guidelines advocating those recommendations.
The scientific community world-wide and especially professionals interested in the topic of hypertension are witnessing currently an unprecedented debate over the issue of appropriateness of using different drugs/drug classes for the treatment of hypertension. An endless supply of recent and less recent "drug-news", some in support of, others against the current guidelines, justifying the use of selected types of drug treatment or criticising other, are coming out in the scientific literature on an almost weekly basis. The latest of such debate (at the time of writing this paper) pertains the safety profile of ARBs vs ACE inhibitors.
To great extent, the factual situation has been fuelled by the new hypertension guidelines (different for USA, Europe, New Zeeland and UK) through, apparently small inconsistencies and conflicting messages, that might have generated substantial and perpetuating confusion among both prescribing physicians and their patients, regardless of their country of origin.
The overwhelming message conveyed by most guidelines and opinion leaders is the widespread use of diuretics as first-line agents in all patients with blood pressure above a certain cut-off level and the increasingly aggressive approach towards diagnosis and treatment of hypertension. This, apparently well-justified, logical and easily comprehensible message is unfortunately miss-obeyed by most physicians, on both parts of the Atlantic.
Amazingly, the message assumes a universal simplicity of both diagnosis and treatment of hypertension, while ignoring several hypertension-specific variables, commonly known to have high level of complexity, such as:
- accuracy of recorded blood pressure and the great inter-observer variability,
- diversity in the competency and training of diagnosing physician,
- individual patient/disease profile with highly subjective preferences,
- difficulty in reaching consensus among opinion leaders,
- pharmaceutical industry's influence, and, nonetheless,
- the large variability in the efficacy and safety of the antihypertensive drugs.
The present 2-series article attempts to identify and review possible causes that might have, at least in part, generated the current healthcare anachronism (I); to highlight the current trend to account for the uncertainties related to the fixed blood pressure cut-off point and the possible solutions to improve accuracy of diagnosis and treatment of hypertension (II).
==== Body
Introduction and magnitude of the background problem
Recent changes in definition and classification of blood pressure levels make hypertension, by far, the most commonly diagnosed condition in primary and secondary healthcare systems and projects the entity on the first place in terms of work load and prescribing cost.
"People with normal blood pressure by their 50 years of age are considered to run a 90% life-time risk for developing hypertension later during their lives"[1].
This statement puts in perspective the epidemic nature of hypertension and the growing concern of all societies in dealing with this outstanding public health problem, in developed as well as in developing countries.
There is little doubt that the American [1], the European [2], the British [3] and the WHO [4] guidelines on hypertension have the same, common goal of improving the quality of health care by changing the behaviour of providers and by improving the effectiveness of hypertension management in daily practice. Overall, current guidelines have become larger documents, apparently more comprehensive and increasingly evidence-based. Despite these obvious improvements, guidelines are hardly, if at all, implemented in clinical practice.
The demographic and socio-economic profile of hypertension
The National Health and Nutritional Health Survey (NHANES) [5] data from 1999 to 2000 reported a 3.7-percentage point increase in the hypertension prevalence rate with more than 42% of hypertensives being not at all treated, almost 30% of them being unaware of their illness and 69% not being controlled. Reported control rates are even lower in European countries, only 8% on average [6]. Approximately 75 million adults (34% of the whole population) have blood pressure above 140/90 mmHg in five European countries (UK, Germany, France, Italy and Sweden) [7].
These disappointing figures are in sharp contradiction with the hypertension-related successes over several years before the survey.
Even worse, the recently published results of an epidemiologic inquiry to assess the hypertension burden and overall prevalence for the same period (1999 to 2000) in US [8] showed that, in fact, the total hypertension prevalence rate reached 31.3% amounting to at least 65 million adults in US having hypertension. This is an almost 30% increase of the frequently quoted figure for the magnitude of hypertension in terms of prevalence based on the NHANES III estimation of "50 million adults" with hypertension in the US [8].
Besides, the newly minted prehypertension category in the JNC VII-hypertension classification [1], introduced on fully justifiable reasons, ads further to the hypertension-related public health burden. In a sample of 3,488 persons of the same NHANES III survey, among people aged 20 years or older, 29% were hypertensive, 31% were prehypertensive and 39% were normotensive, with considerable greater percent of prehypertensives in men than in women (39.0% versus 23.1%) and in blacks as compared with whites (37.4% versus 32.2%) [9,10]. About 59 million American adults (29%) fall into the prehypertension category.
These data suggest that more than 60% of the American adults are either hypertensive or prehypertensive. Noticeably, the prehypertesives were 1.65 times more likely to have at least one other major risk factor (total cholesterol or overweight) than were those with normotension. Furthermore, in a simulation applied to a sample of 10,000 adults aged 25–74 years, the considerably high prevalence of prehypertensives in the higher age groups (two-thirds among people aged 45–65 years and 80% of those aged 65–74 years) indicated that prehypertension might account for 3.4% of hospitalisation, 6.5% of nursing home stays, and 9.1% of deaths [11].
These health-related alarming trends are ascribed to the increasing prevalence of obesity in the general population as well as to the aging, respectively, to the growing segment of elderly in the general population.
Interestingly, the two latter demographic aspects are results of increased life expectancy [12-14] in the context of "healthier lifestyles and/or better control rates associated with hypertension-related public health effectiveness and medical care quality improvement", a mix of factors termed: "higher control-survival-burden paradox" [5].
Paradoxically indeed, improved hypertension-focused medical care has simultaneously increased the burden on public health systems worldwide! Estimating further this burden in the American-context, the 23 million hypertensive adults estimated to take antihypertensive medication generate costs of about $15bn (€12bn), i.e., 10% of the country's total spending on drugs [15].
Estimated total direct and indirect cost of hypertension for 2005, as a result of its increased prevalence, is around $60 billion while, the total cost for cardiovascular disease and stroke for 2005 is estimated to be $393.5 billion [16].
According to the same source [16], hypertension was listed as a primary or contributing cause of death in about 261,000 of about 2.4 million US deaths in 2002, representing a 57% increase in death from hypertension over the past few decades.
The projected addition of some 30 million individuals aged 60 years or older in the next 20 years [17] as well as the widespread consensus and willingness to increase the number of those effectively treated for their hypertension, emphasizes the potential for escalating costs that may blow out of proportion and the seriousness of the hypertension-epidemic.
In this context, it is obvious that there is a tremendous need for wide-scope, systematically driven cost-effective prevention and treatment, and for more effective control strategies at all levels of healthcare systems.
However, a holistic, comprehensive strategic approach must not only target hypertension as pathological entity, but must take into account the wider environment in which hypertension is a major risk factor for cardiovascular disease and its interplay in the constellation of other, well-known modifiable risk factors such as: tobacco use, hypercholesterolemia, overweight/obesity, physical inactivity, diet and, to great extent and more often, associated diabetes mellitus [18-20].
Diagnosing hypertension – sources of errors in blood pressure measurement
Blood pressure measurement is by far, the most commonly performed screening tests in medical practice and, because the act of measuring blood pressure is perceived as simple and straightforward it has also become the most commonly used "in house" self-employed test.
For only 20 years ago, the task of people willing to monitor their blood pressure was considerably easier, at least as far as the reference value for normal/abnormal blood pressure was concerned. It could simply be derived by everyone just by adding his/her age to 100 and consider the resultant value as cut-off point for threshold to abnormality. Advice from physician was eventually sought as to whether drug treatment was appropriate or not.
However, gradually, the age-standardized BP became obsolete and things started to be more complicated, not only for lay people interested in monitoring themselves but also for physicians and nurses. Accuracy of measurement itself has become an issue while adjustment to ever changing "target cut-off points" (reference ranges) is widely required.
So, the apparent user-friendliness of the blood pressure measurement techniques is made more questionable by the evolving changes in the definition, classification, diagnosis and management recommendations in the area of hypertension. As a consequence, both professionals as well a lay people wishing to continue to measure blood pressure are currently supposed to comply with guidelines for blood pressure measurement [21](see Additional file 1), if they are to believe what they are measuring.
Observer bias
Despite the clear guidelines on blood pressure measurement technique, there seems to be large inter-observer variations, both among nursing staff and physicians as well as between the two groups. A questionnaire meant to focus this issue, encompassing 28 senior nurses and 55 health professionals from 28 different clinics in UK, highlighted considerable such variation leading consequently to inappropriate action [22-24]. A similar questionnaire carried out in Sao Paulo State on 105 professionals as to their compliance with the blood pressure measurement recommendations by Perloff et al. [21], showed that nursing staff abided by 40% of the recommended procedures while medicine teachers, physicians and residents abided by approximately 70% [25].
Conventionally, differences of 5 mmHg are considered clinically significant although, results derived from more recent trials such as VALUE (26) and ALLHAT [27] indicated that blood pressure differences of 2 to 4 mmHg might be clinically important. In this context, errors in blood pressure measurement in excess of 15 mmHg or more, as reported by Campbell et al. [28], are obviously leading to misclassification of hypertensive patients and inappropriate treatment.
What is more worrying, while in the same time at least partially explaining why such huge discrepancies may occur in practice, is the fact that gaps in the basic theoretic and practical knowledge seem to be common among interns and first-year family practice residents, supposed to have just acquired the skills for accurate blood pressure measurement [29]. An interesting observational study in this respect, carried out at the Westminster Medical School in London, showed that 33% out of 80 doctors in training grades/junior hospital doctors, acknowledged no formal education on how to measure blood pressure, a finding confirmed further by the poor accuracy in blood pressure measurement displayed by one third of the study group [30].
The human error impact is further compounded by variables such as: cuff selection and application, incorrect cuff positioning and rapid cuff deflation rate [20,31], inadequate rest period [31], digit bias and lack of repeated measurements (see Additional file 2).
Faulty equipment
Adding considerably to the degree of "human error" in the area of blood pressure measurement is the universally poor state of the measurement devices available, their inaccuracy and the unreliability of the measurement results generated.
A study, aimed to assess the accuracy of calibration and evaluation of physical condition of 524 sphygmomanometers (mercury and aneroid) used in hospital settings or private medical offices, showed that 44% of the aneroid sphygmomanometers used in hospital settings and 61% of those used in private medical practices were found inaccurate. The magnitude of inaccuracy ranged from 4–6 mmHg in 32% to more than 13 mmHg in 7% of the tested devices. It was concluded that the mercury and aneroid sphygmomanometers showed inaccuracies of 21% versus 58% and unreliability of 64% versus 70%, respectively [32].
Another study, aimed at the assessment of maintenance and calibration of sphygmomanometers used in 231 English general practices, showed that 9.2% of the 1,462 sphygmomanometers tested gave readings that were more that 5 mmHg inaccurate and that one in 54 practices had an arrangement for maintenance and calibration of the measurement devices [33].
A British study carried out in 18 practices and 67 GP offices showed digit bias in systolic and diastolic readings to the nearest 10 mmHg [34].
Lack of standardized blood pressure measurement devices
Because of its accuracy and reliability, the mercury sphygmomanometer is generally regarded as the gold standard against which all other devices for blood pressure measurement should be compared [35]. Unfortunately, due to the widespread concern that the mercury contaminates the environment, the mercury sphygmomanometers are about to be replaced largely with alternative equipment [36].
In contrast to the mercury sphygmomanometer that is least dependent on calibration and maintenance [37], the aneroid devices need calibration against a known standard (mercury manometer or non-mercury pressure meters) at six months interval. Failed calibration test implies the need to return the device to the manufacturer.
Lack of calibration and maintenance of the aneroid sphygmomanometers, as reported by numerous studies carried out in different parts of the world, makes them highly doubtful for routine use in medical practice, unless aggressive programs of maintenance and calibration will be implemented in timely manner in order to overcome the problems associated with inaccurate measurement of blood pressure [38].
A second type of alternative blood pressure measurement device is the electronic automated manometer. Such a device assesses the oscillations of pressure in the cuff during deflation. The point of maximal oscillation corresponds to the mean intra-arterial pressure. The systolic and diastolic pressures are computed on the basis of an algorithm, commonly known to be patent protected. This means, in turn, that algorithms used by different manufacturers vary from device to device [39]. Besides, these devices display a particular vulnerability in certain clinical circumstances, such as patients with arrhythmias and with stiff arteries. Likewise, they are prone to error because of the dependence of the cuff deflation rate to the heart rate. If the deflation is too rapid or the heart rate is too slow (e.g., in patients treated with beta-blockers), inaccurate blood pressure measurement is likely to occur [40].
Allegedly, most of these blood pressure measurement devices are subject to validation according to internationally accepted protocols of the Association for the Advancement of Medical Administration (AAMI) or the British Hypertension Society (BHS). The latest validation protocol has been issued by the European Society of Hypertension, slightly modified as compared to the two protocols mentioned above [41].
Disappointingly, however, commercially available automated devices that have passed validating study protocols still display significant inaccuracies in blood pressure measurement. Gerin et al. showed that blood pressure measurements with such devices were inaccurate by at least 5 mmHg in 20 to 38% of the individuals tested [42]. Schwartz et al. state that when using blood pressure monitors that meet the AAMI and BHS validation criteria, more than 50% of the persons tested may have average measurements that are in error by more than 5 mmHg [43].
Even worse is the fact that many of the most commonly used blood pressure monitors in clinical practice, in US [44], have never passed an AAIM certification and do not even need to do so as there are no regulatory requirements in this respect.
Blood pressure variability and its assessment by ABPM
Assuming that successful precaution measures would be taken with regard to all the above mentioned sources of errors, the office blood pressure measurement would still not be an accurate reflection of an individual subject's true blood pressure and of its impact on the on that subject's long term outcome.
Apart from the imperfection of the Korotkoff technique as compared with the intra-arterial blood pressure measurements [45], just adding to the aforementioned potential sources of error, the great diurnal variability inherent in the blood pressure behaviour (see Additional file 3) is perhaps the major issue that makes office blood pressure measurement a screening test at best, and unqualifies it as a diagnostic test.
Ambulatory blood pressure measurement (ABPM), now in use since more than 25 years, has been gaining wide acceptance during the past few years. Current guidelines recommend use of the method in selected cases (see Additional file 4) [46-49], however, in the light of more recent data [50,51,63], ABPM is about to emerge as the method of choice for ensuring an accurate "first time" hypertension diagnosis, for refining cardiovascular risk stratification in most of the newly-diagnosed hypertensives and for the ability to confidently start or postpone pharmacologic treatment in a particular individual. Furthermore, ABPM allows for assessment of prognosis and therapeutic guidance.
A 24-hour ambulatory blood pressure curve (see Fig. 1) fully illustrates the large, diurnal intraindividual variability of the blood pressure and suggests that great differences may be encountered when comparing such data with the data collected from single or a few readings in the physician's office.
Figure 1 Blood pressure variability in healthy individual recorded by 24-hour ABPM.
The great uncertainty that lies in casual blood pressure measurements, reflecting the biologic variations of this parameter, has been emphasized through Bayesian analysis [52] and a prospective trial of home blood pressure monitoring [53]. These studies demonstrated that 11 respectively 15 blood pressure readings generated mean values that accounted for 80% of the variation in any one single measure. Likewise, blood pressure readings performed on different days display large variability as reflected by mean standard deviations as high as 12/8 mmHg [54]. The regression to mean phenomenon is known to further dilute blood pressure measurement results.
The total range of variation observed for 24-hour mean in healthy pregnant women is even larger; 28 and 26 mmHg for SBP and DBP, respectively. This range of blood pressure variability is approximately three times larger when computed on the basis of individual blood pressure measurements instead of the 24-hour mean [55].
It has been estimated that use of a single blood pressure measurement to assist in diagnostic decision making would overdiagnose hypertension in 20–30% of cases and miss a third of those who truly have the disease [56,57].
The white coat effect, defined as an office blood pressure exceeding mean daytime ambulatory pressure by at least 20 mmHg systolic and/or 10 mmHg diastolic has been found in as many as 73% of treated hypertensive subjects. It may occur more frequently in women than in men [58] and it is virtually impossible to be diagnosed on clinical examination alone [59].
White-coat hypertension (defined as high blood pressure only in the medical care environment) is reported in as many as 20 to 35% of patients in whom hypertension is diagnosed [60,61] and in nearly 30% of pregnant women [62]. ABPM in elderly of Syst-Eur trial showed that blood pressure was on average 22 mmHg higher on conventional than on daytime ambulatory measurement [63]. Likewise, the ABPM in the same trial revealed that older patients have a general propensity toward hypotensive states of different aetiologies reflecting considerable diurnal blood pressure variability, with periods of hypotension alternating with hypertension. This translates into the well-known elderly's greater susceptibility to adverse drug events and calls for accurate diagnosis and appropriately tailored treatment.
Newer insights in this area, further emphasizing the tremendous complexity of hypertension per se, comes from a study comparing the long-term, office blood pressure measurements with 24-hour ambulatory blood pressure measurement in treated hypertensive patients [64].
The study by Clement et al. [64] showed that when office and ambulatory blood pressure are compared as to their impact on the long-term prognosis, ambulatory blood pressure predicts prognosis significantly better, even after adjustment for any associated risk factors.
Among several valuable findings of this study, the most notable is that patients whose mean 24-hour systolic pressure was 135 mmHg or higher, when they were receiving treatment, were nearly twice as likely to have a cardiovascular event as patients with a mean 24-hour systolic pressure of less than 135 mmHg, regardless of their office blood pressure values.
The use of 24-hour ambulatory blood pressure measurement seems to be a sine qua non condition for accurate diagnosis in cases with hypertension recently termed reversed or masked hypertension [65,66], in which the office pressure used to be lower than the ambulatory pressure. It was estimated that only in US might be as many as 10 million subjects having this type of hypertension [43]. The new entity requires special attention to be given in two different clinical scenarios:
1). Patients with high office pressure but with low ambulatory pressure (no more likely than normotensive persons to have a cardiovascular event); intensive treatment of these patients, exclusively on the basis of office blood pressure measurement and/or their labelling as having "resistant hypertension" (in absence of associated risk factors or established organ damage), is likely to be deleterious and, it may even lead to adverse drug events;
2). Patients with low office blood pressure but high ambulatory blood pressure (who are in the opposite extreme of the hypertension spectrum); apparently, these patients are "well controlled", however, in need of intensified treatment as they seem to be running a worse long-term prognosis.
Consequences of errors in blood pressure measurement
The multiple sources of errors that may be encountered in the office blood pressure measurement emphasize the great degree of non-accurate, misleading results generated by such assessments in daily clinical practice. Two examples bellow highlight the consequences of only minimal biased assessments (e.g., ± 5 mmHg).
An overestimating systematic error of 5 mmHg would misclassify some 27 million people as being hypertensive rather than having normal-high blood pressure, i.e., prehypertension [67]. The individual and societal consequences of initiating drug treatment in this segment of the population might be huge and, eventually translate into a real public health burden. Namely, patients with a low absolute risk may be exposed to the potential side effects of a treatment for little or no therapeutic benefit.
Further, a systematic error of underestimating true blood pressure by 5 mmHg would mean that 21 million people would go untreated as their real hypertension would be labelled as normal-high blood pressure [67]. In this scenario, patients with high absolute risk who are misclassified as having controlled hypertension will have a higher risk of cardiovascular event than necessary.
Both these scenarios may occur with rather high probability in clinical practice given the current, widespread model of healthcare based on a dichotomous paradigm of "yes/no" decision making as to establishing an initial diagnosis, as to the need to investigate or not investigate or, to treat or not to treat, currently arbitrarily fixed for hypertension at 140/90 mmHg.
Amazingly, the scientific community seems to accept the particular outlook reflected by this dichotomous reasoning despite widespread awareness that the risk associated with increasing blood pressure is graded and continuous. It begins as low as at 115/75 mmHg [1] and increases gradually without, however, a particular threshold being known on the BP curve that might discriminate between risk/no risk circumstances [68].
The lack of such a threshold combined with the information derived from a multitude of clinical trials, convincingly demonstrating the benefits of treatment across all levels of blood pressure in Western populations (not only in hypertension) [69-71], have generated "the lower the blood pressure, the better" – treatment philosophy, applicable to the general population and advocated as such by the majority of current guidelines.
The <140/90 mmHg cut-off point (respectively <130/80 mmHg for those with compelling indications) have been chosen as treatment goal/target for pharmacologically treated hypertensives. These targets are supposed to be pursued aggressively with even three or four drugs if needed, in order to achieve "optimal" or "normal" BP in young, middle-aged, or diabetic subjects (below 130/80 mmHg), and at least "high-normal" BP in elderly patients (i.e., <140/90 mmHg) [72,73].
Unfortunately, current evidence indicates that most patients fail to achieve systolic blood pressure below 140 mmHg [74-76]. According to a survey of hypertension in 10 countries worldwide, the proportion of patients achieving a blood pressure target below 140/90 mmHg ranged from a maximum of 27% in the USA to a minimum of less than 3% in Zaire [77]. In clinical trials focusing hypertension treatment the control rate is around 6% [78].
Mancia and Grassi [79] pointed out that even in the case of ideal scenario whereby "patient compliance and physician's expertise were ensured, attaining systolic blood pressure control would neither frequently nor easily be obtained".
Additionally, common sense suggests that there may be a level at which the benefits of treatment are outweighed by its side effects [80]. Likewise, it is obvious that the absolute benefits gained differ substantially between elderly and middle aged people and those with or without pre-existing cardiovascular disease [81,82]. Appropriate assessment of these variables is a matter of judgement of the treating physician, based not on occasional blood pressure measurements but rather, on complex clinical decision making employed on a case-to-case basis [83-88].
The aforementioned major problems with accurately measuring blood pressure in day-to-day clinical practice for the purpose of diagnosis and treatment of hypertension, highlight the magnitude of the uncertainty range around the current blood pressure cut-off point (140/90 mmHg), consisting of huge number of people being misdiagnosed of having or not having hypertension.
Failure in both directions are regrettable but, in the context of currently increasing aggressive approach to hypertension mandated by most guidelines, overdiagnosis exposes people unnecessarily to considerable risk for adverse drug reactions (ADR).
Poor treatment compliance rates – often a reflection of overdiagnosis?
Patients' low compliance rate with the prescribed medication is a widely acknowledged problem in hypertension treatment. Up to 50% of the patients quit the treatment they were given within the first year [89-92]. Adverse events with antihypertensive treatment are to a large extent dose related. More than 75% of these ADRs occur as such, at initiation of antihypertensive treatment [93,94]. In the commonly asymptomatic patients with recent diagnosed hypertension, even adverse events of trivial degree of intensity (mild headache, dizziness, etc.) can be perceived as interfering with their normal life. This reasoning has shaped treatment strategies during many years toward the lowest effective dose (of whatever agent had been used) at initiation of therapy, with the aim to maximize the likelihood of successful early treatment and improve the long-term compliance rate.
The task of the prescribing physician when attempting to stick to the lowest effective dose principle is, however not easy, as the labeling information at hand is often not enough informative for confident decision making. This is the case even when the doctors carefully take into account the known individual variations in drug response due to differences in age, weight, sex, ethnic background, state of health, concomitant medication use, and genetic polymorphism in drug metabolism.
The all too high dosage of a particular drug at launch seems to be inherent in the current drug development paradigm. Namely, high doses are selected during the early clinical trials phases with the purpose of assessing the agent's efficacy and less so its safety profile. Maximizing efficacy by preferential selection of higher doses may, obviously, be deleterious from safety point of view, particularly when such doses selected in early trials are used to prove efficacy in later phase trials and eventually are brought to the market once the drug is granted marketing authorization. Several studies have highlighted deviations in the post-licensing dose administration levels [95,96]. One study reported 115 instances of changes to the defined daily dose (DDD) between 1982 and 2000. Of these, about 60% were reductions relative to the initially designated DDD. Of some reason, cardiovascular drugs had the most DDD changes. In a similar study of changes to labeling instructions after licensing by the FDA, 79% of drugs underwent a reduction in drug dosage. Another study reported a 69% decrease in the length of time after marketing for the dosage change to occur [97]. The poor dosage selection predisposes to adverse drug events and consequent poor patient compliance.
The problem above is further compounded by the initial drug dosage guidance available for physicians in US [98] – the JNC guidelines and the Physicians' Desk Reference (PDR). The latter seems to be the preferred source of reference for some 90% of American physicians [99] while being used extensively also by consumers. A comparison of the two sources (the JNC VI versus the PDR editions of 1999 and 2000) has revealed large dose disparities in 23 out of 34 drugs (68%) in five frequently prescribed antihypertensives: ACEs, ARBs, BB, CCBs and diuretics. The PDR initial doses were at least 100% higher than the JNC VI doses except for chlorthalidone and a brand of metoprolol succinate. Furthermore, the PDR recommended use of lower initial doses in elderly in only 8 (18%) of 45 different drugs. This is, certainly, a real problem as hypertension is mostly prevalent among people above 60 years of age, known to have physiologically decreased hepatic and renal function and in whom clustering of multiple risk factors and/or associated morbidities require daily intake of several drugs. This constellation of factors makes the elderly vulnerable to first-dose reactions such as hypotension, dizziness, syncope, headaches, etc. Commonly, these ADR are dose-related [100,101] and are the leading cause of older patients discontinuing antihypertensive therapy (102).
The well-intended use by many physicians of a "half-dose" of a particular drug at initiation of treatment is hampered by the impossibility of dividing capsules, coated tablets and drugs with irregular format.
The "start low, go slow" principle, promoted constantly during the past decades as a stepwise approach to treating patients with hypertension, allows patients time to adjust psychologically to the fact that they have hypertension. A meta-analysis of 354 trials involving 56,000 participants showed that blood pressure reductions produced by the major classes of drugs at standard dose are similar, and that half the standard dose reduces efficacy by only 20% while more than halving side-effects [103-105].
Given the current prehypertension category, most "new hypertensive" patients are counselled as to lifestyle changes that either may fail or simply are not enough to prevent passing the threshold to "real" hypertension (>140/90 mmHg). The distress about having hypertension and possibly requiring life-long drug therapy may lead to development of anxiety symptoms that may be mistaken for ADR, which may lead to skipping doses or quitting treatment.
Worse than that is the case of those patients who, through the nature of their hypertension (e.g., white coat hypertension with blood pressure in excess of 20/10 mmHg above the cut-off point), are candidates for combination treatment from the start. Indeed, for these patients, or those whose blood pressure levels increase to these levels, the current guidelines recommend that consideration should be given to initiating (first-line) therapy with two drugs, either as separate prescriptions or in fixed low-dose combination. The main logic of these recommendations is simplicity of use of such combinations and their potential to considerably improve the blood pressure response rate while minimizing the incidence of adverse effects.
This might be a best evidence approach in patients with sustained hypertension diagnosed by ABPM but, it might be a high-risk approach in cases with white coat hypertension, not recognized by conventional office blood pressure measurement.
Physicians' non-compliance with treatment guidelines
Proper diagnosis of hypertension is of paramount importance for successful implementation in the clinical practice of current treatment guidelines. Consequently, failure to do so by the prescribing physician community suggests that there might be difficulties with accurately diagnosing hypertension, a fact, ultimately resulting in poor control of hypertension.
As control of hypertension, defined as BP < 140/90 mm Hg had been conferred status of quality indicator in the Health Employer Data Information Set by the National Committee for Quality Assurance in US [106], the potential reasons for any setback regarding this indicator are exposed to careful research. It should be emphasized, however, that the choice of this particular blood pressure threshold is neither evidence-based nor universally accepted.
Fourteen out of 27 national hypertension societies represented at the 17th world conference of Hypertension League Council held in Montreal in 1997 adopted the 140/90 mmHg cut-off point for hypertension diagnosis while the remaining 13 societies stayed with 160/95 mmHg [107].
Not surprisingly, a questionnaire survey among primary care physicians in US indicated that a significant proportion of them were reluctant to seek treatment goal below 140/90 mmHg while, in the same time, being pretty tolerant with mildly elevated SBP in older patients. These findings led the authors of the article [108] to make the statement that "physician behavior makes a significant contribution to the poor rates of hypertension control".
The situation is very similar in UK where an editorial by Campbell entitled: "Patients decide how low they go, not targets" [109], generated a wave of rapid responses reflecting the general discontent with current guidelines on hypertension [110].
Another issue of discontent among practicing physicians is the way results of clinical trials are presented, a fact that seems to impact the prescribing propensity of physicians and the patients' willingness to accept treatment [111].
The apparently small range between the two thresholds under debate (140/90 – 160/95 mmHg), encompasses the great majority of individuals with hypertension in whom most of the hypertension-related morbidity and mortality is recorded [112,113]. Overlapping this range, as a matter of fact, is the widespread uncertainty of the practicing physicians as to how confidently to diagnose the cases within this range, whether to initiate drug treatment in them or not and, if yes, whether the benefit of treatment will outweigh its potential risks.
Given that most of older hypertensive patients in general practices have a clustering of risk factors [3], there is high probability that in many cases a careful treatment approach by the physician will need to take in consideration the combined use of aspirin, a lipid lowering drug, several antihypertensive agents and, in may cases, an antidiabetic agent as well. The high probability of this scenario is generated by the NCEP III guidelines [114] that lowered the eligibility requirements for using statin drugs through moving the LDL-C threshold from = 4.1 mmol/L to = 3.3 mmol/L. This change translated into some 36 million people being in need for treatment, as compared with 15 million, before (a 140% increase) [115]. In addition to this, the results of ASCOT-LLT [116] and the ALLHAT-LLT [117] trials have expanded the relevance of lipid lowering treatment to the majority of hypertensive patients.
Obviously, it is difficult to find an all-fit formula for all these uncertainties; however, an emerging trend is to incorporate into the treatment management decision making the values and preferences of patients, clinicians and the general public [118-121]. This is certainly a process that is going on at the level of physician-patient relationship, unfortunately not always as transparent and well documented as it should be. The consequence is that the patient's choice [86-88] of, eventually, not be willing to accept treatment is not recorded as such, with the physician getting the blame of leaving the patient untreated, alternatively, insufficiently treated.
Philipps et al. [122] emphasized this problem as follows:
"Despite advances......, health care providers often do not initiate or intensify therapy appropriately during visits of patients with these problems. We define such behavior clinical inertia – recognition of the problem but failure to act on it".
"........achieving standard-of-care goals in only limited numbers of treated patients must be attributed either to therapeutic ineffectiveness or to clinical inertia. The attainment of treatment targets in clinical trials shows the effectiveness of present therapies for hypertension, elevated LDL cholesterol levels, and diabetes – leaving clinical inertia as the presumptive basis for treatment failure in many patients with these disorders".
The content of the first quote seems to ignore most of the sensitive aspects that normally occur in the patient-physician relationship and in the mutual decision making process of the two parts involved. Likewise, the second quote assumes generalization of the mean global estimates of clinical trials to the particular individual as a universal task to be fulfilled by all practicing physicians.
Indeed, epidemiologic data indicate that the long-term average BP is the best predictor for risk [123], however, short-term measurements of office BP, especially when they are near cut-off points for normal, all too often result in false-positive diagnosis of hypertension [124].
But, "physician behavior" and failure of clinicians to monitor blood pressure aggressively and institute pharmacotherapy when indicated were invoked as causes for poor control of hypertension already by the NHANES III, in 1997 [125]. From that gentle form of being reminded of their failures, practicing physicians have got to acquaint themselves gradually with the thought that they really deserved being blamed for "clinical inertia".
Most recently they could find out that they might be about to loose entirely what was left of their credibility in terms of diagnosing and treating hypertension.
In an article by Graves and Sheps in the American Journal of Hypertension [126], the authors convey their opinion on that: "Physicians do not measure BP well, and even if they do, the usefulness of their BP measurements is significantly compromised by the white coat effect".
This might certainly be true, and I also agree that something should be done, however, extreme solutions like: "It is clear from evidence trails that current practice of BP measurement is inadequate. Thus, the question is not whether physicians should or should not measure BP in the hypertensive patients; rather, it is how are we to replace the physician measurement with a higher quality BP measurement", might be too long far fetched.
(Wondering, whether this speedy progress might, some day, end up with doctors being disqualified to use a stethoscope!)
The authors answer the hypothetical question and suggest that "trained observers" or validated automated devices should be used. They acknowledge, however, that observer-dependent errors seen in "trained observers" (commonly nurses) are similar in type and magnitude to those seen in physicians [127].
The most recent AHA Scientific Statement [128] makes the justifiable remark that: "...physician blood pressure measurements should not be used exclusively in the routine management of the hypertensive patient". It goes on to say that: "With careful training even unpaid volunteers in large population surveys can measure blood pressure accurately [129]. However, even with the newer automated devices, the accuracy of the readings can be optimal only if all observers are appropriately trained and retrained and conscientious about using appropriate technique".
Eligibility to become "Observer" is dependent on a number of physical and cognitive competencies required to perform, assess and report the blood pressure measurement. The physical requirement includes vision, hearing and eye/hand/ear coordination (see Additional file 5).
Furthermore, observers are supposed to undergo careful evaluation as to their ability to assess "different types of bias, general technique, and the interpretation of the measurements including an understanding of the normal variability of blood pressure by time of day, exercise, timing of antihypertensive medications, etc. The observers should know how and when to communicate blood pressure readings gathered at home or other settings to the health care professional responsible for the care of the patient and management of hypertension".
A comprehensive model of retraining of all observers is suggested, whereby: "a central master trainer trains the site master trainers, and they in turn train the observers at each site. This model could be replicated within hospitals, ambulatory care settings, and community agencies".
The AHA document gives insight in the area of home/self-monitoring of blood pressure as a methodology that has the potential to "improve both therapeutic compliance and blood pressure control", according to two studies from mid 80s [130,131].
Despite the benefits of home blood pressure measurements – being more reliable than those recorded in the clinic setting – the Working Group on Blood Pressure Monitoring of the ESH neither recommend extensive use of the model, nor empowers patients to take action on the basis of the results [132]. Even without the empowerment of measuring blood pressure themselves, patients take the freedom to adjust their drug treatment or to quit it, without necessarily taking into consideration the level of blood pressure or the advice from a doctor. An estimated 50% of drugs prescribed for long term use are not taken because of concerns for side effects and/or drug dependency, a fact that powerfully impacts the overall compliance figures [133].
Furthermore, the main advantage of the home blood pressure monitoring – accurate readings – is not a guarantee of accurate reporting to the physician. This fact is highlighted in two studies employing measurement devices with memory. Patients using them were supposed to carefully record their blood pressure and to report the results to their physicians. More than half the subjects omitted or fabricated readings [134,135]. Supposedly, only the other halves of the patients were sufficiently well trained regarding: "information about hypertension, procedure for self monitoring, advice on equipment and its proper use, and interpretation of protocol and data" [136].
ESC advocates the home blood pressure monitoring for detecting white coat hypertension among patients with persistently raised clinic blood pressure, however, it emphasizes that a diagnosis requires confirmation with ambulatory monitoring.
Conclusion
A widespread awareness seems to emerge as to the doubtful clinical relevance of the most used investigation in clinical practice – the blood pressure measurement.
As mentioned in this paper, a great number of factors are contributing to the more than 100-year old method being carefully scrutinized, primarily because of the biological, random fluctuations of the blood pressure variable and the white coat effect.
These two latter aspects, as well as many others (the greater predictive value of the SBP, rather that that of the DBP, the increasing cardiovascular risk parallel with rising BP from values as low as 115/75 mmHg, the results of a multitude of randomized clinical trials), have forged a big quantum leap in our understanding of hypertension and seem to bring about the need for radical change, a real paradigm shift in the way we see and deal with blood pressure and with hypertension.
Sadly enough, the realization that the classically office-based BP can, on no account, be relied upon for diagnosis and treatment purposes, has generated a wave of mistrust with the practicing physicians, commonly involved in the management of patients with hypertension.
At closer scrutiny, however, most of the reproaches directed toward physicians are exaggerated and many of the "solutions" suggested extreme. Moreover, the peculiar "physician behavior" might be a deliberate acting to account for the uncertainties related to the much debated BP range of 140/90 – 160/95 mmHg and for the failure of the current health care systems to generalize implementation of ABPM in clinical practice, at least for all "first diagnosis" and treatment decision making.
The second part of this paper will attempt to elaborate on these latter issues.
Competing interests
The author(s) declare that they have no competing interests.
Supplementary Material
Additional File 1
Guidelines for blood pressure measurement. (Adapted from Perloff et al.).
Click here for file
Additional File 2
Factors that can interfere with accuracy of BP measurement (after McAlistar and Strauss).
Click here for file
Additional File 3
Effects of routine activities on BP (adapted from Campbell et al.).
Click here for file
Additional File 4
Indications for ABPM (JNC VI).
Click here for file
Additional File 5
Physical and cognitive competencies required in "observers" certified to measure blood pressure (AHA Scientific Statement, 2004).
Click here for file
==== Refs
The Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure – The JNC 7 Report JAMA 2003 289 2560 2572 12748199 10.1001/jama.289.19.2560
2003 European Society of Hypertension – European Society of Cardiology guidelines for the management of arterial hypertension Journal of Hypertension 2003 21 1011 1053 12777938 10.1097/00004872-200306000-00001
Williams B Poulter NR Brown MJ The BHS guidelines working party, for the British Hypertension Society. British Hypertension Society guidelines for hypertension management 2004 (BHS-IV): summary BMJ 2004 328 634 640 15016698 10.1136/bmj.328.7440.634
World Health Organization, International Society of Hypertension Writing Group 2003 World Health Organization (WHO/International Society of Hypertension (ISH) statement on management of hypertension J Hypertens 2003 21 1983 1992 14597836
Hajjar I Kotchen TA Trends in prevalence, awareness, treatment, and control of hypertension in the United States, 1988–2000 JAMA 2003 290 199 206 12851274 10.1001/jama.290.2.199
Wolf-Maier K Cooper RS Banegas JR Hypertension prevalence and blood pressure levels in 6 European countries, Canada, and the United States JAMA 2003 289 2363 2369 12746359 10.1001/jama.289.18.2363
Dobson Abergavenny R News extra: Blood pressure controls need to save on European health bills BMJ 2002 325 238 10.1136/bmj.325.7358.238
Fields LE Burt VL Cutler JA Hughes J Roccella EJ Sorlie P The burden of adult hypertension in the United States 1999 to 2000: A rising tide Hypertension 2004 44 4 398 404 15326093 10.1161/01.HYP.0000142248.54761.56
Haffner SM Ferrannini E Hazuda HP Stern MP Clustering of cardiovascular risk factors in confirmed prehypertensive individuals Hypertension 1992 20 38 45 1618551
Greenlund KJ Croft JB Mensah GA Prevalence of Heart Disease and Stroke Risk Factors in Persons With Prehypertension in the United States, 1999–2000 Arch of Intern Med 2004 164 19 2113 2118 15505124 10.1001/archinte.164.19.2113
Russell LB Valiyeva E Carson JL Effects of prehypertension on admissions and deaths. A simulation Arch Intern Med 2004 164 2119 2124 15505125 10.1001/archinte.164.19.2119
Russel LB Carson JL Taylor WC Milan E Dey A Jagannathan R Modeling all-cause mortality: projections of the impact of smoking cessation based on the NHEFS. NHANES I Epidemiologic Follow-up Study Am J Public Health 1998 88 630 636 9551006
Tsevat J Weinstein MC Williams LW Tosteson AN Goldman L Expected gains in life expectancy from various coronary heart disease risk factor modifications Circulation 1991 83 1194 1201 2013141
Grover SA Paquet S Levinton C Coupal L Zowall H Estimating the benefits of modifying risk factors of cardiovascular disease: a comparison of primary vs secondary prevention Arch Intern Med 1998 158 655 662 9521231 10.1001/archinte.158.6.655
American Heart Association Heart Disease and Stroke Statistics-2005 Update
The World Health Report Life in the 21st century: A vision for all 1998 Geneva: WHO
Puska P Development of public policy on the prevention and control of elevated blood cholesterol Cardiovasc Risk Factors 1996 6 203 210
World Health Organization Innovative care for chronic conditions: building blocks for action Geneva: WHO, document No WHO/MNC/CCH/0201 2002
Volpe M Alderman MH Furberg CD Jackson R Kostis JB Laragh JH Psaty BM Ruilope LM Beyond hypertension. Toward guidelines for cardiovascular risk reduction Am J Hypertens 2004 17 11 1068 1074 15533736 10.1016/j.amjhyper.2004.06.017
Perloff D Grim C Flack J Frohlich ED Hill M McDonald M Human blood pressure determination by sphygmomanometry Circulation 1993 88 2460 2467 8222141
McVicker JT Blood pressure measurement – does anyone do it right?: an assessment of the reliability of the equipment in use and the measurement techniques of clinicians J Fam Plann Reprod Health Care 2001 27 4 241
Ripoles Orti M Martin Rioboo E Diaz Moreno A Murcia Simon M Toledano Medina A Fonseca Del Prozo FJ Agreement in the measurement of blood pressure among different health professionals. Are mercury sphygmomanometers reliable? Aten Primaria 2001 27 4 234 243 11262332
Carney SL Gillies AH Green SL Paterson O Taylor MS Smith AJ Hospital blood pressure measurement: staff and device assessment J Qual Clin Pract 1999 19 2 95 98 10408749 10.1046/j.1440-1762.1999.00308.x
Veiga EV Nogueira MS Carnio EC Marques S Lavrador MA Souza LA Lima NK Nobre F Assessment of the techniques of blood pressure measurement by health professionals Arq Bras Cardiol 2003 80 1 89 93 12612727 10.1590/S0066-782X2003000100008
Julius S Kjeldsen SE Weber M for the VALUE trial group Outcomes in hypertensive patients at high cardiovascular risk treated with regimens based on valsartan or amlodipine: the VALUE randomized trial Lancet 2004 363 2022 2031 15207952 10.1016/S0140-6736(04)16451-9
The ALLHAT Officers and Co-ordinators for the ALLHAT Collaborative Group Major outcomes in high-risk hypertensive patients randomized to angiotensin converting enzyme inhibitor or calcium channel blocker vs diuretic. The Antihyhpertensive and Lipid-lowering Treatment to Prevent Heart Attack Trial (ALLHAT) JAMA 2002 288 1981 1997 10.1001/jama.288.16.1981
Campbell NR McKay DW Chockalingam A Fodor JG Errors in assessment of blood pressure: blood pressure measurement technique Can J Public Health 1994 85 2 S18 S21 7804946
McKay DW Raju MK Campbell NR Assessment of blood pressure measurement techniques Med Educ 1992 26 3 208 212 1614346
Feher M Harris-St John K Lant A Blood pressure measurement by junior hospital doctors – a gap in medical education Health Trends 1992 24 59 61 10121966
McKay DW Campbell NRC Parab LS Chockalingam A Fodor JG Clinical assessment of blood pressure J Hum Hypertens 1990 4 639 645 2096205
Mion D Pierin AM How accurate are sphygmomanometers? J Hum Hypertens 1998 12 4 245 248 9607693 10.1038/sj.jhh.1000589
Rouse A Marschall T The extent and implications of sphygmomanometer calibration error in primary care J Hum Hypertens 2001 15 9 587 591 11550103 10.1038/sj.jhh.1001241
Ali S Rouse A Practice audits: reliability of sphygmomanometers and blood pressure recording bias J Hum Hypertens 2002 16 5 359 361 12082498 10.1038/sj.jhh.1001384
Jones DW Frohlich ED Grim CM Grim CE Taubert KA Mercury sphygmomanometers should not be abandoned: an advisory statement from the Council for High Blood Pressure Research, American Heart Association Hypertension 2001 37 185 186 11230268
O'Brien E Waeber B Parati G Staessen G Meywers MG Blood pressure measuring devices: recommendations of the European Society of Hypertension BMJ 2001 322 531 536 11230071 10.1136/bmj.322.7285.531
Hospital for a Healthy Environment Mercury
Canzanello VJ Jensen PL Schwartz GL Are aneroid sphygmomanometers accurate in hospital and clinic setting? Arch Intern Med 2001 161 729 731 11231706 10.1001/archinte.161.5.729
Staessen JA Blood-pressure measuring devices: time to open Pandora's box and regulate Hypertension 2000 35 1037 10818060
Ramsey M III Blood pressure monitoring: automated oscillometric devices J Clin Monit 1991 7 56 67 1999699
O'Brien E Pickering T Asmar R Working Group on Blood Pressure monitoring of the European Society of Hypertension International Protocol for validation of blood pressure measuring devices in adults Blood Press Monit 2002 7 3 17 12040236 10.1097/00126097-200202000-00002
Gerin W Schwartz AR Schwartz JE Limitations of current validation protocols for home blood pressure monitors for individual patients Blood Press Monit 2002 7 313 318 12488651 10.1097/00126097-200212000-00004
Schwartz AR Haas DC Gerin W Pickering TG Accurate Measurement of Blood Pressure JAMA 2003 289 21 2792 12783901 10.1001/jama.289.21.2792-a
Jones DW Appel LJ Sheps SG Roccella EJ Lenfant C Measuring Blood Pressure Accurately: New and Persistent Challenges JAMA 2003 289 8 1027 1030 12597757 10.1001/jama.289.8.1027
Pickering TG Pieper C Chechter CB Ambulatory monitoring and blood pressure variability 1991 London: Science Press 2.1 2.16
Joint National Committee on Detection, Evaluation, and Treatment of High Blood Pressure The Sixth report of the Joint National Committee on Detection, Evaluation, and Treatment of High Blood Pressure (JNC VI) Arch Intern Med 1997 157 2413 2446 9385294 10.1001/archinte.157.21.2413
Prasad N Isles C Ambulatory blood pressure monitoring: a guide for general practitioners BMJ 1996 313 1535 1541 8978234
Pickering T for an American Society of Hypertension Ad Hoc Panel Recommendations for use of home (self) and ambulatory blood pressure monitoring Am J Hypertens 1996 9 1 11 8834700 10.1016/0895-7061(95)00341-X
O'Brien E Coats A Owens P Use and interpretation of ambulatory blood pressure monitoring: recommendations of the British Hypertension Society BMJ 2000 320 1128 1134 10775227 10.1136/bmj.320.7242.1128
Mancia G Zanchetti A Agabiti-Rosei E Ambulatory blood pressure is superior to clinic blood pressure in predicting treatment-induced regression of left ventricular hypertrophy. AMPLE Study Group. Study on ambulatory Monitoring of Blood Pressure and Lisinopril Evaluation Circulation 1997 95 1464 70 9118514
Gardner SF Schneider EF 24-hour ambulatory blood pressure monitoring in primary care J Am Board Pract 2001 14 3 166 171
Schlechter CB Sequential analysis in a Bayesian model of diastolic blood pressure measurement Med Decis Making 1988 8 191 196 3398747
Chatellier G Day M Dobrie G Menard J Feasibility study of N-of-1 trials with blood pressure self-monitoring in hypertension Hypertension 1995 25 294 301 7843782
Reeves RA A review of the stability of ambulatory blood pressure: implications for diagnosis of hypertension Clin Invest Med 1991 14 251 255 1893658
National High Blood Pressure Education Program Working Group report on high blood pressure in pregnancy Am J Obstet Gynecol 1990 163 1689 1712
Reeves RA Does this patient have hypertension? How to measure blood pressure JAMA 1995 273 1211 1218 7707630 10.1001/jama.273.15.1211
Birkett NJ The effect of alternative criteria for hypertension on estimates of prevalence and control J Hypertens 1997 15 237 244 9468450 10.1097/00004872-199715030-00004
O'Brian E Ambulatory blood pressure monitoring in the management of hypertension Heart 2003 89 571 576 12695477 10.1136/heart.89.5.571
MacDonald MB Laing GP Wilson MP Wilson TW Prevalence and predictors of white-coat response in patients with treated hypertension Can Med Assoc J 1999 161 265 269 10463047
William BW Ambulatory blood-pressure monitoring in clinical practice New Engl J Med 2003 348 24 2377 2378 12802022 10.1056/NEJMp030057
Little P Barnet J Barnsley L Marjoram J Fitzgerald-Baron A Mant D Comparison of agreement between different measures of blood pressure in primary care and daytime ambulatory blood pressure BMJ 2002 325 254 259 12153923 10.1136/bmj.325.7358.254
Bellomo G Narducci PL Rondoni F Pastorelli G Stagnoni G Prognostic value of 24-hour blood pressure in pregnancy JAMA 1999 282 1447 1452 10535435 10.1001/jama.282.15.1447
Staessen JA Fagard R Thjis L Randomized double-blind comparison of placebo and active treatment for older patients with isolated systolic hypertension Lancet 1997 350 757 764 9297994 10.1016/S0140-6736(97)05381-6
Clement DL De Buyzere ML De Baquer DA de Leeuw PW Duprez DA Fagard RH Cheeraert PJ Missault LH Braun JJ Six RO Van Der Niepen P O'Brien E Prognostic value of ambulatory blood-pressure recordings in patients with treated hypertension New Engl J Med 2003 348 2407 2415 12802026 10.1056/NEJMoa022273
Pickering T Davidson K Gerin W Masked hypertension Hypertension 2002 40 795 796 12468559 10.1161/01.HYP.0000038733.08436.98
Guillaume B Chatellier G Genes N Vaur L Vaisse B Menard J Mallion JM Cardiovascular prognosis of "Masked Hypertension" detected by blood pressure self-measurement in elderly treated hypertensive patients JAMA 2004 291 1342 1349 15026401 10.1001/jama.291.11.1342
Kottke TE Stroebel R Hoffman R JNC 7 – It's more than high blood pressure JAMA 2003 289 19 2573 2575 12748200 10.1001/jama.289.19.2573
Mancia G Grassi G Systolic and diastolic blood pressure control in antihypertensive drug trials J Hypertens 2002 20 1461 1464 12172300 10.1097/00004872-200208000-00001
Prospective Studies Collaboration Age-specific relevance of usual blood pressure to vascular mortality: a meta-analysis of individual data for one million adults in 61 prospective studies Lancet 2002 360 1903 1913 12493255 10.1016/S0140-6736(02)11911-8
PROGRESS Collaborative Group Randomized trial of a perindopril-based blood-pressure-lowering regimen among 6105 individuals with previous stroke or transient ischaemic attack Lancet 2001 358 1033 1041 11589932 10.1016/S0140-6736(01)06178-5
Law MR Wald NJ Risk factor thresholds: their existence under scrutiny BMJ 2002 324 1570 1576 12089098 10.1136/bmj.324.7353.1570
Hansson I Zanchetti A Carruthers SG Dahloef B Elmfeldt D Julius S Effects of intensive blood pressure lowering and low-dose aspirin in patients with hypertension: principal results of the Hypertension Optimal Treatment (HOT) randomized trial Lancet 1998 351 1755 1762 9635947 10.1016/S0140-6736(98)04311-6
UK Prospective Diabetes Study Group Tight blood pressure control and risk of macrovascular and microvascular complications in type 2 diabetes. UKPDS 38 BMJ 1998 317 703 713 9732337
Stergiou GS Karotsis AK Symeonidis A Vassilopulou VA Aggressive blood pressure control in general practice (ABC-GP) study: can the new targets be reached? J Hum Hypertens 2003 17 767 773 14578916 10.1038/sj.jhh.1001610
McMahon S Peto R Cutler J Blood pressure, stroke, and coronary heart disease. Part 1: prolonged differences in blood pressure: Prospective observational studies corrected for the regression dilution bias Lancet 1990 335 765 774 1969518 10.1016/0140-6736(90)90878-9
SHEP Cooperative Research Group Prevention of stroke by antihypertensive drug treatment in older persons with isolated hypertension: final results of the Systolic Hypertension in Elderly Program (SHEP) JAMA 1991 265 3255 3264 2046107 10.1001/jama.265.24.3255
Mancia G Grassi G Rationale for the use of a fixed combination in the treatment of hypertension Eur Heart J 1999 1 Suppl L L14 L19
Mancia G Grassi G Systolic and diastolic blood pressure control in antihypertensive drug trials J Hypertens 2002 20 1461 1464 12172300 10.1097/00004872-200208000-00001
Hoes A Grobbee D Lubsen J Does drug treatment improve survival? J Hypertens 1995 3 805 811 7594445
Mulrow CD Cornell JA Herrera CR Kadri A Farnett L Aguilar C Hypertension in the elderly: implications and generalizability of randomized trials JAMA 1994 272 1932 1938 7990246 10.1001/jama.272.24.1932
Swales J Pharmacological management of hypertension Lancet 1994 344 380 385 7914311 10.1016/S0140-6736(94)91405-2
Di Blasi Z Harkness E Ernst E Georgiou A Kleijnen J Influence of context effects on health outcomes: a systematic review Lancet 2001 357 757 762 11253970 10.1016/S0140-6736(00)04169-6
Kleijnen J de Craen AJM van Everdingen J Krol L Placebo effect in double-blind clinical trials: a review of interactions with medications Lancet 1994 344 1347 1349 7968033 10.1016/S0140-6736(94)90699-8
Kaptchuk TJ Powerful placebo: the dark side of the randomized controlled trial Lancet 1998 351 1722 1725 9734904 10.1016/S0140-6736(97)10111-8
Onel E Hammond C Wasson JH An assessment of the feasibility and impact of shared decision making in prostate cancer Urology 1998 51 63 66 9457290 10.1016/S0090-4295(97)00494-9
Benson J Britten N Patients decisions about whether or not to take antihypertensive drugs: qualitative study BMJ 2002 325 873 876 12386041 10.1136/bmj.325.7369.873
Redelmeier DA Rozin P Kahneman D Understanding patients' decisions JAMA 1993 270 72 76 8510300 10.1001/jama.270.1.72
Feldman R Bacher M Campbell N Adherence to pharmacologic management of hypertension Can J Public Health 1998 89 I16 I18 9813921
Alderman MH Madhavan S Cohen H Antihypertensive drug therapy: the effect of JNC criteria on prescribing patterns and patient status through the first year Am J Hypertens 1996 9 413 418 8735170 10.1016/0895-7061(95)00438-6
Tomlinson B Optimal dosage of ACE inhibitors in older patients Drugs Aging 1996 9 262 273 8894524
Flack JM Novikov SV Ferrario CM Benefits of adherence to anti-hypertensive drug therapy Eur Heart J 1996 17 suppl A 16 20 8737196
Lazarou J Pomeranz BH Corey PN Incidence of adverse drug reactions in hospitalized patients: a meta-analysis of prospective studies JAMA 1998 279 1200 1205 9555760 10.1001/jama.279.15.1200
Melmon KL Morrelli HF Hoffman BB Nierenberg DW Melmon and Morrelli's Clinical Pharmacology: Basic Principles in Therapeutics 1993 3 New York, NY:McGraw-Hill Co 33
Heerdink ER Urquhart J Leufkens HG Changes in prescribed drug doses after market introduction Pharmacoepidemiol Drug Saf 2002 11 6 447 453 12426928 10.1002/pds.745
Cross J Lee H Westelinck H Postmarketing drug dosage changes of 499 FDA-approved new molecular entities, 1980–19999 Pharmacoepidemiol Drug Saf 2002 11 6 439 446 12426927 10.1002/pds.744
Struijker Boudier HA A drug is not a drug: a commentary Pharmacoepidemiol Drug Saf 2002 11 6 437 438 12426926 10.1002/pds.767
Cohen JS Adverse drug effects, compliance, and initial doses of antihypertensive recommended by the Joint National Committee vs the Physicians' Desk Reference Arch Int Med 2001 161 6 880 885 11268233 10.1001/archinte.161.6.880
Cohen JS Insel PA The Physicians' Desk Reference : problems and possible improvements Arch Intern Med 1996 156 1375 1380 8678706 10.1001/archinte.156.13.1375
Brawn LA Castleden CM Adverse drug reactions: an overview of special considerations in the management of the elderly patient Drug Safety 1990 5 421 435 2285496
Rochon PA Anderson GM Tu JV Age- and gender-related use of low-dose drug therapy: the need to manufacture low-dose therapy and evaluate the minimum effective dose J Am Geriatr Soc 1999 47 954 959 10443856
Tomlinson B Optimal dosage of ACE inhibitors in older patients Drugs Aging 1996 9 262 273 8894524
Law MR Wald NJ Thompson SG By how much and how quickly does reduction in serum cholesterol concentration lower risk of ischaemic heart disease? BMJ 1994 308 367 372 8043072
Law MR Wald NJ Morris JK Jordon RE Blood pressure lowering drugs: analysis of 354 randomised trials to assess the value of low dose combination therapy BMJ 2003 326 1419 12829553 10.1136/bmj.326.7404.1419
Sacks FM Tonkin AM Shepherd J Effect of pravastatin on coronary disease events in subgroups defined by coronary risk factors-the prospective pravastatin pooling project Circulation 2000 102 1893 1900 11034935
National Committee for Quality Assurance HEDIS 2000: Technical Specifications Washington, DC: National Committee for Quality Assurance 1999 2
Mulrow PJ Izzo JL, Black HR, Goodfriend TL Hypertension: a worldwide epidemic Hypertension primer: the essentials of high blood pressure 1999 2 Baltimore: Williams and Wilkins 721 723
Hyman DJ Pavlik VN Self-reported hypertension treatment practices among primary care physicians: blood pressure thresholds, drug choices, and the role of guidelines and evidence-based medicine Arch Intern Med 2000 160 15 2281 2286 10927724 10.1001/archinte.160.15.2281
Campbell NC Treating hypertension with guidelines in general practice. Patients decide how low they go, not targets BMJ 2004 329 523 524 15345603 10.1136/bmj.329.7465.523
Williams B Poulter NR Brown MJ Davis M McInnes GT Potter JF Sever PS Thom S McG British Hypertension society guidelines for hypertension management 2004 (BHS-IV): summary BMJ 2004 328 634 640 BMJ rapid responses (13 March 2004) to 15016698 10.1136/bmj.328.7440.634
Sackett D Cook R Understanding clinical trials BMJ 1994 309 755 756 7950551
McAlister FA Laupacis A Towards a better yardstick: the choice of treatment thresholds in hypertension Can J Cardiol 1998 14 47 51 9487272
Grover SA Paquet S Levington C Coupa L Zowal H Estimating the benefits of modifying cardiovascular risk factor: a comparison of primary versus secondary prevention Arch Intern Med 1998 158 655 662 9521231 10.1001/archinte.158.6.655
Gueyffier F Boutitie F Boissel JP Pocock S Coope J Cutler J Effect of antihypertensive drug treatment on cardiovascular outcomes in women and men. A meta-analysis of individual patient data from randomised, controlled trials. The INDANA Investigators Ann Intern Med 1997 126 761 767 9148648
National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (Adult Treatment Panel III) Third Report of the National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (Adult Treatment Panel III) final report Circulation 2002 106 3143 3421 12485966
Sever PS Dahlof B Poulter NR Prevention of coronary and stroke events with atorvastatin in hypertensive patients who have average or lower-than-average cholesterol concentrations, in the Anglo-Scandinavian Cardiac Outcomes Trial – Lipid Lowering Arm (ASCOT-LLA): a multicentre randomized controlled trial Lancet 2003 361 1149 1158 12686036 10.1016/S0140-6736(03)12948-0
ALLHAT Officers and Coordinators for the ALLHAT Collaborative Research Group Major outcomes in moderately hypercholesterolemic, hypertensive patients randomized to pravastatin vs. usual care: the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT-LLT) JAMA 2002 288 2998 3007 12479764 10.1001/jama.288.23.2998
Misselbrook D Armstrong D Thinking about risk: Can doctors and patients talk the same language? Fam Pract 2002 19 1 2 11818341 10.1093/fampra/19.1.1
Gigerenzer G Edwards A Simple tools for understanding risks: from innumeracy to insight BMJ 2003 327 741 744 14512488 10.1136/bmj.327.7417.741
Boulware LE Daumit GL Frick KD An evidence-based review of patient-centered bahavioral interventions for hypertension Am J Prev Med 2001 21 221 232 11567845 10.1016/S0749-3797(01)00356-7
Hlatk M Patient preferences and clinical guidelines JAMA 1995 273 1219 1220 7707631 10.1001/jama.273.15.1219
Phillips LS Branch WT Cook CB Doyle JP El-Kebbi IM Gallina DL Miller CD Ziemer DC Barnes CS Clinical Inertia Ann Int Med 2001 135 9 825 834 11694107
MacMahon S Peto R Cutler J Blood pressure, stroke, and coronary heart disease, I: prolonged differences in blood pressure: prospective observational studies corrected for the regression dilution bias Lancet 1990 335 765 774 1969518 10.1016/0140-6736(90)90878-9
Perry HMJ Miller JP Difficulties in diagnosing hypertension: implications and alternatives J Hypertens 1992 10 887 896 1325524
Fedder DO Koro CE L'Italien GJ New National Cholesterol Education Program III guidelines for primary prevention lipid lowering drug therapy: projected impact on the size, sex, and age distribution of the treatment-eligible population Circulation 2002 105 152 156 11790693 10.1161/hc0202.101971
Graves JW Sheps SG Does evidence-based medicine suggest that physicians should not be measuring blood pressure in the hypertensive patient? Am J Hypertens 2004 17 4 354 360 15062890 10.1016/j.amjhyper.2003.12.007
Drevenhorn E Hakansoon A Petersson K Blood pressure measurement: an observational study of 21 public health nurses J Clin Nursing 2001 10 189 194 10.1046/j.1365-2702.2001.00481.x
Pickering GT Hall GJ Appel LJ Falkner BE Graves J Hill MN Jones DW Kurtz T Sheps SG Roccela ED Recommendations for Blood Pressure Measurement in Humans and Experimental Animals. Part 1: Blood Pressure Measurements in Humans: A Statement for Professionals From the Subcommittee of Professional and Public Education of the American Heart Association Council on High Blood Pressure Research Hypertension Published online before print December 20, 2004.
Adams C Burke V Beilin LJ Accuracy of blood pressure measurement and anthropometry among volunteer observers in a large community survey J Clin Epidemiol 2002 55 338 344 11927200 10.1016/S0895-4356(01)00492-9
Stahl SM Kelley CR Neill PJ Grim CE Mamlin J Effects of home blood pressure measurement on long-term BP control Am J Public Health 1984 74 704 709 6742256
Edmonds D Foerster E Groth H Greminger P Siegenthaler W Vetter W Does self-measurement of blood pressure improve patient compliance in hypertension? J Hypertens Supp 1985 3 S31 S34
Stergiou G Megden T Padfield PL Parati G O'Brien E Working Group on Blood Pressure monitoring of the European Society of Hypertension. Self monitoring of blood pressure at home BMJ 2004 329 870 871 15485940 10.1136/bmj.329.7471.870
Horne R Weinman J Patients' beliefs about prescribed medicines and their role in adherence to treatment in chronic illness J Psychosom Res 1999 47 555 567 10661603 10.1016/S0022-3999(99)00057-4
Mengden T Hernandez Medina RM Beltran B Alvarez E Kraft K Vetter H Reliability of reporting self-measured blood pressure values by hypertensive patients Am J Hypertens 1998 11 1413 1417 9880121 10.1016/S0895-7061(98)00241-6
Myers MG Self-measurement of blood pressure at home: the potential for reporting bias Blood Press Monit 1998 3 suppl 1 S19 S22
O'Brien E Asmar R Beilin L Imai Y Mancia G Mengden T on behalf of the European Society of Hypertension Working Group on Blood Pressure Monitoring European Society of Hypertension Working Group on Blood Pressure Monitoring. European Society of Hypertension recommendations for conventional, ambulatory and home blood pressure measurement J Hypertens 2003 21 821 848 12714851 10.1097/00004872-200305000-00001
| 15813975 | PMC1087862 | CC BY | 2021-01-04 16:47:32 | no | Curr Control Trials Cardiovasc Med. 2005 Apr 6; 6(1):6 | utf-8 | Curr Control Trials Cardiovasc Med | 2,005 | 10.1186/1468-6708-6-6 | oa_comm |
==== Front
Epidemiol Perspect InnovEpidemiologic perspectives & innovations : EP+I1742-5573BioMed Central London 1742-5573-2-21584017510.1186/1742-5573-2-2Analytic PerspectiveHistorical Perspective: The social determinants of disease – some roots of the movement Syme S Leonard [email protected] University of California, Berkeley, School of Public Health 140 Warren Hall, Berkeley, CA 94720 USA2005 19 4 2005 2 2 2 18 3 2005 19 4 2005 Copyright © 2005 Syme; licensee BioMed Central Ltd.2005Syme; 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.
This is an account of the early days of research on social determinants as I experienced them. I describe my time as one of four Fellows in a new training program in Medical Sociology at Yale University and how I came to be the first Sociologist employed in the U.S. Public Health Service. I then became the first Executive Secretary of a new Study Section at NIH dealing with a small number of research grant proposals in the field of Epidemiology. My account deals with some of my experiences in this developing field, culminating with my appointment as the first Sociologist to become a Professor of Epidemiology in a School of Public Health.
==== Body
Introduction
In 2001, a colleague and I wrote an article describing a remarkable phenomenon: between the years 1995 and 2001, there had been ten books published that focused on the social determinants of disease [1]. We suggested that this explosion of work marked the coming of age of the field of social epidemiology. Since that paper appeared, more than a dozen new books dealing with Social Epidemiology have been published, more than a dozen Reports from the Institute of Medicine (National Research Council) have been written on such topics as social and behavioral approaches to racial and ethnic inequalities in health, and hundreds of journal articles have been written on these issues.
In addition to these publications and reports, the Robert Wood Johnson Foundation has recently established a new program to train postdoctoral scholars in a field they refer to as Health and Society; this training program has a heavy emphasis on the study of the social determinants of disease and is, I believe, the first major national initiative ever taken to train students in social epidemiology. Perhaps in response to these developments, I was asked by the Department of Epidemiology at the Mailman School of Public Health at Columbia University to give a lecture on "The Social Determinants of Disease: The Roots of the Movement". They wanted me to describe the "beginnings" of the field of social epidemiology on the basis of my personal experience. I received many comments following that talk suggesting it might be of interest to record my remarks in published form and that is the purpose of the present paper.
It is of course presumptuous for me to discuss the beginnings of this work based on my own limited exposure. The real beginning probably begins with Hippocrates and includes such other early scholars as Louis Villerme, Rudolph Virchow, Edgar Sydenstricker, and Emile Durkheim as well as more recent scholars such as Thomas McKeown, Saxon Graham, Mervyn Susser, Leo Reeder, Bruce Dohrenwend, Sol Levine, and John Cassel. My limited personal experience should therefore be considered within this very much broader context.
Analysis
My first attempt to think about my beginnings of work in social epidemiology was described in the Foreword to the Berkman/Kawachi textbook on Social Epidemiology [2]. As I indicated in that Foreword, my thinking about social determinants began in 1955 when I was accepted into a training program in the Department of Sociology at Yale University. That training program, called "medical sociology" was funded by the Commonwealth Fund and it was the first such formal training program in the world. There were four of us in the program. We were given a choice early on as to whether we would focus on what was then called the sociology of medicine or sociology in medicine. As I noted in the textbook, the logical choice was for me to choose to study the sociology of medicine because there already existed a relatively large and interesting literature on this topic dealing with the institution of medicine and medical care, the sick role, and attitudes and beliefs of patients regarding illness, pain, and medical treatment.
For reasons that are not clear to me, I decided to study sociology in medicine, which I took to mean the study of how social factors affect health and well-being. I now realize that what my Professors at Yale really meant by this term was nowhere near as grandiose as my version. Professors August Hollingshead and Frederick Redlich were at that time doing a large study of the link between social class and mental illness and that is what they meant by the term "sociology in medicine": They wanted me to help them with their research. I had no interest in the topic of mental illness because I took it for granted that social factors would somehow be related to mental illness. Looking back, I can see what a naïve view this was, but that was my uninformed position at that time. Instead, I wanted to know if social factors were related to diseases that were not so obviously connected to the social world, diseases such as heart disease, cancer and arthritis. Not only was this a naïve view, but it was also a reckless decision because there was virtually no literature on these topics at the time and no one was sure there ever would be.
When I graduated, I was scheduled to go into the Army to fight in Korea but Professor Hollingshead said that I might want to consider an alternative that would give me a military deferment: go to work for the U.S. Public Health Service in Washington. I agreed that that was a better idea. He said that he recently had talked to a statistician in the Heart Disease Control Program in Washington who wanted to hire a sociologist. So I went to Washington and met Phillip Enterline. I asked him why he wanted to hire a sociologist to study heart disease and he said that he had no idea. He and his group had just completed a study of the geographic distribution of coronary heart disease mortality in the U.S. and they found very high rates on the East and West Coasts and in the Detroit-Chicago metropolitan area but low rates elsewhere. They had not been able to explain this finding and they thought that perhaps a sociologist might be able to help.
So I took the job. I was to be classified as a Statistician in the Civil Service because there was no category available for a Sociologist. I made the mistake of reporting this to Professor Hollingshead and he was not very happy. "If there's no category for a Sociologist, make one!" He raised such a fuss that they in fact did. So I was the first Sociologist labeled as such in the Civil Service.
Then I went to work and it was a disaster. I decided to begin my work by looking at data from a state with a very low death rate from CHD with the idea of then doing a similar study in a higher rate state. We obtained some wonderful data from North Dakota, a low rate State. In a six-county area of North Dakota, we were able to obtain information on every case of coronary heart disease that occurred in men, 35–64 years of age, in a one year period. Then we selected two age-matched men, free of CHD, from a representative sample of the 6 county area from which the cases came. I then set about testing all the hypotheses that I had learned in graduate school. In those days, we were thinking about marginality, status crystallization and many other concepts that no one can now remember. It must be recognized, of course, that there was no literature or previous research to rely on. This was, I think, the first such study of CHD ever done with social factors. So I based my work on the concepts that I had been studying in school. I spent a year doing this. Not one of the hypotheses worked out. The cases and controls did not differ from one another on any of the dozens and dozens of ideas that were then popular in Sociology. I think Enterline must have thought he made a major mistake in hiring me.
So I decided on a different tack. I would go through all of the data and see on which items there might be a difference between cases and controls. I had been taught that this type of fishing expedition was not a very good way to proceed, but I was desperate. In this analysis, I was able to see a considerably higher rate of CHD among men who had changed jobs and who had moved geographically and, especially, among men who had moved from farms to white collar jobs in the city [3]. I observed all of this, of course, after controlling for smoking, blood pressure, and many other CHD risk factors. I called this phenomenon "cultural mobility" [4]. I was then able to repeat this analysis with a remarkably similar data set in a State with a much higher rate of CHD, California [4]. And I found precisely the same thing as in North Dakota.
I came to a fateful decision based on this experience. I decided that our social theory was not very useful in helping us think about health matters. I decided to no longer base my research on theory but to collect reasonable seeming data and do fishing expeditions. I taught several generations of students to forget the "theory thing" and just go for it. The result is that we now in social epidemiology have piles and piles of findings and no way to make sense of it or to think about what needs to be done next. This sorry situation is not all my fault of course but I have been a major contributor. And the reason for it is to be found in the wheat fields of North Dakota. Fortunately, better minds than mine are now prevailing and things are getting better. For example, one of my former students, Nancy Krieger, is forcefully demonstrating the power, and importance, of theory in spite of everything I tried to teach her [5]. This part of my work has not been one of my better contributions.
I was prevailed upon by Professor Jeremiah Stamler to present my North Dakota findings at a meeting of the American Heart Association. It was a daunting experience. In the front row sat all of the most eminent cardiovascular epidemiologists in the world and I suggested that above and beyond the usual CHD risk factors was a set of social factors that no one could understand and for which possible disease mechanisms were very difficult to visualize. One eminent epidemiologist cornered me after my presentation and angrily criticized the whole approach. His argument: "What are we supposed to do with findings like this? Tell people not to move or change jobs? All you are doing is distracting people from the real issues which are cholesterol, blood pressure and smoking. You are doing shameful work and you should stop it!"
I did not stop, of course, but these were difficult times. I had other troubles on other fronts. For example, a nutritionist on the staff of the Heart Disease Control Program asked me to help her design a dietary questionnaire. These were the days before we had the well-established instruments we have today. She wanted to do a study of Seventh Day Adventists. There was at that time, 1959, a growing body of evidence, and speculation, that a diet high in fat might be a risk factor for coronary heart disease. Seventh day Adventists were lacto-ovo-vegetarians and it was thought to be interesting to study their lipid levels and other health issues. I agreed to help her design a questionnaire. As we worked, it occurred to me that Adventists might have better lipid levels not only because of their diet but because they were religious. So I convinced her to let me add three questions at the end of the interview about their church attendance and about the importance of religion in their lives.
Since this was a government survey, all forms had to be cleared by a group in the Bureau of the Budget. Two weeks after we submitted our questionnaire, word came that it had been approved but that my three questions on religion had been deleted. I was not very happy. Upon inquiry, I was informed that there is in the U.S. Constitution a policy of separating church and state and that my three questions, on a government form, violated the Constitution. So I handed in my resignation. An Assistant Surgeon General summoned me to his office the next day. "What's all this about quitting?" he asked. I told him that as a sociologist I needed to ask people questions about their lives, including their religious beliefs, and if I wasn't going to be able to do that, there was no point in my working in the government. He told me to calm down. He asked if there was any evidence to support my hypothesis that religious beliefs had anything to do with lipid levels. "Of course there is!" I lied. "That's why I put those questions in!" "OK," he said, "bring me the evidence and then we'll talk". I went to poor suffering Phil Enterline and asked him for 3 weeks off so that I could search for the evidence that I had so confidently said existed.
I worked very hard during those three weeks and I did in fact find quite a bit of evidence. There was information about religion and stress taken from studies of Trappist and Bendictine monks and there was evidence about stress and lipids from studies of medical students at exam time and from tax accountants at tax time. I also did a lot of research about the Seventh-day Adventist religion and its relevance for stress research. As a complete amateur, I concluded that the SDA religion was based on the return of the Lord and that that return will occur soon after we see people warring with one another and when there is much civil strife and, in general, when everything is falling apart. So I argued that Seventh-day Adventists have a very different response than the rest of us when they read the daily newspaper. We moan about the events of the day while they see the bad news as bringing them closer to salvation.
I wrote a 45 page paper about stress and lipids, religion and stress, and about the Seventh-day Adventist religion. It was, I must say, quite elegant. I ended with a paragraph saying that in light of the foregoing, the three questions I wanted to ask were clearly warranted. I handed in my paper and was summoned a few days later to the Assistant Surgeon General's office. He said he was impressed with my paper and that he was satisfied that there was a credible scientific basis for my three questions. "But," he said, "we now have to consider the constitutional issue". I felt betrayed and said I was going to resign. Again, he told me calm down. "Give me a few weeks", he said. Several months later, he announced a change in government policy about such issues. One can now ask about things like religion if a case can be made that more good than harm will come from the inquiry. There has to be a good, or even compelling, reason for violating the constitution, but it can be done.
A year later, in 1960, I was asked to move to the National Institutes of Health to establish, for the first time, an Epidemiology Study Section. This was a very powerful position for a young 28-year-old beginner. This new Study Section was to be established to deal with the small number of research grant applications that were beginning to be submitted to the NIH dealing with the epidemiology of such non-infectious diseases as arthritis, mental illness, cancer, heart disease, and injuries. I had received my Ph.D. in a new field called Medical Sociology just three years prior to this invitation and, while I had been working as a fledgling epidemiologist in a heart disease program in the U.S. Public Health Service, I was not very knowledgeable about the field; not many others were either. My boss at NIH, Dr. Murray Goldstein (who later became the Director of the National Institute of Neurological Disorders and Stroke) asked me to nominate a group of people who could serve on this new Study Section and I began to do research to deal with this challenge. The first thing I learned was that we could not use the word "Epidemiology" for the title of the new Study Section because that word was reserved for the study of infectious diseases only. We therefore came up with an alternative name, the "Human Ecology Study Section". We then selected a truly distinguished multi-disciplinary group of members.
Our first choice was Abraham Lilienfeld from Johns Hopkins University. Even then, he was the outstanding epidemiologist in the country. Then there was William Cochran from Harvard, perhaps the most outstanding biostatistician in the country. And Arno Motulsky the geneticist then at Washington University. Other members included John Fulton (a dentist), William Clark (an infectious disease epidemiologist), Schulyer Kohl (an obstetrician), Felix Moore (a biostatistician), George Reader (an internist) and Robert Shank (an internist and nutritionist). And, because of my training as a sociologist, I nominated my Professor from Yale, August B. Hollingshead (who, as I noted earlier, was beginning to do pioneering work on the link between social class and mental health) and Otis Dudley Duncan, from Chicago, who was working on the relationship between macro social forces and behavior. There were no women or minorities on the Committee reflecting the fact that there were very few women and scholars from minority groups working in this area at that time.
The Human Ecology Study Section, later renamed the Epidemiology Study Section, eventually grew into several large subdivisions. In those days, however, there were very few applications to review and we took it as our mission to help develop the field. For that reason, we went on site visits very frequently. If a grant proposal looked promising, but inadequate, we went to visit the group to help them do it better. I was on airplanes all the time. It was a truly fascinating experience. We visited John Cassel in North Carolina. He was doing some of the very best work at the time and, interestingly, much of his research is still the best. He was doing a study about the health consequences of people moving from rural places to take jobs in factories. We went out into the hills of western North Carolina to visit a paper mill that was one of his factory sites. We met a remarkable young occupational physician who we later induced to come to Chapel Hill to study Public Health. That was the beginning of Herman (Al) Tyroler's distinguished career in Epidemiology. We gave a young Warren Winkelstein his first grant to study the health effects of air pollution in Buffalo, New York. We supported Lawrence Hinkle's work on stress in telephone workers. We supported research on Seventh Day Adventists to see if their good health was due to nutrition or spirituality. We supported Sam Shapiro's pioneering study of mammography in HIP. We supported Saxon Graham who was studying the link between social factors and cancer at Roswell Park. We supported Bruce Dohrenwend's classic work on mental health. And we supported the work of Sol Levine and Norman Scotch in their study of social factors in the Framingham study. There was at that time a lot of money available and we were able to work hard to stimulate epidemiologic research. Since I was the Executive Secretary and trained in Medical Sociology, quite a lot of that support went to beginning work in social epidemiology.
I recall a time during those years when Dr. Lester Breslow applied to the NIH for money to support the establishment of what he called a Human Population Laboratory in Alameda County, California. His idea was to do research in a large representative sample of an entire county over a long period of time to study what he called their health in relation to their way of living. What disease was he going to focus on? None. He had been influenced by the writings of John Cassel and others suggesting that an appropriate outcome for studies of social factors might be "health and disease" in general and not one or another specific disease. This idea was later eloquently presented in the last paper Cassel wrote before he died. I am referring to his classic contribution published in 1976 in the American Journal of Epidemiology called "The Contribution of the Social Environment to Host Resistance" [6].
The NIH was not sure how to deal with Dr. Breslow's proposal because it did not neatly fit into any of the disease-specific institutes. It turned out that there was no institute at the National Institutes of Health that dealt with health. Of course, this is still the case. I was asked for my advice on how to handle this very unusual application. I suggested that we develop a special study section with specially chosen people to deal with this crisis and I was given permission to proceed. To create this special review committee, I invited a few people from my Study Section to serve, as well as some carefully chosen outsiders. I attempted to pick people who I thought could understand the radical idea that Dr. Breslow was proposing.
We went out to California for a two-day meeting. In the end, my specially picked people recommended that the proposal not be funded. It was too weird. For example, Breslow proposed to study the health of people but he was not going to do one physical exam or take any blood or urine. He was simply going to ask people to rate their own health! I recall he proposed a question that asked "Compared to other people your age, how would you rate your health? Excellent, good, fair, or poor?" This question has turned out to be one of the most powerful predictors of future health in dozens and dozens of studies but at that time it was a very bizarre question indeed. That the Breslow proposal was turned down was very disappointing but I urged Dr. Breslow to resubmit and a year later I assembled yet another group of specially picked reviewers to give it another try. And this time it worked. So Dr. Breslow was able to establish this crucial population study that has turned out to be one of the most significant studies in the history of social determinants [7]. A few years later, Dr. Tommy Francis of the School of Public Health in Ann Arbor was able to establish a similar population laboratory in Tecumseh, Michigan.
After three years of serving as Executive Secretary, I returned to do research in the Heart Disease Control Program, this time based in San Francisco, California. By then, it was clear that a field of research in social epidemiology was emerging, most of it focused on coronary heart disease. This research was not of very high quality, nor were the results compelling, but some interesting questions were beginning to emerge. I had a conversation with Professor Leo Reeder about this and we decided it might be good to bring together all the people who were engaged in this research to see what we were all doing and to think about next steps. Since I was a government employee, I prevailed upon my bosses to provide funds for the meeting. The Conference was held in Phoenix, Arizona in February 1966. We invited all of the social scientists and medical people in the country doing research on heart disease as well as some others who, while they were not doing such research, were nevertheless bright and potentially helpful. We scoured the country and came up with 27 people, including Reeder and myself.
The report of our conference was later published in 1967 as a special volume of the Milbank Memorial Quarterly with the title "Social Stress and Cardiovascular Disease" [8]. It is a little embarrassing to read the book now and see the state of the art at that time but it was quite clear that, in spite of this, something important was happening. It was in one of the last papers of this little book, by the way, that I explored the issue of appropriate outcomes for social epidemiologic research. I argued in that piece, no doubt influenced by Lester Breslow's idea for the Alameda County Study, that we needed to look at a broader set of disease outcomes than the usual clinical entities. This idea stands as one of the key features in John Cassel's classic paper as well.
In 1968, I became a Professor of Epidemiology in the School of Public Health at Berkeley. I was, I think, the first sociologist to hold a position as an epidemiologist at any School of Public Health in the world. Leo Reeder was a sociologist at the UCLA School of Public Health but his was a normal position as a Professor of Behavioral Science. By then I was working with Reuell Stallones, who was also a Professor at Berkeley, to study coronary heart disease and stroke among Japanese migrants to Hawaii and California. Stallones was primarily interested in testing the dietary hypothesis. Did the Japanese in Japan have low rates of CHD due to their low fat diet? I was interested in testing the mobility hypothesis. Did rates of CHD go up among the migrants? We were both surprised by the findings.
It turned out that Japanese men who migrated to California had CHD rates five times higher than those in Japan, with migrants to Hawaii having intermediate rates. And this increase in CHD rate was not explained by any of the usual CHD risk factors such as diet, serum cholesterol, smoking or blood pressure. I assigned a doctoral student to figure out what was going on. Michael Marmot did his doctoral dissertation on this issue. He concluded that those Japanese men who had adopted Western cultural ways were the ones with the enormous increase in CHD while those California Japanese who had retained traditional ways had rates comparable to those still living in Japan [9]. Again, this observation was independent of diet and all the usual CHD risk factors. This clearly was not supportive of the mobility hypothesis since some migrants had no health consequence at all. Then Marmot left Berkeley to go to London to begin his work on the British civil servants and he left me with the question: what does it mean to say "Western ways" versus "Traditional ways"?
I went to Japan several times, interviewed dozen and dozens of people, and I read many books to get some understanding of this but all I could get out of this work was that my Japanese informants thought that Americans were lonely. I challenged this observation many times but dozens of people said that anyone could easily see this loneliness when you saw so many Americans walking on the street, alone. "Alone on the street?" I said. "That's not evidence of loneliness", I said. People all shrugged at my naiveté. So I returned to Berkeley, this was in 1975, and found another doctoral student who agreed to work on this. That student, Lisa Berkman, had already been thinking about the importance of social networks and social support and took took my loose and primitive question and reshaped it into a brilliant and elegant study showing the health consequences of social connection. For this study, she used data from the Alameda County Human Population Laboratory that Lester Breslow and I had worked so hard to get funded many years before. My view is that Berkman's study, published in the American Journal of Epidemiology in 1979, really began to establish the field of social determinants [10]. Her findings were later replicated by James House and his group using data from the Tecumseh study [11]. This was a finding that resonated with common experience and that fit with many of the empirical observations we had been making over the years. And it was entirely consistent with one of the most important contributions ever made in social epidemiology: the work of Emile Durkheim on Suicide [12]. If one is going to talk about the roots of a movement, it is crucial to put Berkman's and Durkheim's work at the center.
In my classes at Berkeley, I provoke all first year students by assigning them to read a significant number of pages from Durkheim's work on suicide. Physicians are especially challenged. Here, in suicide, is a study of one of the most intimate and personal behaviors that can be imagined. Surely, Durkheim notes, this behavior can only be explained by understanding the most intimate personal events in someone's life. And yet, he points out, there is a patterned regularity in suicide rates, over time, in various groups. Some groups have characteristically high or low rates of suicide, over time, even as individuals come and go from these groups. If the causes of suicide are to be found within the individual, he asks, how can there be a patterned regularity in groups over time even as individuals come and go from these groups? There must be something about the groups themselves that causes a higher or lower rate. That something would not explain why only some individuals succumb to the social fact but it would explain the difference in group rates. A better description of the role of social epidemiology does not exist. Berkman's work on social connections was the first modern empirical demonstration of Durkheim's genius. And since her work, of course, the importance of social networks has become a recognized international fact. And it has led me, and others, to think about such concepts as control and other similar factors that might explain inequalities in disease by social class.
The Alameda County Human Population Laboratory has been useful for other important work as well. George Kaplan and Mary Haan used data from this study to do their pioneering work showing that certain neighborhoods had higher and lower rates because of their poverty status and to show that these differences could not be explained by the characteristics of individuals living in those areas [13]. Another Durkheim legacy. Others who worked with this data set were Jack Guralnik, now at the National Institute of Aging, John Lynch, now at the University of Michigan, and Teresa Seeman, now at UCLA. And most recently, Irene Yen. And there were others. The special Study Section authorized by the NIH clearly made an important contribution. And the findings obtained from the Tecumseh Human Population Laboratory, another legacy of the Study Section, has also been impressive.
This is a very sketchy and highly selective personal set of observations about the early years of the movement as I experienced it. I have left out a lot and I have undoubtedly ignored the work of many others at work at that time. The result of all of these efforts, however, is revealed today not only by the appearance of the new books mentioned earlier and by the Robert Wood Johnson program but also by the fact that both the National Institutes of Health and the Centers for Disease Control are emphasizing work in this area under the rubric of "disparities". In addition, the Canadian government has reorganized its grant-giving mechanisms to recognize this work by establishing a new Institute of Population and Public Health. To me, it is amazing to see the changes that have occurred in the last 40 or 50 years.
What is the explanation for this phenomenon? In my experience, these changes have not come easily. In fact, my experience has been that they have come very grudgingly, with great suspicion and wariness. This suspicion, it has seemed to me, was based on the following issues: First, many felt that social determinants were vague, and ill-defined concepts based on poor (that is non-experimental) research. Second, even if research findings were shown to be well documented, it was very difficult to imagine how these social factors could "get into the body" to cause disease. Third, even if associations between social factors and disease were well documented and even if a disease mechanism could be imagined, it was difficult, if not impossible, to see how these factors could be intervened upon. Current work on social determinants is focused on these very issues and with very promising results.
Conclusion
As I look back over the last 50 years, I am enormously impressed, and a little surprised, at the positive changes that have taken place in our work to improve health and well-being. These changes began very slowly many years ago by a relatively small group of people but the pace of change is increasing exponentially. Because it is the right thing to do. And because we really have no choice: We all know that our medical care system is under enormous strain. And we all know that the baby-boomers will enter the over age 65 group between 2020 and 2030. When they do, the number of old people in our country will have doubled. If we think the medical care system is under stress now, we will soon see the system burdened even more dramatically. Our only hope is to develop better programs to prevent disease in the first place and not merely wait to fix things after the fact. And to develop appropriate programs of prevention, we are going to need vital, vigorous and creative research and intervention activities that are firmly rooted in social epidemiology. It would be fun to look back 50 years from now to see how all this works out.
==== Refs
Syme SL Frohlich KL The contribution of social epidemiology: Ten new books Epidemiology 2001 13 110 112 10.1097/00001648-200201000-00018
Berkman LF Kawachi I Social Epidemiology 2000 New York: Oxford University Press
Syme SL Hyman MM Enterline PE Some social and cultural factors associated with the occurrence of coronary heart disease J Chronic Dis 1964 17 277 289 10.1016/0021-9681(64)90155-9
Syme SL Hyman MM Enterline PE Cultural mobility and the occurrence of coronary heart disease J Health Hum Behav 1965 6 178 189 5841841
Krieger N Epidemiology and the web of causation: Has anyone seen the spider? Soc Sci Med 1994 39 887 903 7992123 10.1016/0277-9536(94)90202-X
Cassel J The contribution of the social environment to host resistance Am J Epidemiol 1976 104 107 123 782233
Berkman LF Breslow L Health and Ways of Living: The Alameda County Study 1983 New York: Oxford University Press
Syme SL Reeder LC eds Social Stress and Cardiovascular Disease 1967 45 New York: Milbank Memorial Fund Quarterly
Marmot MG Syme SL Acculturation and coronary heart disease in Japanese Americans Am J Epidemiol 1976 104 225 247 961690
Berkman LF Syme SL Social networks, host resistance and mortality: A nine- year follow-up of Alameda County residents Am J Epidemiol 1979 109 186 204 425958
House JS Robbins C Metzner HL The association of social relationships and activities with mortality: Prospective evidence from the Tecumseh community health study Am J Epidemiol 1982 116 123 140 7102648
Durkheim E Suicide: A Study in Sociology 1897 Glencoe Ill: Free Press 1951
Haan MN Kaplan GA Camacho T Poverty and health: Prospective evidence from the Alameda County Study Am J Epidemiol 1987 125 989 998 3578257
| 15840175 | PMC1087863 | CC BY | 2021-01-04 16:36:38 | no | Epidemiol Perspect Innov. 2005 Apr 19; 2:2 | utf-8 | Epidemiol Perspect Innov | 2,005 | 10.1186/1742-5573-2-2 | oa_comm |
==== Front
Genet Vaccines TherGenetic Vaccines and Therapy1479-0556BioMed Central London 1479-0556-3-31584273210.1186/1479-0556-3-3ResearchEvaluation of the VP22 protein for enhancement of a DNA vaccine against anthrax Perkins Stuart D [email protected] Helen C [email protected] Helen S [email protected] Angela E [email protected] Freda K [email protected] Robert J [email protected] Biomedical Sciences Department, Defence Science and Technology Laboratory, Porton Down, Salisbury, Wiltshire, SP4 OJQ, UK2 Tenovus Laboratory, University of Southampton Hospital NHS Trust, Southampton, SO16 6YD, UK2005 20 4 2005 3 3 3 13 1 2005 20 4 2005 Copyright © 2005 Perkins et al; licensee BioMed Central Ltd.2005Perkins 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
Previously, antigens expressed from DNA vaccines have been fused to the VP22 protein from Herpes Simplex Virus type I in order to improve efficacy. However, the immune enhancing mechanism of VP22 is poorly understood and initial suggestions that VP22 can mediate intercellular spread have been questioned. Despite this, fusion of VP22 to antigens expressed from DNA vaccines has improved immune responses, particularly to non-secreted antigens.
Methods
In this study, we fused the gene for the VP22 protein to the gene for Protective Antigen (PA) from Bacillus anthracis, the causative agent of anthrax. Protective immunity against infection with B. anthracis is almost entirely based on a response to PA and we have generated two constructs, where VP22 is fused to either the N- or the C-terminus of the 63 kDa protease-cleaved fragment of PA (PA63).
Results
Following gene gun immunisation of A/J mice with these constructs, we observed no improvement in the anti-PA antibody response generated. Following an intraperitoneal challenge with 70 50% lethal doses of B. anthracis strain STI spores, no difference in protection was evident in groups immunised with the DNA vaccine expressing PA63 and the DNA vaccines expressing fusion proteins of PA63 with VP22.
Conclusion
VP22 fusion does not improve the protection of A/J mice against live spore challenge following immunisation of DNA vaccines expressing PA63.
==== Body
1.0 Background
The VP22 protein is a major component of the amorphous tegument region of the Herpes Simplex Virus type I (HSV-1). Composed of 301 amino acids, it has become known as a protein transduction domain able to mediate intercellular spread. Like other translocatory proteins such as antennapedia and the HIV Tat protein, it is highly basic, it is able to bind heparin or sialic acid and all three proteins have an almost identical predicted pI [1]. VP22 has been reported as being able to exit the cell in which it is synthesised via an uncharacterised, golgi-independent secretory pathway and subsequently enter surrounding cells by a non-endocytic mechanism. These properties may be retained after fusion to other proteins [2].
The ability to 'piggyback' proteins or peptides into cells may be particularly useful for gene therapy. Thymidine kinase and p53 have benefited from fusion with VP22 [3,4]. VP22 has been fused to proteins and delivered by a viral vector. For example, p53 delivered by an adenovirus vector [5,6], GFP delivered by a lentivirus vector [7] and Human papillomavirus E7 antigen delivered by a Sindbis replicon [8,9] have all proved more effective after VP22 fusion.
However, the ability of VP22 to mediate intercellular spread has been questioned, based on in vitro studies that use methanol fixation. Because methanol dissolves cellular membranes, it may produce an artefact interpreted as cell to cell spread [10]. In further studies, transport could not be detected in live cells [11] and a fusion protein of VP22 and diphtheria toxin A (a single molecule of which is lethal to a cell) could not cross the cell membrane and cause a measurable cytotoxic effect [12]. A critical analysis of the literature has led to the conclusion that the effects of VP22 can be explained by well-established biological principles whereby VP22 causes liberation from cells, possibly by cell death. Following this, the protein may bind to surrounding cells, but does not efficiently penetrate cellular membranes [1].
Irrespective of whether VP22 can mediate intercellular spread however, VP22 can enhance in vivo responses to a number of antigens not only in the context of gene therapy, but also when fused to antigens within a DNA vaccine. This could be particularly useful because although DNA vaccines can offer protection against a wide variety of pathogens in small animal models, their efficacy in larger animal models and primates is insufficient. In this study, we evaluate the potential of VP22 to enhance DNA vaccines against anthrax.
The spore-forming bacterium Bacillus anthracis causes the disease anthrax. The current UK-licensed vaccine is an alum-precipitated filtrate of a B. anthracis Sterne strain culture, administered by the intramuscular route, which occasionally causes some transient reactogenicity in vacinees [13]. The US-licensed vaccine is the Anthrax Vaccine Adsorbed (BioThrax-AVA) vaccine produced from the culture supernatant fraction of the V770-NP1-R strain [14].
The key component in both these vaccines is the protective antigen (PA), which along with lethal factor (LF) and edema factor (EF) forms a tripartite toxin and is one of the virulence factors of the bacteria [15]. Host cell intoxication is thought to occur after binding of the full length 83 kDa PA to the host cell membrane receptor. The 20 kDa N-terminal fragment of PA is cleaved by furin protease exposing the LF-EF binding site [16]. The 63 kDa PA fragments form a heptameric pore, the LF or EF bind and the whole toxin complex is internalised [17,18].
DNA vaccines against B. anthracis expressing either the 63 kDa fragment of PA [19,20] or the 83 kDa PA protein have proved successful [21]. Protection against lethal toxin challenge in Balb/c mice or a spore challenge in NZW rabbits can be achieved by either intramuscular or gene gun immunisation [19-22]. Attempts to enhance the protective efficacy of DNA vaccines against anthrax include co-administration with a DNA vaccine expressing LF, and a DNA prime / protein boost regimen [20] or the use of cationic lipids [22].
The aim of this study was to assess the potential of VP22 to enhance the immunogenicity of a DNA vaccine expressing the 63 kDa fragment of PA (PA63) attached to a secretion signal. The VP22 protein, which has previously been shown to improve the performance of DNA vaccines [23-26], was fused to either the N- or the C-terminus of PA63. We show that following gene gun administration of these vaccines, fusion with VP22 does not improve anti-PA antibody responses to the PA63 DNA vaccine, nor does it increase protection against anthrax lethal spore challenge.
2.0 Methods
2.1 Construction of DNA vaccines
The DNA vaccine pGPA contains the signal sequence for human plasminogen activator fused to the N-terminus of the gene for the 63 kDa fragment of PA [19] and was a kind gift from Dennis Klinman (Food and Drug Administration, USA). To include the VP22 sequence derived from amino acids 159 – 301, which possesses the full transport activity of the native protein (both the intrinsic transport ability and the ability to carry proteins of significant size [27]), the following strategy was employed. To construct the N-terminal fusion, the gene for the VP22 sequence was PCR amplified from pCR®T7/VP22-1 (Invitrogen) using primers VP22 F9 (5' ACTCTAGCTAGCACGGCGCCAACCCGATCCAAGACA 3') and VP22 R8 (5' ATTGTCACGGTCTGGAACCGTAGGAGCAGCTGGACCTGGACCCTCGACGGGCCGTCTGGGGCGAGA 3'). Additionally, the gene for PA63 was PCR amplified from pGPA using primers PA F8 (5' CCTACGGTTCCAGACCGTGACAAT 3') and PA R9 (5' CGCGGATCCTTATCCTATCTCATAGCC 3'). The two sequences were then fused together by PCR [28] using primers VP22 F9 and PA R9.
To create the C-terminal fusion, the gene for the PA63 sequence was PCR amplified from pGPA using primers PA F11 (5' CTAGCTAGCCCTACGGTTCCAGACCGTGACAAT 3') and PA R10 (5' TGTCTTGGATCGGGTTGGCGCCGTAGCAGCTGGACCTGGACCTCCTATCTCATAGCC 3'). The gene for VP22 was PCR amplified from pCR®T7/VP22-1 (Invitrogen) using primers VP22 F10 (5' CGGCGCCAACCCGATCCAAGACA 3') and VP22 R11 (5' CGCGGATCCTTACTCGACGGGCCGTCTGGGGCGAGA 3'). The two sequences were then fused together using VP22 F10 and VP22 R11 by PCR fusion [28]. The PCR primers used to create the two gene fusions were designed to incorporate a linker sequence of Gly-Pro-Gly-Pro-Ala-Ala between the VP22 and PA63 proteins, to allow folding of the fusion protein. Using restriction sites NheI and BamHI, the PA63 gene was exised from pGPA and the gene fusions were ligated into the vector to form pSTU-22-PA (N-terminal fusion) and pSTU-PA-22 (C-terminal fusion). These constructs were verified by sequencing and are schematically represented in figure 1.
Figure 1 DNA vaccines constructed in this study as in section 2.1. DNA vaccine expressing PA63 (pGPA) is a kind gift from Dennis Klinman (Food and Drug Administration, USA). (Abbreviations: PCMV, CMV promoter; Sig, Signal sequence; BGH polyA, Bovine growth hormone polyadenylation signal).
The control DNA vaccine expressing VP22 only (pSTU22) has been previously described [26]. The plasmid DNA was prepared using Qiagen Endofree DNA purification columns (Qiagen Ltd).
2.2 Western blot analysis of expressed proteins
African Green Monkey Kidney COS-7 cells (European Collection of Animal Cell Cultures, Porton Down) were plated at 1–5 × 105 cells well-1 into 6 well plates (Corning). Cells were transfected with 1 μg of plasmid DNA using the transfection reagent Polyfect (Qiagen) according to the manufacturer's guidelines. Transfected cell lysates were separated by 4–20% polyacrylamide gel electrophoresis (Tris-Glycine gel, Invitrogen), using XCell SureLock™ Mini-Cell apparatus (Invitrogen) according to the manufacturer's protocol. Protein from the gel was then transferred to nitrocellulose by electroblotting (Invitrogen). An ECL Western blotting kit (Amersham Biosciences) was used with antibody to PA (rabbit polyclonal sera) or VP22 (rabbit polyclonal sera) to detect expression from DNA vaccines.
2.3 Vaccination of Balb/c mice
Groups of 10 female A/J mice (Harlan OLAC) were immunised with 1 μg of DNA coated onto gold particles and delivered using a Helios™ gene gun (BioRad) as described previously [29]. Mice were immunised three times at two-week intervals. Blood was taken from the tail vein prior to challenge for serum antibody analysis by enzyme-linked immunosorbent assay (ELISA).
2.4 Measurement of anti-PA antibodies by ELISA
Microtitre plates were coated with 5 μg ml-1 recombinant PA (Aldevron) in phosphate-buffered saline using 50 μl well-1 and incubated overnight at 4°C. Three columns on each plate were coated with anti-IgG (Fab) (Sigma) in order to produce a standard curve for quantification of IgG concentration. After washing three times with PBS containing 0.2% Tween-20, non-specific binding was blocked with 5% (w/v) powdered skimmed milk in PBS and the plates were incubated for 2 hours at 37°C. The plates were washed three times and serum was added at a starting dilution of 1:50 in blocking buffer, and double-diluted down the plate. IgG or isotype standards (Sigma), diluted in blocking buffer, were added to wells which had been coated with anti-IgG (Fab) (Sigma), and double diluted as before. Plates were incubated for 1.5 hours at 37°C before washing and the addition of goat anti-mouse IgG (or anti-mouse IgG isotype) conjugated to horseradish peroxidase (Sigma), diluted in blocking buffer. Plates were incubated for 1 hour at 37°C, then washed 3 times before addition of the substrate ABTS (Sigma). Absorbance at 410 nm was measured after 20 minutes incubation at room temperature and analysed using Ascent software.
2.5 Challenge with B. anthracis
Three weeks after the final immunising dose, mice were challenged intraperitoneally with B. anthracis STI (Tox+ Cap-) spores. Sufficient spores for the challenge were removed from stock cultures, washed in sterile distilled water, and resuspended in PBS to a concentration of 7 × 105 spores ml -1. Mice were challenged with 100 μl volumes containing 7 × 104 spores per mouse (equivalent to 70 50% lethal doses [LD50s] [30]) and were monitored for 18 days post challenge to determine their protected status. Humane endpoints were strictly observed so that any animals displaying a collection of clinical signs that indicated a lethal infection were culled.
2.6 Statistical Methods
One-way ANOVA with Tukey's multiple comparison post analysis test and statistical analysis of survival using the Mantel-Haenszel Logrank test were performed using GraphPad Prism version 3.02 for Windows, GraphPad Software, San Diego, California, USA .
3.0 Results
3.1 In vitro expression of DNA vaccines
DNA vaccines encoding PA63, VP22-PA63, PA63-VP22 or VP22 (Figure 1) were transfected into African Green Monkey Kidney cells (COS-7). Cells were harvested and processed for Western blot analysis 48 hours post transfection. Cells transfected with the PA63-encoding DNA vaccine expressed a protein of approximately 68 kDa that reacted with PA-specific antibody. Fusion of VP22 to either the N-terminal or C-terminal of PA63 resulted in a protein of approximately 90 kDa that reacted with both PA-specific antibody and VP22-specific antibody (Figure 2). Control cells, transfected with plasmid DNA expressing VP22 only expressed a protein of approximately 22 kDa that reacted with VP22-specific antibody. Some degradation of the PA63 proteins was evident irrespective of whether fused to VP22 or not. However, the degraded fusion proteins were recognised by both the anti-PA and anti-VP22 antibodies suggesting that this degradation was not due to instability at the point of fusion of the two proteins.
Figure 2 Western blot analysis of DNA vaccines. Membranes were probed with anti-PA antibody (A) or anti-VP22 antibody (B) as described in section 2.2. Cells were untransfected (1) or transfected with DNA vaccines expressing VP22 (2), PA63 (3), VP22-PA63 (4) or PA63-VP22 (5).
3.2 Anti-PA antibody responses following gene gun immunisation
Groups of 10 female A/J mice were immunised three times by gene gun administration of 1 μg plasmid DNA at two weeks intervals. Serum samples were collected 17 days after the third immunisation (4 days before challenge). Sera from individual mice were assayed for PA-specific total IgG (Figure 3). Mice immunised with PA63-expressing DNA vaccine produced a mean titre of 27,216 ng/ml total PA-specific IgG, compared with 18,823 ng/ml and 19,448 ng/ml for the VP22-PA63 and PA63-VP22 -expressing DNA vaccines respectively. These antibody titres of PA-specific total IgG did not differ significantly between the three groups (p > 0.05, One-way ANOVA with Tukey's multiple comparison posthoc analysis).
Figure 3 A/J mice were immunised with DNA vaccines expressing PA63, VP22-PA63 or PA63-VP22. Anti-PA total IgG levels in the sera at day 38 were determined by ELISA. Bars represent the mean of each group, the error bars represent 95% confidence intervals. n = 10 mice per group.
3.3 Protection against anthrax spore challenge
Mice were challenged three weeks after the final dose with 70 50% lethal doses of B. anthracis strain STI by the intraperitoneal route. The DNA vaccine expressing PA63 conferred 70% survival to the immunised mice. In comparison, 80% and 50% of the mice survived following immunisation with the DNA vaccines expressing VP22-PA63 and PA63-VP22, respectively (Figure 4). Thus inclusion of the VP22 protein at either the N- or C-terminus of PA63 did not significantly alter protection of the mice. All three vaccines offered a significant level of protection compared to naïve mice. Statistical analysis of survival was performed using the Mantel-Haenszel Logrank test (GraphPad Prism).
Figure 4 Numbers of mice surviving 18 days post challenge with 70 LD50s of B. anthracis STI spores after immunisation with DNA vaccines expressing PA63, VP22-PA63 or PA63-VP22. n = 10 mice per group.
4.0 Discussion
The Herpes Simplex virus type I VP22 protein has been suggested to mediate intercellular spread by exit from cells in a golgi-independent manner and entry to adjacent cells by a non-endocytic mechanism [2]. However, in vitro studies of this protein remain inconclusive, with reports that the apparent effects of VP22 can be attributed to an artefact produced by methodology [1,10,31,32]. Despite the controversy surrounding in vitro studies, most in vivo work shows that VP22 has a beneficial effect, particularly in the gene therapy field. Fusion of VP22 to either the pro drug-activating enzyme thymidine kinase [3] or the transcription factor p53 [5] results in an improvement in their effectiveness. Similarly, inclusion of this protein within a DNA vaccine can increase immune responses. Antigens shown to benefit from fusion to VP22 include Yellow Fluorescent Protein (YFP) [24], Enhanced Green Fluorescent Protein (EGFP) [26] and the human papillomavirus (HPV) E7 protein [23,25,33,34].
In this study, the VP22 protein has been fused to either the N- or C-termini of the Protective Antigen (PA) of B. anthracis. This antigen expressed from a DNA vaccine is protective against challenge with either lethal toxin (PA plus LF) in Balb/c mice [19,20] or spore challenge in New Zealand white rabbits [21]. We used an immunisation regimen and challenge dose of STI spores designed to offer significant but not full protection to anthrax challenge of A/J mice. This design would allow us to demonstrate any increased protection due to fusion of VP22 to PA within the DNA vaccine. Our results showed that the fusion of VP22 with PA63 at either terminus failed to significantly enhance anti-PA antibody responses compared to the PA63 DNA vaccine. Following challenge, all three DNA vaccines expressing either PA63, VP22-PA63 or PA63-VP22 offered significant protection against 70 LD50's of B. anthracis compared to unimmunised control mice. However, the inclusion of VP22 did not significantly increase or decrease the protection afforded when compared to PA63-expressing DNA vaccine alone. This suggests that the fusion of VP22 to either the N- or C-terminus of PA63 within a DNA vaccine, does not alter either the antibody response elicited in vivo or the protection afforded to A/J mice following spore challenge. The longevity of the immune response or the ability of these DNA vaccines to initiate long-term protection was not evaluated in this study.
The failure of VP22 fusion to increase antibody responses to PA63 contrasts with other YFP, EGFP or HPV E7 antigens expressed from DNA vaccines where improvement is evident [23-26]. Furthermore, the failure to increase protection against B. anthracis challenge contrasts with studies involving DNA vaccines expressing HPV E7 protein where protective anti-tumour immunity was increased with VP22 fusion [33,34]. However, the DNA vaccines expressing the reporter proteins YFP and EGFP lack secretion signals and the level of enhancement afforded to the HPV E7 protein following fusion with VP22 was equivalent to that afforded by inclusion of a secretion signal [23]. The PA63-expressing DNA vaccine used here does contain a secretion signal.
The inclusion of a secretion signal is a commonly used strategy for DNA vaccination as liberation of the protein from the cell can increase immune responses [35-37]. The inclusion of VP22 within a DNA vaccine may enable non-secreted proteins to exit the cell thus increasing their exposure to antigen presenting cells such as dendritic cells. This is consistent with the hypothesis that VP22 does not mediate intercellular spread as first described, but rather is liberated from cells possibly by cell death [1]. Apart from liberation of the expressed protein from the cell, VP22 may enhance DNA vaccines in other ways. For example, the fusion of immunostimulatory sequences to antigens expressed from DNA vaccines has been shown to provide cognate T cell help [38]. In this study, a DNA fusion vaccine against B cell tumours uses the non-toxic C fragment of tetanus toxin. So it is possible that fusion of VP22 to antigens encoded by DNA vaccines may improve immunogenicity by provision of cognate T cell help.
5.0 Conclusion
This study investigates the inclusion of the VP22 protein in a DNA vaccine expressing PA63 of B. anthracis. The VP22 protein has been shown previously to enhance the performance of DNA vaccines expressing non-secreted proteins. In this case, the PA63-expressing DNA vaccine contains the human plasminogen activator signal sequence [19]. Inclusion of VP22 within this DNA vaccine construct did not enhance anti-PA antibody responses or offer an increase in the level of protection afforded to A/J mice following anthrax spore challenge. This suggests that although VP22 can improve responses to DNA vaccines encoding non-secreted proteins, it does not improve responses to a PA63-expressing DNA vaccine encoding a secretion signal.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
SDP, HCF-S, HSG, AEE-L carried out the studies. FKS, RJP participated in the design of the study. All authors read and approved the final manuscript.
Acknowledgements
The authors would like to acknowledge the Tenovus Laboratory (Southampton University Hospitals Trust) and the Leukaemia Research Fund. Thanks also to Emma Waters, Steve Elvin, Tony Stagg, Warren Kitchen, Stefan Mills, Sarah Hayes, Sara Browning, Angela Scutt and Clare Burton for excellent technical assistance. Thanks also to Helen Burnell for advice and Lyn O'Brien for proof reading.
==== Refs
Leifert JA Whitton JL "Translocatory proteins" and "protein transduction domains": a critical analysis of their biological effects and the underlying mechanisms Mol Ther 2003 8 13 20 12842424 10.1016/S1525-0016(03)00151-5
Elliott G O'Hare P Intercellular trafficking and protein delivery by a herpesvirus structural protein Cell 1997 88 223 233 9008163 10.1016/S0092-8674(00)81843-7
Dilber MS Phelan A Aints A Mohamed AJ Elliott G Smith CI O'Hare P Intercellular delivery of thymidine kinase prodrug activating enzyme by the herpes simplex virus protein, VP22 Gene Ther 1999 6 12 21 10341871 10.1038/sj.gt.3300838
Phelan A Elliott G O'Hare P Intercellular delivery of functional p53 by the herpesvirus protein VP22 Nat Biotechnol 1998 16 440 443 9592391 10.1038/nbt0598-440
Wills KN Atencio IA Avanzini JB Neuteboom S Phelan A Philopena J Sutjipto S Vaillancourt MT Wen SF Ralston RO Johnson DE Intratumoral spread and increased efficacy of a p53-VP22 fusion protein expressed by a recombinant adenovirus J Virol 2001 75 8733 8741 11507218 10.1128/JVI.75.18.8733-8741.2001
Zender L Kock R Eckhard M Frericks B Gosling T Gebhardt T Drobek S Galanski M Kuhnel F Manns M Kubicka S Gene therapy by intrahepatic and intratumoral trafficking of p53-VP22 induces regression of liver tumors Gastroenterology 2002 123 608 618 12145813 10.1053/gast.2002.34756
Lai Z Han I Zirzow G Brady RO Reiser J Intercellular delivery of a herpes simplex virus VP22 fusion protein from cells infected with lentiviral vectors Proc Natl Acad Sci U S A 2000 97 11297 11302 11027330 10.1073/pnas.97.21.11297
Cheng WF Hung CH Chai CY Hsu KF He L Ling M Wu TC Enhancement of sindbis virus self-replicating RNA vaccine potency by linkage of herpes simplex virus type 1 VP22 protein to antigen J Virol 2001 75 2368 2376 11160740 10.1128/JVI.75.5.2368-2376.2001
Cheng WF Hung CF Hsu KF Chai CY He L Polo JM Slater LA Ling M Wu TC Cancer Immunotherapy Using Sindbis Virus Replicon Particles Encoding a VP22--Antigen Fusion Hum Gene Ther 2002 13 553 568 11874633 10.1089/10430340252809847
Lundberg M Johansson M Is VP22 nuclear homing an artifact? Nat Biotechnol 2001 19 713 714 11479552 10.1038/90741
Elliott G O'Hare P Intercellular trafficking of VP22-GFP fusion proteins Gene Ther 1999 6 149 151 10341888 10.1038/sj.gt.3300850
Falnes PO Wesche J Olsnes S Ability of the Tat basic domain and VP22 to mediate cell binding, but not membrane translocation of the diphtheria toxin A-fragment Biochemistry 2001 40 4349 4358 11284691 10.1021/bi002443l
McBride BW Mogg A Telfer JL Lever MS Miller J Turnbull PC Baillie L Protective efficacy of a recombinant protective antigen against Bacillus anthracis challenge and assessment of immunological markers Vaccine 1998 16 810 817 9627938 10.1016/S0264-410X(97)00268-5
Ivins BE Pitt ML Fellows PF Farchaus JW Benner GE Waag DM Little SF Anderson GWJ Gibbs PH Friedlander AM Comparative efficacy of experimental anthrax vaccine candidates against inhalation anthrax in rhesus macaques Vaccine 1998 16 1141 1148 9682372 10.1016/S0264-410X(98)80112-6
Mikesell P Ivins BE Ristroph JD Dreier TM Evidence for plasmid-mediated toxin production in Bacillus anthracis Infect Immun 1983 39 371 376 6401695
Klimpel KR Molloy SS Thomas G Leppla SH Anthrax toxin protective antigen is activated by a cell surface protease with the sequence specificity and catalytic properties of furin Proc Natl Acad Sci U S A 1992 89 10277 10281 1438214
Friedlander AM Macrophages are sensitive to anthrax lethal toxin through an acid-dependent process J Biol Chem 1986 261 7123 7126 3711080
Petosa C Collier RJ Klimpel KR Leppla SH Liddington RC Crystal structure of the anthrax toxin protective antigen Nature 1997 385 833 838 9039918 10.1038/385833a0
Gu ML Leppla SH Klinman DM Protection against anthrax toxin by vaccination with a DNA plasmid encoding anthrax protective antigen Vaccine 1999 17 340 344 9987172 10.1016/S0264-410X(98)00210-2
Price BM Liner AL Park S Leppla SH Mateczun A Galloway DR Protection against anthrax lethal toxin challenge by genetic immunisation with a plasmid encoding the lethal factor protein Infect Immun 2001 4509 4515 11401993 10.1128/IAI.69.7.4509-4515.2001
Riemenschneider J Garrison A Geisbert J Jahrling P Hevey M Negley D Schmaljohn A Lee J Hart MK Vanderzanden L Custer D Bray M Ruff A Ivins B Bassett A Rossi C Schmaljohn C Comparison of individual and combination DNA vaccines for B. anthracis, Ebola virus, Marburg virus and Venezuelan equine encephalitis virus Vaccine 2003 21 4071 4080 12922144 10.1016/S0264-410X(03)00362-1
Hermanson G Whitlow V Parker S Tonsky K Rusalov D Ferrari M Lalor P Komai M Mere R Bell M Brenneman K Mateczun A Evans T Kaslow D Galloway D Hobart P A cationic lipid-formulated plasmid DNA vaccine confers sustained antibody-mediated protection against aerosolized anthrax spores Proc Natl Acad Sci U S A 2004 101 13601 13606 15342913 10.1073/pnas.0405557101
Michel N Osen W Gissmann L Schumacher TN Zentgraf H Muller M Enhanced Immunogenicity of HPV 16 E7 Fusion Proteins in DNA Vaccination Virology 2002 294 47 59 11886264 10.1006/viro.2001.1321
Oliveira SC Harms JS Afonso RR Splitter GA A genetic immunization adjuvant system based on BVP22-antigen fusion Hum Gene Ther 2001 12 1353 1359 11440628 10.1089/104303401750271002
Hung CF Cheng WF Chai CY Hsu KF He L Ling M Wu TC Improving vaccine potency through intercellular spreading and enhanced MHC class I presentation of antigen J Immunol 2001 166 5733 5740 11313416
Perkins SD Hartley MG Lukaszewski RA Phillpotts RJ Stevenson FK Bennett AM VP22 enhances antibody responses from DNA vaccines but not by intercellular spread Vaccine 2005 23 1931 1940 15734065 10.1016/j.vaccine.2004.10.033
Kueltzo LA Normand N O'Hare P Middaugh CR Conformational lability of herpesvirus protein VP22 J Biol Chem 2000 275 33213 33221 10913125 10.1074/jbc.M002476200
Hobert O PCR fusion-based approach to create reporter gene constructs for expression analysis in transgenic C. elegans Biotechniques 2002 32 728 730 11962590
Bennett AM Phillpotts RJ Perkins SD Jacobs SC Williamson ED Gene gun mediated vaccination is superior to manual delivery for immunisation with DNA vaccines expressing protective antigens from Yersinia pestis or Venezuelan Equine Encephalitis virus Vaccine 1999 18 588 596 10547416 10.1016/S0264-410X(99)00317-5
Beedham RJ Turnbull PC Williamson ED Passive transfer of protection against Bacillus anthracis infection in a murine model Vaccine 2001 19 4409 4416 11483266 10.1016/S0264-410X(01)00197-9
Lundberg M Johansson M Positively charged DNA-binding proteins cause apparent cell membrane translocation Biochem Biophys Res Commun 2002 291 367 371 11846414 10.1006/bbrc.2002.6450
Lundberg M Wikstrom S Johansson M Cell surface adherence and endocytosis of protein transduction domains Mol Ther 2003 8 143 150 12842437 10.1016/S1525-0016(03)00135-7
Hung CF He L Juang J Lin TJ Ling M Wu TC Improving DNA Vaccine Potency by Linking Marek's Disease Virus Type 1 VP22 to an Antigen J Virol 2002 76 2676 2682 11861834 10.1128/JVI.76.6.2676-2682.2002
Kim TW Hung CF Kim JW Juang J Chen PJ He L Boyd DA Wu TC Vaccination with a DNA vaccine encoding herpes simplex virus type 1 VP22 linked to antigen generates long-term antigen-specific CD8-positive memory T cells and protective immunity Hum Gene Ther 2004 15 167 177 14975189 10.1089/104303404772679977
Bennett AM Perkins SD Holley JL DNA vaccination protects against botulinum neurotoxin type F Vaccine 2003 21 3110 3117 12804837 10.1016/S0264-410X(03)00260-3
You Z Huang X Hester J Toh HC Chen SY Targeting dendritic cells to enhance DNA vaccine potency Cancer Res 2001 61 3704 3711 11325842
Rice J King CA Spellerberg MB Fairweather N Stevenson FK Manipulation of pathogen-derived genes to influence antigen presentation via DNA vaccines Vaccine 1999 17 3030 3038 10462238 10.1016/S0264-410X(99)00171-1
King CA Spellerberg MB Zhu D Rice J Sahota SS Thompsett AR Hamblin TJ Radl J Stevenson FK DNA vaccines with single-chain Fv fused to fragment C of tetanus toxin induce protective immunity against lymphoma and myeloma Nat Med 1998 4 1281 1286 9809552 10.1038/3266
| 15842732 | PMC1087864 | CC BY | 2021-01-04 16:39:08 | no | Genet Vaccines Ther. 2005 Apr 20; 3:3 | utf-8 | Genet Vaccines Ther | 2,005 | 10.1186/1479-0556-3-3 | oa_comm |
==== Front
Health Qual Life OutcomesHealth and Quality of Life Outcomes1477-7525BioMed Central London 1477-7525-3-271584770310.1186/1477-7525-3-27ResearchPrevalence and burden of self-reported blindness and low vision for individuals living in institutions: a nationwide survey Brézin Antoine Pierre [email protected] Antoine [email protected] Francis [email protected] Mounir [email protected] Gilles [email protected] Hôpital Cochin – Service d'ophtalmologie – Université Paris 5 – 27 rue du Faubourg Saint-Jacques F-75679 Paris cedex 14, France2 Cemka, 43, Boulevard du Maréchal Joffre, F-92340 Bourg-la-Reine, France3 Université Pierre et marie Curie, 175, rue du Chevaleret, F-75013 Paris France4 Gilles Berdeaux, Alcon France, 4 Rue Henri Sainte Claire Deville, F-92563 Rueil-Malmaison Cedex, France2005 25 4 2005 3 27 27 23 3 2005 25 4 2005 Copyright © 2005 Brézin et al; licensee BioMed Central Ltd.2005Brézin 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 prevalence of self-reported low vision (LV) and blindness, and their associated disabilities, handicaps and socio-economic consequences for individuals living in institutions are poorly documented.
Methods
2,075 institutions were selected at random and eight individuals were picked at random from the list of residents. Three groups of individuals were defined: blind, LV, and a control group (CG). These were compared after adjustment for age and co-morbidities. Of the 15,403 individuals, 14,603 interviews (94.9%) were completed.
Results
The prevalence of blindness was 1.6% and the LV 13.4%. Blind individuals needed assistance more often (OR: 2.65 to 11.35) than CG members while the assistance required by LV individuals was similar to that for the CG. Blind individuals required institution adaptation (building and furniture changes) more often than the CG. Blind (57.9%) and LV individuals (35.4%) were more often registered for social allowances. Monthly social allowances were EUR 86 higher for blind than LV individuals. Monthly family incomes were found to be similar between the three groups (from EUR 782 to 797). Social and demographic data, institution description, income, handicaps, disabilities, social allowances and details of daily activities were collected interviews
Conclusion
The results demonstrate the impact of self-reported blindness and LV on daily life for patients living in institutions.
BlindnessLow visionHandicapIncapacityDependency
==== Body
Background
Aging creates policy challenges for most developed countries that increase pressure on social and care systems [1,2]. An institution is often the final care facility when older people have too many disabilities. In developed countries, it was estimated that 6.5% of individuals older than 65 years lived in institutions in 1994 [3] with an annual growth rate of 0.8%. The number of individuals in French institutions will show an increase of 56.3% by 2020 mainly due to the aging of the population. In 2000, institutionalization represented 0.62% to 2.71% of the GDP (Gross Domestic Product) in OECD countries. Increases of up to 69.5% in the number of institutionalized individuals are expected by 2020 [4].
In 2001, according to the World Health Organization (WHO), visual impairment was responsible for 2,286,000 DALY (disability adjusted life years) in the high income countries. [5]. The prevalence of blindness and low vision (LV) for people living in institutions has never been estimated at a national level as far as we are aware. The use of registers to estimate the prevalence of blindness is controversial since a high proportion of blind individuals are not registered [6-9]. Studies were conducted in institutions [10-12], but never from a representative nationwide sample. None of them documented the level of disability linked to visual impairment. Visual impairment prevalence rates varied from 7.4% to 23.0%, this range being mainly explained by different definitions of visual impairment. Since blindness and/or LV could be one of the impairments leading to institutionalization in elderly individuals, the comparison of prevalence in the community and in institutions is a key issue.
The social and economic consequences of blindness (disability, dependency and need for assistance) have never been evaluated, in France, at a national level with a representative sample of individuals living in institutions. This information is important for several reasons: (1) institutions have to provide the right level of assistance for each different type of impairment, blindness being one of them; (2) a handicap needs to be indemnified (directly or indirectly) and therefore its economic burden have to be estimated to determine the level of social allowances; (3) macro-economic consequences of blindness need to be estimated to determine how much should be budgeted and invested to care for this handicap at a national level in the long term.
The present survey had three aims: (1) to estimate the prevalence of self-reported blindness and LV in French individuals living in institutions, on a national basis; (2) to study the consequences of self-reported blindness and LV, focusing specifically on disabilities (restricted ability, or inability to perform the activities of daily living) and handicaps (restricted ability, or inability to fulfill desired social roles); (3) to collect information on the economic consequences of income level and social allowances.
Methods
This survey was conducted at the request of the French State which also provided the finance. The data were gathered by INSEE (Institut National de la Statistique et des Etudes Economiques) which is the State agency responsible for executing regular national census surveys. This survey complied with all existing national regulations, including personal data privacy and access. The database was subsequently made available to researchers for secondary analyses.
The methodology of this survey has already been described elsewhere[13]. Cluster sampling was used.
2,075 institutions were randomly selected from French Health Ministry files (day care centers were excluded). French institutions are classified into four types according to the nature of the residents: children with handicaps, adults with handicaps, the elderly and psychiatric patients. . The sample was stratified by 4 main categories of institutions (Table 1). The probability of an institution being selected was inversely proportional to the number of institutions in the category, and was proportional to the number of beds. Fifty-seven had to be replaced: 37 had no residents, 18 no longer existed, and the survey was not technically possible in seven centers. One hundred and fifty-five (7.5%) of the institutions refused to participate. Refusal rates varied by type of institution: 6.5% in centers for handicapped children, 4.5% in centers for handicapped adults, 4.5% in elderly care homes and 17.0% in psychiatric centers. The most frequent reasons for refusal to participate were lack of time (22.7%), the non-compulsory (this survey was disconnected from the national census) nature of this INSEE survey (10.7%), lack of staff to help the interviewer (7.3%), disturbance of the residents (5.3%), institution being restructured (3.3%), violation of privacy (2.7%), too many surveys (2.7%), and beliefs that the questionnaire was not adapted to the residents (2.7%).
Table 1 Experimental design. Sampling plan. Category description
Institution (Sampling frame) Individuals (Sampling elements)
Population Sampling rate Population Sampling rate
Children institution 1,206 34.2% 48,398 6.8%
Adult institution 2,405 18.7% 82,852 4.3%
Psychiatric institution 394 79.2% 70,932 3.5%
Elderly care home 7,414 12.0% 490,963 1.4%
Children institution included intellectual, motor and sensorial handicap. Adult institution included individuals able to work outside, inside the institution, need for medical assistance and other type of institution. Psychiatric institution included mental disease specialized, psychiatric, and other type of hospital. Elderly institution included various types of institution, private, public, charity, short and long term, with no, mild and high medical activities.
Of these institutions, 15,403 individuals were taken at random by the interviewers from the resident lists (eight per institution) and 14,611 interviews (94.9%) were performed. The analysis was performed on the 14,603 patients with documentation of handicap (eight interviews were stopped before the handicap was documented).
The questionnaire is available from the INSEE upon request by researchers.
No exclusion criteria were specified (e.g. age, cognitive functioning, etc ..) and proxy responded when required by the health status of the interviewed. The presence of handicap was identified by the following initial "yes/no" question: "In everyday life, are you faced with either physical, sensorial, intellectual or mental difficulties (resulting from an accident, a chronic disease, a problem at birth, a disability, aging ...) ?" Independent of the answer, the following question was asked to list handicaps: "What kind of difficulties, disabilities, or other health problems do you suffer from?" All the answers were written down without any alteration. If, during the rest of the questionnaire, the interviewers came across diseases or other health problems not mentioned earlier, they were also reported. The declared morbidity was then medically coded by experts in this field, using ICD classification. More details on coding, quality control, training are available from the INSEE upon request.
For the purpose of this study, blindness and LV were defined based on individual declaration, independent of any medical information. At the beginning of the interview, individuals were asked whether they experienced physical, sensorial, intellectual or mental difficulties in their daily life, and the nature of the difficulties. During questioning on incapacity, the incapacity had to be related to a handicap; this provided a second chance to capture data on handicaps. Three closed questions dedicated to vision were asked during the interview: (1) Do you have trouble reading newspapers, books, etc ... with your glasses, if you wear any? (2) Do you have trouble recognizing the features of someone standing four meters from you (with eyeglasses or contact lenses, if you usually wear any)? (3) Would you say you are completely blind (light perception at best), partially blind (still have form perception) or visually impaired? Data were collected as free text and experts in medical coding made post-hoc classification of declared disease. Question 1 (near vision) and question 2 (distant vision) were used to identify individuals with vision troubles. Only the latter had to answer to question 3. According to the previous answers, individuals were classified into one of the following groups: (1) blind; (2) low visual acuity; (3) control (i.e. neither blind nor LV). Blind people declared only light perception. Low visual acuity people declared either severe difficulties in long or short distance vision to question 1 and 2 and form perception or visual impairment to question 3. The worst severity in terms of visual impairment (blind >LV >NVP) was always retained.
Disabilities were described according to an ordinal scale following a question started by "Can you ...?". Answers varied according to the handicap, but the general structure was as follows: "Irrelevant", "Yes, without assistance and without any trouble", "Yes without assistance but with some difficulty", "Yes, without assistance, but with much difficulty because of my physical disorders", "No, I need partial assistance", "No, I need assistance for everything", and "Will not answer or does not know". The figures presented in the tables grouped partial assistance and assistance for everything together.
Social allowances covered the following groups of items: allowance for disabled adults, compensatory allowance, income guarantee, special education allowance, housing allowance, special dependency allowance, disablement paid by the State, disablement allowance deriving from an accident at work, daily allowance paid by the French Sick Fund, allowance paid by an insurance company, military disablement allowance, and other. Details within groups were collected. Results were expressed in EUROs. Revenue was defined as the sum of incomes and social allowances.
Data were collected from October to December 1998 by 413 personal interviewers. The mean interview time was 38 minutes. A trained interviewer filled in the questionnaire at the institution using remote data entry computer software. The data included type of handicap and disabilities (if any), indexes of daily activity (Katz, Colvez and EHPA indexes [14,15]), the reasons for and duration of handicaps, social professional and family environment, individuals' social demography, institution characteristics, institution adaptations required for the handicap, mobility difficulties, household income (including social allowances), social allowances and public and private health insurance.
All analyses were conducted with SAS (SAS Institute, North Carolina) software, release 8.2. Since the three sub-groups were not comparable, adjustments were made using a weighted logistic regression for qualitative variables and a weighted analysis of variance for quantitative parameters.
Weighting factors required for national extrapolation included size of the category, the institution occupation rate (number of individuals in the institution / number of available beds) and the answer refusal rate (higher in psychiatric centers).
Populations were made comparable using the blind population as the reference. Adjustments were performed on age and number of co-morbidities since these variables were found to be linked to almost all of the parameters. Rates were adjusted by using the blind population estimates as reference: estimators of the logistic regression were applied to the co-variables estimated on the blind population.
Odds ratios (OR) were calculated by using the control population (i.e. without visual problems) as a reference. All tests were interpreted two-sided, alpha fixed at 0.05. No corrections to take into account test multiplicity were applied.
Results
Out of 14,603 individuals interviewed (Table 2), 265 were classified as blind, 1,622 had LV and 12,716 did not claim a visual handicap. Upon extrapolation to the whole population (664,252 individuals), the corresponding prevalence figures were 10,394 blind individuals (1.56%) and 89,252 with LV (13.4%).
Table 2 Socio-demographic parameter and handicap description
Blind n = 265 (n = 10,394) Low vision n = 1,622 (n = 89,252) No visual problems n = 12,716 (n = 564,606)
Age (years) 71.4 80.0 67.3
Male 30.3% 26.6% 37.5%
Entry to institution related to health 88.0% 72.8% 77.5%
Institution with medical services 44.2% 31.2% 34.5%
Time in institution (years) 9.2 5.2 5.3
Number of handicaps outside vision 1.62 (1.50) 1.99 (1.82) 1.68 (1.59)
Motor handicap 54.2% (51.0%) 73.5% (68.8%) 51.9% (57.4%)
Auditory handicap 13.9% (18.6%%) 35.0% (26.1%) 22.3% (12.4%)
Vocal handicap 6.0% (4.8%%) 5.3% (5.3%) 5.2% (5.3%)
Visceral handicap 23.4% (21.1%) 35.6% (28.8%) 23.5% (22.9%)
Cognitive handicap 42.1% (38.4%) 34.0% (36.0%) 38.5% (40.9%)
Other handicap 26.8% (18.6%) 15.7% (17.4%) 20.1% (23.7%)
Katz classification 'A' 24.4% (20.7%) 40.7% (50.6%) 47.5% (45.7%)
Katz 'A' classification: able to wash, dress, go to the restroom, go to bed and stand up, sphincter control, eat meal already prepared, alone. Number of handicaps outside vision adjusted on age (2nd line, figures within brackets). Katz 'A' classification adjusted on age, and number of handicaps outside vision (3rd line, figures within brackets).
On average, LV individuals were older than blind and control individuals (80.0, 71.4 and 67.3 respectively). About one-third of the individuals were male. Blind people utilized an institution with medical services more frequently than LV and control individuals. The average time spent by blind individuals is around four years longer than for LV individuals and controls.
The number of handicaps excluding vision was slightly higher in the LV group and the difference persisted after adjustment. More motor, auditory and visceral handicaps were reported by LV individuals. According to the Katz classification, blind people were able to self-wash, dress, go to the restroom as well as control elimination functions, get-up from bed, and eat without assistance less often than LV and control individuals (24.4%, 50.6% and 47.5% respectively, after adjustment for age and number of co-morbidities).
Almost no difference in need for assistance was found between LV individuals and individuals who declared no problems with vision (Table 3). Blind individuals more often needed assistance to do most of their normal daily tasks (Odds-ratio (OR) between 2.65 and 11.35). This mainly concerned the following activities: shopping (OR: 11.35), getting a drink (10.68), using a lift (7.63), climbing steps (7.19), mobility on one level (7.09) and walking outside (7.02).
Table 3 Assistance and burden for the care giver according to level of vision impairment
Need assistance Blind Low vision No visual problems P-Value
Self-washing 66.2% (2.65) 36.8% (0.79) 42.5% (Ref) <0.0001
Dressing 71.1% (3.39) 36.3% (0.79) 42.0% (Ref) <0.0001
Cutting food 52.0% (4.43) 18.8% (0.95) 19.7% (Ref) <0.0001
Help yourself to a drink 74.9% (10.68) 22.7% (0.99) 22.9% (Ref) <0.0001
Drinking and eating once food is ready 32.0% (3.92) 7.5% (0.67) 10.7% (Ref) <0.0001
Going alone to the restroom 53.5% (3.91) 17.6% (0.72) 22.8% (Ref) <0.0001
Getup from a bed 52.2% (2.87) 23.3% (0.80) 27.5% (Ref) <0.0001
Getup from a seat 41.8% (2.70) 16.4% (0.74) 21.0% (Ref) <0.0001
Mobility on one level 61.6% (7.09) 16.9% (0.90) 18.5% (Ref) <0.0001
Climbing steps 77.4% (7.19) 34.6% (1.11) 32.3% (Ref) <0.0001
Using a lift 81.8% (7.63) 35.1% (0.92) 37.1% (Ref) <0.0001
Walking outside 87.3% (7.02) 53.4% (1.17) 49.5% (Ref) <0.0001
Shopping 97.0% (11.35) 76.9% (1.16) 74.1% (Ref) <0.0001
Adjusted according to the blind French population on age, number of handicaps and size. Odds-ratio within brackets adjusted on age and number of handicaps outside vision.
Institution adaptations (building and furniture changes) were more often required for blind people than LV individuals and controls (54.5%, 39.4%, 38.0%, respectively) (Table 4). Globally the level of unmet needs concerning institution adaptation was rather low (around 1%), although slightly higher for blind individuals (5.8%). Four items were mainly concerned: bed, bathroom, restrooms and ramps.
Table 4 Institution adaptation, equipment according to the vision status
Blind Low vision No visual problems P-Value on needs
Institution adaptation (Need – Unmet need) 54.5% – 5.8% 39.4% – 1.3% 38.0% – 1.1% <0.0001
Restrooms 29.9% 22.1% 20.7% <0.0001
Bathroom 31.3% 19.2% 20.3% <0.0001
Tables 23.6% 11.2% 14.3% <0.0001
Seats 24.2% 11.6% 12.8% <0.0001
Bed 46.5% 25.2% 28.3% <0.0001
Ramps, etc. 25.0% 17.5% 18.6% <0.0001
Devices to open doors, etc. 1.8% 2.1% 1.6% <0.0001
Other pieces of furniture 5.1% 2.5% 1.8% <0.0001
Adjusted according to the blind French population on age, number of handicaps. First figure : Needed institution adaptation; Second figure: unmet need
The need for medical devices increased from individuals who declared no problems with vision to people who declared themselves to be blind (38.9% – no visual trouble -, 48.5%- LV – and 64.8% – blind -) (Table 5). The level of unmet needs was very low (less than 2%). Wheel chairs were by far the most frequently required medical devices.
Table 5 Devices dedicated to blindness
Blind Low vision No visual problems P-Value
Devices (purchased- Unmet Need) 64.8% – 1.5% 48.5% – 1.0% 38.9% – 0.6% <0.0001
Stick 5.6% 18.0% 12.7% <0.0001
White stick 11.5% 2.0% 0.0% <0.0001
Walking aids 1.2% 7.0% 6.2% <0.0001
Wheel chair 36.5% 19.7% 20.3% <0.0001
Dog 0.2% 0.2% 0% 0.98
Optical assistance 0.2% – 3.1% 8.6% – 8.5% 0.5% – 0.8% <0.0001
Computer interface 0.3% – 3.4% 0.1% – 0.4% 0% – 0% <0.0001
Software adapted for blindness 1.0% – 4.5% 0.1% – 1.0% 0% – 0% <0.0001
Tape recorder 2.0% – 1.6% 0.3% – 1.1% 0% – 0% <0.0001
Adjusted according to the blind French population on age, number of handicaps. First figure : purchased devices; Second figure : unmet needs.
Only one-third of the LV group and less than 60% of blind individuals received social allowances (Table 6). The rate of private insurance was slightly higher in LV individuals. Although statistically significant, the difference of monthly average allowances between blind and control individuals was quantitatively small (EUR 80). Finally, no differences were found in monthly average income between the three groups.
Table 6 Working activities, social allowances, copayment and revenue according to the vision status
Blind Low vision No visual problems P-Value
Social allowance 57.9% 35.4% 37.0% <0.0001
Private insurance 54.9% 61.3% 58.9% <0.0001
Monthly average allowances (EUR) 254 168 174 <0.0001
Total monthly average income (EUR) 782 787 797 <0.80
Total monthly average income (EUR)-Paid institution cost 116 119 121 <0.94
Adjusted according to the blind French population on age, number of handicaps and size of the household. Total monthly average income includes social allowances. EUR 1 close to US$ 1 in 2002.
Discussion
The present analysis aimed at estimating the prevalence of self-reported blindness and LV in French individuals living in institutions and studying the disabilities related to visual impairment and its economic consequences. Since the questionnaire and the statistical techniques used in this survey was identical to the one used in a survey conducted on individuals living in the community [16], the comparison of the results between the two studies is rather straightforward.
Prevalence of self-reported blindness was 0.10% in the community and 1.56% in institutions. Respective figures were 1.94% and 13.4% for individuals with LV. The probability of being in an institution was 15.6 times higher for blind people and 6.8 times for LV individuals. These preliminary results suggest that blindness and LV might be associated with institutionalization. These relative risks need to be adjusted for the presence of other handicaps in order to confirm whether blindness and LV are independent risk factors for institutionalization.
Individuals with LV living in the community were younger (61.4 versus 80.0) than those living in institutions [16] and had less co-morbidity (1.47 versus 1.99). These differences were not found in the blind population. Mean age of blind individuals was 72.3 in the community and 71.4 in institutions. Mean number of handicaps of blind individuals was 1.58 in the community and 1.62 in institutions. This suggests that the process of institutionalization is different for blind and LV individuals. Blind people entered the institution when they were younger and they had on average a similar number of co-morbidities compared to the control group. However, the need for assistance (Katz index) is much higher for blind individuals than for controls. This suggests that blindness and its related dependency are associated with institutionalization. LV individuals entered the institution when they were older with a higher number of handicaps (after adjustment for age) than the population with no visual problems. This suggests that the association of handicaps, blindness being one of them, is linked with the institutionalization.
As expected, people living in institutions needed more assistance than those living in the community. People with no visual problems declared assistance needs that were very similar to the LV individuals. Blind individuals needed assistance more often than individuals with no visual problems (OR between 2.65 and 11.35).
Concerning institution adaptation (building and furniture changes) for the handicapped, the level of unmet needs was rather low, although one out of every 20 blind individuals had some demands that were unsatisfied. The demands of LV individuals were very similar to those of people with no visual problems. In contrast, blind people had some specific needs requiring particular institution adaptations with related costs.
57.9% of blind people living in institutions were registered for a social allowance and 35.4% of LV individuals. Concerning the monthly average allowances, the difference between LV and blind individuals was smaller for people living in institutions (EURO 86) than that calculated in the community (EURO 277). This might be explained by the fact that the financing of institutionalization includes some direct transfer from social allowances to the institution.
This study shares the same limitations as the one conducted in the community:
(1) The cross-sectional design of the survey did not permit an analysis of possible causal relationships between blindness, handicap, dependency and incapacity.
(2) The visual acuity (VA) of responders was neither measured nor controlled by an ophthalmologist and we were unable to apply the WHO classification for blindness (VA<3/60; ICD 10) [17]. The major limitation was that the primary outcome, visual impairment and blindness, was based upon interviews. There are several major problems with this. First, the validity of defining visual impairment and blindness by interview was uncertain. It was difficult to be sure that the prevalence estimates of 'visual impairment' and 'blindness' were accurate. Were they over-estimated or under-estimated? Second, there was no post-hoc simple medical algorithm indicating how low visual acuity and blindness were discriminated by the interview questions. Third, because both the outcome of interest (visual impairment and blindness) and risk factors (non-vision-related handicaps, need for assistance etc.) were determined by similar interview questions, the association between visual impairment and incapacity, assistance and economic consequences could be the result of reporting bias. Moreover, some patients in institutions could be severely cognitively impaired and experience significant trouble answering the questions of this survey. These problems could have generated measurement errors in both the visual impairment classification and the evaluation of disability.
The self-reported visual impairment prevalence rates we observed were very close [10-12] to those reported in other western developed countries for blind and LV individuals on subjects living in institution. At least at a national level, this gives empirical external validity to our classification of visual impairment. It is also worth commenting that we observed large differences between LV and blindness, in terms of incapacity, dependency and financial consequences. These differences persisted after adjustment on age, number of other handicaps and number of people in the household. Hence, the post hoc discriminant validity of blind and LV individuals seems acceptable. Therefore, because there were large differences between LV and blind individuals in terms of prevalence rates, incapacity, dependency and financial consequences, we decided to go ahead with distinguishing between the two groups. Lastly, misclassification might have occurred more frequently between LV and controls than between blind and LV individuals. The existence of a control group is important since the comparison between control and LV is a way to estimate the incapacity attributable to visual impairment, other things being equal (age, handicaps and size of household). Under the reasonable hypothesis that there is a continuum between visual impairment and severity of incapacity (as shown by this survey) misclassification should not have biased the overall burden borne by society. To summarize, the experimental design of this survey made it possible to estimate the self-declared visual impairment prevalence rates, for blind and LV persons, and their burden on society expressed as incapacity, dependency and economics."
Individuals classified as blind in this survey were those who self-declared that they did not perceive shapes. We believe that the interviewer-administered questionnaires were of sufficient quality to correctly differentiate between LV and blindness with a high probability of accuracy. Indeed in this elderly population, chronic diseases such as ARMD and glaucoma are known to account for the vast majority of ocular disorders. Lastly, access to ophthalmologists is free in France and the role of optometrists is negligible; theoretically, therefore, the conditions leading to visual impairment should have been identified medically in almost all French citizens.
Conclusion
The quantification of disabilities associated with blindness is very important for public health decision makers and persons responsible for institutions. The later need to know the number of patients that could potentially join institutions. They also have to face tariff issues. How much social allowances do blind / LV individuals have? Are the needs (assistance, medical devices, etc ...) of blind people different from the others? Would any differences in the needs justify a dedicated tariff? From a public health point of view, aging is associated with handicaps that will increase the probability of institutionalization [3]. Most of the countries belonging to the OECD have decided to control and limit the number of people that could be institutionalized, mainly to try to control social expenses. To do so, the knowledge of risk factors (possibly including blindness and LV) of institutionalization is critical. This study provides nationwide estimates (rates and odds-ratios) of the age-adjusted disabilities associated with self-reported blindness (since blind people are older) of citizens living in institutions. Some future research should be conducted to confirm these findings, more specifically dedicated to visual impairment, including medically-validated assessment of LV and blindness.
Authors' contributions
All authors have made substantial contributions to conception and design, acquisition of data, analysis and interpretation of data, have been involved in drafting the article or revising it and have given final approval of the version to be published.
Acknowledgements
This work was supported by an unrestricted grant from Alcon France SA, Rueil-Malmaison, France. It was conducted in collaboration with Cemka SA, Bourg-la-Reine, France and according to local laws. We are grateful to Irène Fournier and Benoit Riandey for the database. Alcon France SA employed Dr Gilles Berdeaux, but the authors had no financial or other conflicts of interest relating to Alcon Laboratories or directly relevant to the contents of this study. Dr Antoine Brézin has participated to clinical trials, unrelated to this study, financed by Alcon laboratories.
==== Refs
OECD Aging in OECD countries. A critical Policy Challenge Social Policy study N° 20, Paris 1997
OECD Maintaining propriety in an Aging Society, Paris 1998
Jacobzone S Cambois E Robine JM Is the health of older persons in OECD countries improving fast enough to compensate for population ageing ? OECD economic studies N° 30, OECD, Paris 2000 149 190
Jacobzone S Ageing and care for frail elderly persons: an overview of international perspectives. Labour market and social policy Occasional papers N° 38 OECD, Paris 1999
Mathers C Lopez A Stein C Ma Fat D Rao C Inoue M Shibuya K Tomijima N Bernard C Xu H Deaths and disease burden by cause: global burden of disease estimates for 2001 by world bank country groups
Robinson R Deutsch J Jones HS Youngson-Reilly S Hamlin DM Dhurjon L Fielder AR Unrecognised and unregistered visual impairment Br J Ophthalmol 1994 78 736 740 7803347
Bruce IW McKennell AC Walker EC Blind and partially sighted adults in Britain: the RNIB survey London: HMSO 1991 1
Walker EC Tobin MJ McKennell AC Blind and partially sighted children in Britain: the RNIB survey London HMSO 1992 2
Wormald R Evans J Registration of blind and partially sighted people Br J Ophthalmol 1994 78 736 740 7803347
VanNewkirk MR Weih L McCarty CA Stanislavsky Yuri Keefe J Taylor H Visual impairment and eye diseases in elderly institutionalized Australians Ophthalmology 2000 107 2203 2208 11097596 10.1016/S0161-6420(00)00459-0
Mitchell P Hayes P Wang JJ Visual impairment in nursing home residents: the Blue Mountains Eye Study Med J Aust 1997 166 73 76 9033561
Tielsch JM Javitt JC Coleman A Katz J Sommer A The prevalence of blindness and visual impairment among nursing home residents in Baltimore N Engl J Med 1995 332 1205 1209 7700315 10.1056/NEJM199505043321806
Brézin A Lafuma A Fagnani F Mesbah M Berdeaux G Blindness, low vision and other handicaps as risk factors attached to institutional residence Br J Ophthalmol 2004 88 1330 1337 15377561 10.1136/bjo.2003.039180
Colvez A Gardent H Les indicateurs d'incapacité fonctionnelle en gérontologie. Information, Validation, Utilisation Ed CTNERHI PUF, Paris 1990
Katz S Assessing self-maintenance Activity of daily living. Mobility and instrumental activities of daily living J Am Geriatr Soc 1983 31 721 726 6418786
Brézin A Lafuma A Fagnani F Mesbah M Berdeaux G Prevalence and burden of self-reported blindness, low vision and visual impairment in the community: a nation-wide survey Arch Ophthalmol
International statistical classification of diseases and related health problems. Tenth revision World Health Organisation Ed World Health Organisation Geneva, Switzerland 1993
| 15847703 | PMC1087865 | CC BY | 2021-01-04 16:38:14 | no | Health Qual Life Outcomes. 2005 Apr 25; 3:27 | utf-8 | Health Qual Life Outcomes | 2,005 | 10.1186/1477-7525-3-27 | oa_comm |
==== Front
Hum Resour HealthHuman Resources for Health1478-4491BioMed Central London 1478-4491-3-31583310510.1186/1478-4491-3-3ResearchEducational and labor wastage of doctors in Mexico: towards the construction of a common methodology Nigenda Gustavo [email protected] José Arturo [email protected] Rosa [email protected] Centre for Social and Economic Analysis in Health, Mexican Health Foundation, Mexico, Mexico2005 15 4 2005 3 3 3 28 6 2004 15 4 2005 Copyright © 2005 Nigenda et al; licensee BioMed Central Ltd.2005Nigenda 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
This paper addresses the problem of wastage of the qualified labor force, which takes place both during the education process and when trained personnel try to find jobs in the local market.
Methods
Secondary sources were used, mainly the Statistical yearbooks of the National Association of Universities and Higher Education Institutions (ANUIES in Spanish). Also, the 2000 Population Census was used to estimate the different sources of labor market wastage. The formulas were modified to estimate educational and labor wastage rates.
Results
Out of every 1000 students who started a medical training in 1996, over 20% were not able to finish the training by 2000. Furthermore, out of every 1000 graduates, 31% were not able to find a remunerated position in the labor market that would enable them to put into practice the abilities and capacities obtained at school. Important differences can be observed between generalists and specialists, as well as between men and women. In the case of specialists and men, lower wastage rates can be observed as compared to the wastage rates of generalists and women. A large percentage of women dedicate themselves exclusively to household duties, which in labor terms represents a wastage of their capacity to participate in the production of formal health services.
Conclusion
Women are becoming a majority in most medical schools, yet their participation in the labor market does not reflect the same trend. Among men, policies should be formulated to incorporate doctors in the specific health field for which they were trained. Regarding women, specific policies should target those who are dedicated full-time to household activities in order to create the possibility of having them occupy a remunerated job if they are willing to do so. Reducing wastage at both the educational and labor levels should improve the capacity of social investment, thereby increasing the capacity of the health system as a whole to provide services, particularly to those populations who are most in need.
==== Body
Introduction
The medical workforce has been studied from diverse points of view by means of a wide variety of methodologies and at varying levels of depth. Some studies approach the subject from the formal education standpoint, examining the problems encountered in the updating of knowledge, the number of human resources available and the distribution and proportion of resources according to the population. Other studies look into working conditions, prevailing ethical codes and productivity, among other topics. In a first review of the available specialized literature, very few studies were found to deal systematically with the difficulty for trained individuals in the health field to put into practice their acquired knowledge to deliver services [1].
In many countries the supply of doctors is being generated without planning and, very frequently, with no regulation at all. Doctors are being educated in schools that provide them with a professional degree; in spite of this, they will eventually face difficulties in finding a job. This problem is related to the qualified labor force wastage, as well as to the specific market configuration. On the other hand, there is another source of wastage that takes place during the education process and is reflected in the demand for medical education, the number of dropouts, the final efficiency rates at a national level and the organization and location of medical schools [2].
Besides these problems, the wastage of scholars has various facets that are not easy to identify, study or solve. One of these, perhaps the most important, is the enormous economic burden for a society represented by an inconclusive education process in which many students drop out before finishing their studies, making it impossible for them to put into practice the knowledge they acquire. This represents a fruitless investment and a wastage of economic resources that no nation can afford, particularly in the developing world.
Social investment in medical training should produce a benefit for societies, their institutions and citizens. The role of medical practice is highly valued in most societies, but it is up to governments to establish regulations that enable education to produce well-trained doctors practising under high ethical and professional standards. Government intervention is crucial to guaranteeing that this social investment provides the highest possible outcome. Health care reforms are taking place in many countries and are creating new conditions for medical practice, but not many of them regard HRH as a strategic planning issue [3].
It is also necessary to consider that wastage and geographical distribution are very much related. Indeed, doctors in Mexico have been concentrated for many years in urban areas, yet these areas also show high underemployment and unemployment rates. On the other hand, the lack of interest of doctors to practice in remote regions paradoxically results in a very reduced wastage in suburban and rural areas [4].
Therefore, the dynamics between human resources supply and the actual requirements of institutions lead to breakdowns that are expressed as labor wastage – that is to say, time and capabilities that graduate doctors do not put into practice either directly or indirectly for the production of health services.
The paper is not limited to the estimation of wastage rates but, based on the information obtained, widens its scope to discuss the implications for HRH – among them, the need to solve wastage in urban areas while encouraging the flow of doctors to remote areas. It also suggests the need to protect social investment in medical education. A further issue emphasized is the participation of women, as it has been made evident that their enrolment to medical schools has grown significantly in the last 15 years, yet their engagement in the labor market still occurs under different conditions as compared to those of men [5].
Objectives
The paper pursues the following objectives:
• to characterize the problem of wastage of doctors during their education process and in the labor market;
• to begin to construct a methodology that allows for the study of the problem in Mexico and, at the same time, could be replicated in other countries;
• to contribute to the discussion of the wastage of human resources in the health sector at the national and international levels;
• following the initial diagnosis about the wastage found in medical personnel, to use the sex variable to explore a relevant dimension of the problem.
Conceptual framework
The purpose of this section is to establish definitions that will set out the foundations for the presentation of results based on the information gathered. Due to the particularities of each field and for the sake of the exposition, the educational and labor market subjects will be addressed separately.
Wastage during education
Many studies in Mexico have dealt with the problems of school dropouts, final efficiency, repetition, exclusion and effectiveness and efficiency of the training institutions, among others. However, both at national and international levels little has been done to study the problem of specific professions (such as medicine) comprehensively so as to fully understand the causes and effects of the wastage of human and economic resources during the educational process [6].
The data presented in the following sections were obtained mainly from the National Association of Universities and Higher Education Institutions (ANUIES in Spanish), a nongovernmental agency that for more than 35 years has been responsible for compiling and systematizing information provided by 138 public and private institutions [7]. The stability and duration of this process of data collection renders the information obtained from this source highly reliable.
The following concepts have been taken from the existing studies:
Global attrition
This is a condition experienced by someone who does not comply with the timelines and does not complete the corresponding stages of the study plan of an institution in a specific year [8]. The student voluntarily or involuntarily interrupts his/her studies without having completed the studies required for his or her career. Such an interruption is not a spontaneous act, as there are family, social and institutional factors behind it [9]. Among the main causes of an interruption of studies are family attitude, economic conditions, study habits, inadequate selection of career, motivation, age, civil status and employment [10].
Graduates
Students who have completed the total number of credits required and/or have fully complied with the established procedures (such as writing a thesis) included in the plan of studies, receive a degree from a university or other institution of higher education.
Wastage in the labor market
There are factors affecting the labor market that are difficult to identify and much more difficult to quantify. The issue of unemployment and its varied manifestations in shape and time make it a somewhat polemical topic when dealing with secular variations. Although parameters to measure unemployment have been established at the international level, this has not allowed to count on timely information that might be comparable among countries and even among regions and states within a country [11]. In the case of professional groups, quantifying unemployment is not enough to understand labor market unbalances.
In the case of Mexico, the National Institute of Statistics, Geography and Informatics – INEGI – carries out two important data collections reporting results on various aspects related to employment at the national level: the General Census on Population and Housing is carried out every ten years, and a National Survey of Urban Employment, every three months.
Periodically INEGI reports a series of statistics about the employment situation across the country; one of these is the rate of open unemployment at the national level. However, this rate alone is not the most adequate indicator to establish the dimension of labor wastage, among other reasons because it is a macro indicator that shadows other levels of participation in the labor market that could be considered inadequate for an individual with professional training.
As part of the conceptual and methodological definitions of the present work we propose to estimate the rate of wastage among doctors, which would incorporate all those conditions in which a graduate from a medical school does not put into practice the knowledge gained from the school for the production of health services. Some situations, such as unemployment, are easy to identify as part of this wastage, but others (e.g., that of individuals who work less than 20 hours a week) are more difficult. We quantify wastage through the number of individuals who fall into categories in which their training does not match their labor activity. (Table 1)
Table 1 Sources of training and labor wastage. Categories used to estimate wastage in training are: attrition and non-graduated. Categories used to estimate labor wastage are: unemployment, household activities, and other jobs (working in activities not related to the field of training).
Training wastage*
Attrition Non-graduate
Labor wastage
Unemployment Household Other jobs (working in activities foreign to the field of training)
* Those who temporarily withdraw from studying and those who return to study are not included.
For the purpose of this study, the following concepts would need to be taken into account [12].
Employment
This is the situation in which graduates work as general practitioners or, as students for a medical specialization degree, in a full-time clinical practice at a hospital. It also includes specialist doctors with a labor position in health institutions according to the degree obtained. The category also comprises those doctors who are dedicated to research and/or teaching activities and those in managerial positions in health institutions.
Unemployment
This refers to individuals without employment, including those awaiting a reply about a job application (and who are not looking for any other job), those who are too discouraged to continue looking for a job and those who are actively seeking one.
Underemployment
This refers to individuals who have finished their studies and carry out activities different from their training; such activities may take place outside the health sector and/or in areas not directly related to the delivery of health services. To estimate underemployment in the general population, labor specialists normally consider working time and income criteria. In this case, since we are dealing with a professional group, a dimension that has proved to be appropriate for this measurement is the match between training and labor activities. This type of underemployment is known as qualitative underemployment and it is useful to understand the participation of highly trained groups of the population.
Household activities
These refer to individuals who do not have remunerated work because they are dedicated to household activities on a full-time basis.
Inactive, not available
This refers to groups of individuals who are retired, pensioned or suffer a permanent disability.
Labor wastage
This refers to qualified human resources who do not practice activities related to their formal education because they are not employed (including those dedicated to household activities) or because they carry out activities that do not correspond to their training.
Methods
With respect to wastage that takes place during the years of study, the Anuario Estadístico (Annual Statistical Book) published by ANUIES between 1976 and 2001 was used as the main source of information. It was necessary to carry out our own calculations to estimate enrolment, incoming students, graduates and abandonment per group pertaining to a common period of study, with a cohort of five years each.
Since there is no information on the incoming students disaggregated by sex, drop-outs and graduates for the years before 1996, it was possible to calculate rates of abandonment and final efficiency for only two graduating classes.
To calculate the wastage in the education of the medical students, the following formulas were established:
As to the wastage in the labor market, the database of the XII Censo General de Población y Vivienda, 2000 [13] (XII General Census on Population and Housing, 2000) was examined for information about the following variables: sex, age, education and individuals who had studied for the career of medicine. The information about specialists available in the Census database was very limited. Thus, we decided to focus our exercise on generalists (in Mexico, generalists are those graduated from medical schools who have not obtained a specialist degree).
As to the latter, census codes were compared to find out their activity status (their own or activities not related to their education), occupation (making it explicit if they were dedicated to household activities), and whether they were unemployed, retired, pensioned or permanently disabled.
Codes used to classify the activity area were taken from the North American Industry Classification System (NAICS) used by the 2000 National Census of Population and Housing. Category 6 corresponds to health services. Codes used to classify the educational level correspond to a classification developed by the National Institute of Statistics, Geography and Informatics.
Once this information was processed, the following formulas were built. To calculate the rate of employment among individuals who studied medicine:
For the rate of unemployment, the formula used was:
And for the rate of wastage it was established that:
Results and discussion
Based on the definitions set forth in the framework provided for this work and the formulas set out in the methodology section, the following results for the total medical personnel were obtained.
Wastage during education
In general, enrolment in the career of medicine in Mexico has shown non-linear behavior during the last 25 years. At the beginning of the 1990s it showed a tendency to drop, mainly because of the official policy implemented in the mid-1980s to try to halt the high demand then evident. Despite this, on average enrolment grew by 22.8% during the period 1990–2001 (Table 2).
Table 2 Total enrolment, medical schools in Mexico, 1990–2001. In general, enrolment in the career of medicine in Mexico has shown non-linear behavior for the past 25 years. It was at the beginning of the 1990s when it showed a downward trend, mainly because of the official policy implemented in the mid-1980s to halt the high demand to enrol that was evident at that time. Yet on average, enrolment increased by 22.8% during the 1990–2001 period.
Year 1977 1980 1983 1986 1990 1993 1996 1999 2001
Total 85 822 77 474 76 424 64 853 57 667 55 591 59 645 64 594 70 830
Source: ANUIES, Anuarios estadísticos, 1990–2001
Throughout this period, the proportional participation of women maintained constant growth. According to the annual statistical book from ANUIES, the percentage of women enrolled in medicine jumped from 43.9% in 1990 to 50.4% in the year 2001. It was in the year 1999 when the women enrolled outnumbered men for the first time, by 1038 students (Figure 1).
Figure 1 Total enrolment in the career of medicine by sex, 1990–2001. Throughout the period, the proportional participation of women maintained a constant growth. According to the annual statistical book from ANUIES, the percentage of women enrolled in medicine jumped from 43.9% in 1990 to 50.4% in 2001. It 1999 the number of women enrolled outnumbered men for the first time, by 1038.
An indicator that clearly illustrates wastage during education is the rate of final efficiency in the career of medicine at a national level. In the same manner, the series of graduating classes with the same cohort was constructed, enabling us to observe that the highest final efficiency was achieved in the graduating class of 1985 and that of 1995, with a rate of 834.9 and 804.3, respectively. After ten years the second-highest rate was reached, during the period 1995–1999 (Figures 2 and 3).
Figure 2 Global rate of attrition (GRA) in medicine by group pertaining to the same period of study, 1977–2001. To calculate the attrition in the medical profession, a series was constructed for each graduating group, from the first admission in 1977 up to the 1997–2001 graduating class. Once the series was completed, it could be established that the lowest level of drop-outs took place during the period 1985–1989, with a rate of 165.0 per thousand students, while the highest level was registered in the 1990–1994 class, with a rate of 493.5.
Figure 3 Incoming students and attrition by cohort, 1977–2001. An indicator that allows us to determine the wastage is the rate of final efficiency in the career of medicine at the national level. To this purpose, a series of graduating classes with the same cohort was constructed; this led to the identification of the highest final efficiency, which was achieved in the 1985 and the 1995 graduating classes, with a rate of 834.9 and 804.3, respectively. The second-highest rate was reached during the period 1995–1999.
Although it is true that the final efficiency varies from one medical school to another (which also would be important to further investigate), there is no doubt that at the national level the registered rates are worrisome, given the number of medical students who do not complete their studies.
On the other hand, the proportion of incoming students in relation to the total enrolment for medical training in Mexico during the period 1977–2001 was calculated. The result for the first year (1977) was 21.9, and for the second (2001), 21.8. The similarity between these two years is peculiar, since nothing similar is observed for the remaining years. In fact, this proportion decreases at times (13.9 and 14.0 in 1986 and 1982), while at some others it increases (as in the years of 1997 and 1998, with 23.6 and 23.3, respectively).
The notoriously high rate of attrition in 1990–1994 may well be related to the economic crisis Mexico was facing by the end of the period, which made it very difficult for students and their families to afford medical education. The rate of efficiency shown in Figure 2 mirrors the capacity of schools, students and families to reduce the volume of drop-outs [14].
Figure 3 presents another way to express wastage in medical education. The proportion of drop-outs in the period 1990–1994 is the highest of all periods (50%). The volume of drop-outs in that period was similar to those of 1977–1981 and 1978–1982, but the volume of new enrolments in the latter two periods was 40% higher.
As can been seen in Table 3, the information about drop-outs, graduates and final efficiency does not show any significant difference whenever the sex variable is included. However, given that it was possible to obtain information by sex for only two classes pertaining to the same period of study, it would be difficult to reach any solid conclusion in this respect.
Table 3 Global rates of attrition and final efficiency of the medical graduates per thousand students by sex and groups pertaining to the 1996–2000 and 1997–2001 periods. The information about drop-outs, graduates and final efficiency does not show any significant difference when the sex variable is included. However, given that it was possible to obtain information by sex for only two classes for the same period of study, it would be difficult to draw any final conclusions regarding this issue.
Cohort Incoming students Attrition Graduate students Global rate of attrition × thousand students Rate of final efficiency × thousand students
M W M W M W M W M W
1996–2000 6 200 6 054 1 390 1 100 4 810 4 954 224.2 181.7 775.8 818.3
1997–2001 6 819 6 820 2 215 2 343 4 604 4 477 324.8 343.5 675.2 656.5
Source: ANUIES, Anuario estadístico, 1996–2001
* No data are available for 2002.
Upon constructing series of data by year, it was found that women have moved from representing 19% of the graduates in 1970 to almost half the graduates in the year 2001 (49.3%). Since 1996, the number of female medical graduates has been very similar to that of men. Something similar occurs when comparing the information about incoming students and dropouts (Table 4). The number of graduates who received their degree has not shown significant changes in the recent years: in 1996, 45.9% were women, a figure that went up to 49.3% in the year 2001.
Table 4 Incoming students, drop-outs and graduate students in medicine by year and sex, 1996–2002. Since 1996, the number of women medical graduates has been very similar to that of men. Something similar occurs when comparing the information about incoming students and drop-outs. The number of graduate students receiving their degree has not shown significant changes in recent years: in 1996, 45.9% were women, while in 2001 this proportion was 49.3%.
Year Incoming students Percentage of total attrition Percentage of total graduate students
Men Women Men Women Men Women
1996 6 200 6 054 47 53 51 49
1997 6 819 6 820 49 51 51 49
1998 7 456 7 064 53 47 50 50
1999 7 331 7248 51 49 50 50
2000 7 655 7 858 51 49 49 51
2001 7 501 7 962 45 55 51 49
2002 7 746 8 631 n/a n/a n/a n/a
Source: ANUIES, Anuario estadístico, 1996–2002
* No data are available for 2002.
Wastage in the labor market
Out of the total number of general physicians and specialists in 2000, 70% were working in medical care and 5% were studying. From a sex perspective, 13% of the men were not working and 2% of the women were inactive, while the percentage of men and women working in medicine was 75% and 59%, respectively; that is to say, there was a difference of 16 points favoring the number of men employed.
Out of the population with education in general medicine that was dedicated to household activities, only 0.5% were men while 11% were women. Another item of information that stands out regards inactive and unavailable personnel: in fact, there are more men than women as such, which may indicate that there are fewer women retired and pensioned [15] (Table 5).
Table 5 Occupational status of physicians by sex, 2000. Out of the population with education in general medicine who were dedicated to household activities, only 1% were men, while 12% were women. It should also be noted that more men than women are inactive and not available; this may indicate that there are fewer women retired and pensioned.
Total % Men % Women %
National total 204 778 100 133 673 65 71 105 35
Employed 142 923 70 100 818 75 42 105 59
Studying 10 122 5 4 596 3 5 526 8
Unemployed 10 892 5 5 385 5.5 5 507 8
Dedicated to household activities 7 895 4 8 0.5 7 887 11
Working in activities not related to the field of training 26 733 13 18 289 13 8 444 12
Inactive, unavailable 6 213 3 4 577 3 1 636 2
Source: Data generated by FUNSALUD from the XII General Census on Population and Housing, 2000.
Results after repeating the same operations, but taking into account the variable sex, are presented in Tables 6, 7 and 8.
Table 6 Rate of employment, unemployment and wastage, 2000. To estimate the rate of employment, the following categories were used: total of employed divided by the total number of doctors minus students and inactive. In the case of unemployment, the following formula was used: unemployment divided by employed plus other jobs plus underemployed. In the case of wastage, the formula used was: unemployed plus household plus other jobs divided by the total number of doctors minus students and inactive.
Rate of employment = A/ (T - B - F)
Source: Data generated by FUNSALUD from the XII General Census on Population and Housing, 2000.
Where: A = Employment, B = Studying, C = Unemployed, D = Household, E = Others jobs, F = Inactive, T = Total
Table 7 Rates of employment in the health sector, employed in non-health activities and studying, 2000. Data used to estimate the rate of employment within the health sector were the following: the total number of employed divided by the total number of doctors minus the inactive. The formula used to estimate the students was: total number of students divided by the total number of doctors minus the inactive. To estimate the number of doctors working outside the health sector the following formula was used: other jobs outside the health sector divided by the total number of doctors minus the inactive.
Rate of employment in the health sector = A/T-F
Source: Data generated by FUNSALUD from the XII I General Census on Population and Housing, 2000.
Where: A = Employment, B = Studying, C = Unemployed, D = Household, E = Others jobs, F = Inactive, T = Total
Table 8 Rates of labor participation (× thousand doctors) by sex, 2000. To estimate the rates of labor participation by sex, men and women constituted separate groups. Formulas were estimated in the same way as for the total population.
Rates Women Men
Rate of employment
Rate of unemployment
Rate of wastage
Source: Data generated by FUNSALUD from the XII General Census on Population and Housing, 2000.
Discussion
The methodology proposed in the present document can be reproduced in other countries if a population census database is available. WHO should encourage countries to estimate medical and other occupational groups' wastage rates in order to find out ways to reduce this phenomenon and to make the most of social and private investments [16,17].
In Mexico, an Inter-institutional Commission for the Training of Human Resources for Health, an entity created by high educational and health authorities to plan the supply, demand and distribution of human resources for health in the country, has been working for almost 30 years. It is suggested that the Mexican case be revised, since this might contribute to encouraging linkage and interaction between training and health services institutions with the aim of achieving a better planning process according to country-specific characteristics.
Figure 4 shows that doctors enrolled at medical schools can follow one of two patterns: the first refers to students obtaining their diploma, and the second, to students dropping out or not fulfilling the established requirements to graduate. Once in the labor market, graduates can be divided into two subgroups: those who are ready to be immediately employed and those who are not. In turn, those who are ready can be employed, unemployed, underemployed or fully dedicated to household activities. Those who are not ready to be immediately employed are divided into students that go on for a specialization degree and those who are inactive.
Figure 4 Possible outcomes for individuals, from medical schools to the labor market. The diagram shows that doctors enrolled at medical schools can follow one of two patterns: the first refers to students obtaining their diploma, and the second, to students dropping out or not fulfilling the established requirements to graduate. Once in the labor market, graduates can be divided into two subgroups: those who are ready to be immediately employed and those who are not. In turn, those who are ready can be employed, unemployed, underemployed or fully dedicated to household activities. Those who are not ready to be immediately employed are divided into students that go on for a specialization degree and those who are inactive. A proportion of those initially enrolled will later be represented by the rate of attrition.
The wastage of resources during the education of medical students is significant, as was shown in the results. This problem should be studied in further detail to enable us to arrive at an estimation of the economic cost that this represents at the individual, family and social levels. This economic cost cannot be estimated from the available information regarding wastage in medical training and in the labor market. Before moving ahead to the next step, it would be important to conceptualize the factors that intervene in the wastage process in order to attain an objective assessment of it. Questions such as how to determine and estimate indirect costs and how to assign a money value to the time lost when people abandon the market temporarily seem to be of utmost importance.
The inequalities between men and women in the labor market are reflected not only as higher wastage rates, as has been shown in this document. In this respect, there is an increasing trend toward a decline in working conditions imposed in many developing countries by employers, both private and public. This process of decline includes the stagnation of salaries, the lack of guarantee of labor rights and benefits, and the use of temporary contracts instead of permanent positions. In Latin America it is possible to identify countries where declining working conditions have spread, such as Argentina, Brazil, Ecuador, Mexico and Peru. Contracts reflecting these conditions are offered mostly to women [18].
In Mexico previous studies have shown a relation between geographical distribution and wastage. The highest levels of medical unemployment and labor wastage in Mexico are found in urban areas [19]. This fact can be appreciated when comparing states with large rural populations to states with more urbanized populations [20]. Within the first group the estimated availability of doctors per population tends to be lower than in the second group. For example, in the southern state of Chiapas, with 54% of its population being rural, the number of available doctors per 1000 inhabitants is 0.79, while in the Federal District (Mexico City), with 99.8% of its population being urban, the number of doctors per 1000 inhabitants is 3.03, 3.8 times higher than in Chiapas.
It would be a sound idea to sensitize medical students about the problem and the need to move on to unsaturated areas. On the other hand, health policies should enhance incentives for doctors to move to underserved areas, including higher salaries and the possibility of further training for those doctors who show willingness to take that option.
The top decision-making levels of the health system should be supporting the design and development of studies aiming at understanding in detail the issues concerning labor wastage; this would contribute to producing policy recommendations that stress the need for a comprehensive and coordinated institutional participation [21].
The methodology followed to calculate the wastage during the educational process as well as in the labor market was demonstrated to be adequate to support these kinds of studies. Hence, based on the information derived from a population census and on the management of similar variables, it would be possible to replicate this method in other countries. Additionally, this would make it possible to carry out comparative and complementary studies that would set forth the problem in more detail and would assist in the formulation of alternative policies within the health sector.
Such a methodology can be applied without further implications for the exploration of the development and labor conditions that prevail in other occupational categories, such as nursing and dentistry.
Conclusion
Although assessment of the wastage phenomenon could be more accurate if a wider set of data were available, it can in any case be said that in Mexico the wastage of human resources in the health sector is a major problem. In the year 2000, 310 of every 1000 enrolled students did not finish their training. This represents an important source of wastage of human resources.
There is no doubt that the wastage of a highly qualified workforce has a negative impact on the economy of any country. Governments and families invest huge amounts of material and financial resources to train professionals; ultimately these professionals may not find a position within the labor market in which the functions they perform accord with the long training they have received [22]. By the end of the 1990s, 720 of every 1000 enrolled medical students finished their training, and of 1000 doctors available in the labor market, only 58% found paid employment that enabled them to practise the skills obtained during their training.
As part of the problem, the experience individuals have accumulated to fully integrate into the labor market as well as the barriers and opportunities they face to get a job must be taken into account. Finally, health systems, including their educational components, must look for ways to reduce the wastage of human resources in order to increase the efficiency of the system as a whole; this should be considered a social imperative.
Unemployment and the rate of wastage among women are much higher than for men [23]. This reflects an inequitable labor situation that adds up to a series of disadvantages related to the male-centered social structure prevalent in Mexico. For example, men generally receive a higher income for carrying out the same tasks as women; they also are usually accorded administrative positions [24].
It is clear that the issue explored in the present article represents one of the faces of a long-standing paradox. The number of doctors in urban areas has surpassed the demand from the population to the point of producing unemployment, underemployment and, ultimately, wastage. In rural areas, in contrast, doctors are still absent in many communities, so people become ill and die of communicable and preventable diseases for want of competent medical care [25]. Understanding this phenomenon is the first step towards solving the problem. But to proceed with the second step, it will be necessary to develop new policies, based on sound information, aiming at attaining a more equitable and efficient distribution of resources, including health personnel.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
GN conceived the original idea of the paper; wrote some its sections, generated the conclusions and recommendations and corrected the whole of its contents. JAR wrote some sections of the paper and generated the conclusions and recommendations. RB wrote some sections of the paper, compiled the information and generated tables and graphs.
Acknowledgements
The contribution of Yetzi Rosales in finding and formatting data was important for the preparation of this article. Also, the participation of Oscar Méndez Carniado was important for the handling of the XII General Census of Population and Housing 2000 database, as well as for the design and production of tables. We also appreciate the financial support of the Rockefeller Foundation to carry out the study. The authors remain solely responsible for the contents of the article.
==== Refs
Gupta N Zurn P Diallo K Dal Poz MR Uses of population census data for monitoring geographical imbalance in the health workforce: snapshots from three developing countries International Journal for Equity in Health 2003 2 11 14697099 10.1186/1475-9276-2-11
Kolehmainen-Aitken R-L Decentralization's impact on the health workforce: Perspectives or managers, workers and national leaders Boston: Joint Learning Initiative Working Group on Demand 2003
Rigoli F Dussault G The interface between health sector reform and human resources in health Human Resources for Health 2003 1 9 14613523 10.1186/1478-4491-1-9
Wibulpolprasert S Inequitable distribution of doctors. Can it be solved? Human Resources for Health Development Journal 1999 1 2 39
Knaul F Frenk J Aguilar AM The gender composition of the medical profession in México: implications for employment patterns and physician labor supply Journal of American Medical Women's Association 2000 55 32 35
Arroyo Laguna J Cuevas Álvarez L, Brito P Situación y desafíos de los recursos humanos en salud en el área andina Presente y futuro en la formación, práctica y regulación profesional en ciencias de la salud 2002 México: Pan American Health Organization/World Health Organization 353 414
Asociación Nacional de Universidades e Instituciones de Educación Superior Anuario Estadístico México 1990–2000
Martínez Rizo F Estudio de la eficiencia en cohortes aparentes Deserción, rezago y eficiencia terminal en las instituciones de educación superior Propuesta metodológica para su estudio 2001 México: UNAM/Asociación Nacional de Universidades e Instituciones de Educación Superior
Durán J Díaz G Análisis de la deserción estudiantil en la UAM Mimeo 1990 México: Universidad Autónoma Metropolitana
Casares Ortiz R Exploración preliminar de la causalidad de la deserción de la Facultad de Medicina de la Universidad Autónoma de Yucatán Educación y Ciencia 1994 3 33 40 julio-diciembre
Dovlo D Assessing HRH wastage and improving staff retention: an African perspective Boston: Joint Learning Initiative Working Group onDemand 2003
Frenk J Alagon J Nigenda G Muñoz G Robledo C Vazquez LA Ramirez C Patterns of medical employment: A survey of imbalances in urban Mexico American Journal Public Health 1991 81 23 29
Instituto Nacional de Estadística, Geografía e Informática XII Censo General de Población y Vivienda 2000 México 2001
Ruiz JA Molina J Nigenda G La formación de médicos y mercado de trabajo en México Caleidoscopio de la Salud 2003 México: FUNSALUD
Michel JB Why do women physicians work fewer hours than men physicians? Inquiry 1984 21 361 368 6240467
De Oliveira O García B Trabajo femenino y vida familiar en México 1994 México: Centro de Estudios Sociológicos. El Colegio de México
Phillip RK Marder WD Silberger AB The growing proportion of female physicians implications for US physicians supply American Journal Public Health 1990 80 300 304
Langer A Nigenda G Sexual and reproductive health and health sector reform in Latin America and the Caribbean 2004 México: Population Council/Inter American Development Bank
Frenk J Knaul FM Vázquez-Segovia LA Nigenda G Trends in medical employment: persistent imbalances in urban Mexico American Journal Public Health 1999 89 1054 1058
Wibulpolprasert S Pengpaiboon P Integrated strategies to tackle the inequitable distribution of doctors in Thailand: four decades of experience Human Resources for Health 2003 1 2 12904253 10.1186/1478-4491-1-12
Harrison ME Female physicians in Mexico: migration and mobility in the life course Social Science and Medicine 1998 47 455 468 9680229 10.1016/S0277-9536(98)00112-9
Weisman C Teitelbaum M The work-family role system and physicians productivity Journal Health Society Behavior 1984 74 1348 1352
Uhlenberg P Cooney T Male and female physicians: family and career comparisons Social Science and Medicine 1990 30 373 378 2309131 10.1016/0277-9536(90)90192-U
Phillip RK Marder WD Silberger AB The growing proportion of female physicians implications for US physicians supply American Journal Public Health 1990 80 300 304
Nigenda G Los recursos humanos para la salud en busca del equilibrio 1994 México: FUNSALUD. Observatorio de la salud
| 15833105 | PMC1087866 | CC BY | 2021-01-04 16:37:43 | no | Hum Resour Health. 2005 Apr 15; 3:3 | utf-8 | Hum Resour Health | 2,005 | 10.1186/1478-4491-3-3 | oa_comm |
==== Front
Int J Behav Nutr Phys ActThe International Journal of Behavioral Nutrition and Physical Activity1479-5868BioMed Central London 1479-5868-2-21580789810.1186/1479-5868-2-2DebateTheory, evidence and Intervention Mapping to improve behavior nutrition and physical activity interventions Brug Johannes [email protected] Anke [email protected] Isabel [email protected] Department of Public Health, Erasmus University Medical Center, PO Box 1738, 3000 DR Rotterdam, The Netherlands2005 4 4 2005 2 2 2 18 11 2004 4 4 2005 Copyright © 2005 Brug et al; licensee BioMed Central Ltd.2005Brug 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 present paper intends to contribute to the debate on the usefulness and barriers in applying theories in diet and physical activity behavior-change interventions.
Discussion
Since behavior theory is a reflection of the compiled evidence of behavior research, theory is the only foothold we have for the development of behavioral nutrition and physical activity interventions. Application of theory should improve the effectiveness of interventions. However, some of the theories we use lack a strong empirical foundation, and the available theories are not always used in the most effective way. Furthermore, many of the commonly-used theories provide at best information on what needs to be changed to promote healthy behavior, but not on how changes can be induced. Finally, many theories explain behavioral intentions or motivation rather well, but are less well-suited to explaining or predicting actual behavior or behavior change.
For more effective interventions, behavior change theory needs to be further developed in stronger research designs and such change-theory should especially focus on how to promote action rather than mere motivation. Since voluntary behavior change requires motivation, ability as well as the opportunity to change, further development of behavior change theory should incorporate environmental change strategies.
Conclusion
Intervention Mapping may help to further improve the application of theories in nutrition and physical activity behavior change.
==== Body
Background
Is there "nothing more practical than a good theory" in improving behavioral nutrition and physical activity interventions? And which theories are indeed good enough to help us improve the practice of encouraging people to adopt healthier diets and physical activity patterns? The International Journal of Behavioral Nutrition and Physical Activity (IJBNPA) recognized the importance of this issue and encouraged a 'theory debate' [1]. Jeffery started the debate by sharing his experiences with and views on applying theories in weight management and weight loss interventions [2]. His conclusion is that we focussed too much on social cognition models and that these models proved to be not very practical. He was not able to find much evidence from his own studies that using such theories improved the effectiveness of interventions. Rothman contributed to the debate by positing that theory should evolve based on rigorous empirical evidence and that intervention research is one of the best ways to evaluate and refine behavior change theory [3]. Rothman further stated that already much attention has been given to explaining how theories should be applied, and that now greater emphasis should be given to further refining or rejecting theoretical principles. In the present contribution to the theory debate, we argue that it is still very necessary to further improve the process that guides which theories are applied in behavior change interventions, how these theories are applied, as well as to further improve and integrate existing theories.
What is theory and why do we need it?
Since the publication of Green and Kreuter's Precede and Precede-Proceed models [4], the health behavior promotion area has recognized the importance of careful theory-based intervention planning. According to these, and other similar planning models [5], the first step in health-promotion planning is the identification of health problems that are serious and/or prevalent enough to justify spending time, money and other resources. In the second step, the behavioral risk factors for the health problems need to be identified. Step 3 is to investigate the mediators or determinants of these risk behaviors after which these determinants should be translated into intervention goals, change strategies and methods, that need to be integrated in a comprehensive intervention package (step 4) that can be implemented and disseminated (step 5). Each step should preferably be evidence-based (see Figure 1).
Figure 1 A Model for Planned Health Education and Promotion
Behavioral theories are mostly used for step 3 of the planning process. Since free choice and autonomy are important values in many societies, and what we eat or how much we exercise are believed to be part of free choice, people choose to a large extend what they eat, and if they are physically active. With very few exceptions (small children, certain groups of institutionalized people), diet and physical activity (PA) behaviors can thus not be influenced directly; instead, we need to influence people's choices. What people choose to eat or do is influenced by a complex, interrelated set of so-called 'mediators' or 'determinants' of nutrition and PA behaviors, including different cognitions as well as environmental factors such as food availability and accessibility [6]. Successful behavior change interventions are dependent on the ability to influence these mediators. Rothschild posited that these mediators can be divided in three broad categories: motivation, abilities and opportunities [7]. Thus, complex combinations of motivations, abilities and opportunities determine diet and PA behaviors. Their relative importance, as well as the underlying beliefs of these determinants, are likely to differ across different populations, as well as between individuals within populations, depending on their personal, social, and environmental circumstances. Furthermore, since these circumstances are liable to change, the most relevant specific determinants may also change over time.
Since the second half of the last century, evidence-based medicine came into fashion. The urgency to base our efforts to improve health or to prolong life on scientific evidence was also transferred to public health and health promotion. In 1998, for example, the World Health Assembly stated that all member states should adopt an evidence-based approach to health promotion [8]. However, what counts as evidence may be debatable. In clinical practice, the randomized controlled trial (RCT) is the 'gold standard' to obtain evidence. An RCT ensures good internal validity, but may lack external validity especially in evaluation of complex behavior-change interventions [8]. The effects of a diet and PA change intervention, based on an inventory of the mediators of change in a specific population and tested among that population, may not be the same in another population with different motivations, abilities or opportunities. Does this mean that evidence on significant mediators of change and effective interventions are relevant only for that specific population under those specific circumstances and thus that it is in fact impossible to build a real evidence-base for behavior change interventions? We don't think so: we should use evidence obtained in specific populations, and under specific circumstances to build, refine and improve behavior change theory.
The Collins Cobuild English Language Dictionary provides two meanings for the word 'theory'. According to the first meaning, a theory is "an idea or set of ideas to explain something. It is based on evidence and careful reasoning, but it cannot be completely proved". The second meaning of a theory according to Collins is "an idea about something that is based on a lot of thinking but not on actual knowledge or evidence".
What we can derive from research, research that includes RCTs but also formative and process evaluation, as well as other forms of impact and effectiveness evaluations [8,9], are evidence-induced general ideas, about which categories of mediators, and which intervention strategies and methodologies, as well as intervention channels may work for influencing certain mediators of behavior change. In other words: we need to build behavior-change theory. These theories should thus be the weighted, systematic, summarized and carefully interpreted results of what has been found in empirical studies directly or indirectly related to nutrition and PA behavior and behavior change. Therefore, we argue that theory-based – in the first aforementioned meaning of the word theory – health behavior interventions is the equivalent of the evidence-based approach in clinical practice. Therefore, theory-based interventions are the only acceptable way to proceed in promotion of healthy diet and PA habits. This, however, implies that we are highly dependent on the quality of the theories and on how these theories are applied in intervention development and implementation. Behavioral nutrition and PA research is a relatively young scientific discipline. Diet and PA behaviors are complex behavioral categories and evidence-induced behavioral nutrition and PA theory is still in its developmental phase and currently too many nutrition and PA behavioral interventions are still based on theory in its second meaning.
In the remainder of this position paper we will, therefore, argue that in healthy nutrition and PA behavior promotion
• we may not always use the available theories in the right way,
• many theories still lack a strong empirical foundation,
• we tend to use theories that are too much focused on the individual and on motivational processes, and
• we may be too inclined to apply a single theory approach.
The Intervention Mapping protocol introduced by Bartholomew and colleagues [10] suggests specific steps that guide problem-driven development, application and integration of nutrition and PA behavior-change theories. IM proposes a systematic way to proceed from knowledge about behavioral determinants to specific change goals, and subsequently to intervention methods and strategies based on the production of intervention matrixes. Such matrices finally develop into an 'intervention map' that make the translation of objectives to change strategies to actual intervention activities explicit [10,16]. In the next discussion paragraph we will therefore refer to approaches suggested in IM that may help to improve the application of behavior theory in intervention development.
Discussion
Behavior and behavior change
Theories that are used to inform nutrition and PA behavior-change interventions, are often theories primarily meant to understand, i.e., to explain or predict, behaviors [11]. Applying such theories comes with two important problems. First, a determinant of behavior is not the same as a determinant of behavior change. In a study based on the Theory of Planned Behavior [12,13], we found that a majority of Dutch adults had positive attitudes, subjective norms and perceived control to eat a low fat diet, and that these presumed determinants were highly associated with the intention to eat a low fat diet [14]. More than 80% of the Dutch adult population was eating a diet higher in fat than the official Dutch recommendation at that time (fat intake below 35 percent of total energy intake fat). Thus, it appeared that the conditions for encouraging the Dutch population to eat less fat were very positive. However, further research showed that attitudes, norms, perceived control, as well as intentions to reduce fat intakes were much less positive [14]. Most people incorrectly thought their diet was already low in fat, and had positive attitudes, perceived norms, control beliefs and intentions to keep eating what they already did [15]. These findings illustrate the need to apply the available theories to try to explain and predict the desired behavior change instead of the status quo of healthy or unhealthy behavior, and thus to measure attitudes, subjective norms and perceived control toward behavior change. Intervention Mapping (IM) is a framework for effective theory- and evidence-based decision-making at each step of the process of intervention development, implementation and evaluation of health-promotion interventions [10]. IM argues that for health promotion intervention development we should first translate the health-related behavior (e.g. high fat intake) into a health-promoting behavior or behavior change (e.g. fat reduction) and then search for determinants of the required change, instead of predictors of present behavior.
Second, theories that help us to gain insight into possible determinants of nutrition and PA behaviors or behavior-change, do not directly tell us how to modify these behaviors. Such theories only help us to find out what needs to be modified to induce behavior-change [2,3]. We need to use and build better theories that guide us in how the determinants of behavior-change can be modified; how to translate behavior change determinants into behavior-change methods, strategies and actual intervention tools.
Association or determination?
Attempts have been made to develop behavior-change theories. One of the most popular theories used for the development of nutrition and PA behavior-change interventions is the Transtheoretical model and its popular stages of change concept. Stages of change has such a strong appeal because it is brief, has high face validity, and can be easily explained, also to the non-behavioral scientist. Thus, it is readily applicable in intervention development and has been used to develop a range of different nutrition and PA interventions (see [17,18]). Stages of Change is, however, also exemplary for the still rather weak evidence for the theories we use in behavior nutrition and PA research. The evidence for Stages of Change comes almost solely from cross-sectional studies [17,19]. For example, based on cross-sectional associations, pros, cons and self-efficacy are regarded as stage-transition determinants and are used to tailor interventions to each stage of change. However, such cross-sectional associations do not prove that these factors predict, let alone cause stage-transitions or behavior change [19]. Different authors have argued that longitudinal and experimental studies are needed to validate behavior and behavior-change theories better [17,19,20]. In a series of studies that were recently conducted in the Netherlands, the associations found between presumed stage transition determinants and stages of change in cross-sectional studies, could not be replicated in longitudinal analyses [21]. Furthermore, stages of change lacked stability over time, even within a short time interval of only three days [21,22]. As others have also argued, such study results make the validity of one of the most often used behavior change theories rather doubtful [19]. We thus strongly support Rothman's suggestion to use intervention research to test and further refine behavior change theory [3], and Jeffery's experiences show that such rigorous, true experimental tests of theory, can be disappointing [2].
From motivation to action
Nevertheless, there is evidence that interventions that have applied the stages of change concept are more effective than non-stage matched interventions, at least for short term effects [17]. Perhaps the most important contribution of TTM is the distinction between a motivational phase and a volitional phase in behavior change [23]. The distinction between motivation and action indeed appears to be very relevant. Most theories that are used to inform diet and PA change interventions explain quite well motivation or behavioral intentions, but the explained variance for behavior or behavior change is much lower [25,26]. Intention is an important predictor of behavior. In fact, lack of intention almost certainly results in lack of behavior change. However, a positive intention is no guarantee for behavior change [26], especially not for complex, habitual behaviors like nutrition and PA, that depend very much on personal abilities and environmental opportunities. We need theories to design interventions that help people bridge this intention-behavior gap, i.e. theories that improve people's abilities and opportunities to effectively act on their motivations. Such action-oriented self-regulation models focus specifically on the cognitive mechanisms involved in translating an intention to perform a particular behavior into action. The central tenet of self-regulation models is that through the formation of action goals, pursuing these goals and continuing to pursue these goals in the face of difficulties (i.e. coping with difficulties and frustration) successful transformation of motivation into action and maintenance can be accomplished. Self-regulation models provide various strategies for action initiation and goal pursuit, such as forming implementation intentions [27], goal setting and feedback [28], action planning and building on action self-efficacy and coping self-efficacy [29], self-monitoring and skills training (12). The body of evidence regarding the efficacy and applicability of using these strategies in modifying complex health related behavior is now growing [30,31].
Another possible approach to contribute to bridging the intention – behavior gap is to try to accomplish environmental changes; to increase the actual opportunities for healthy nutrition and PA behaviors and/or to reduce the opportunities for unhealthy behaviors. It has, for example, repeatedly been argued that an environment that offers and encourages plenty of opportunities to eat palatable energy-dense foods and to avoid physical activity may make it extremely difficult for people to act on their positive intentions to prevent weight gain [32-35]. These interventions have mostly tried to apply individual behavioral theories to increase people's awareness, motivation, abilities and confidence to face such an environment. More recently, however, so-called social-ecological theory that highlight the importance of environmental influences, has gained more attention [11]. Once again, this theory mainly identifies what needs to be changed in the environment, rather than how this change can be induced. We still lack systematic evidence and careful reasoning (i.e. theory) driven interventions that can change the environment to offer better opportunities for healthy eating and PA. IM might again offer some direction here [10]. In line with ecological models of health behavior, IM distinguishes between individual and environmental determinants of health behavior and argues that interventions may directly or indirectly address the at-risk individuals. In accomplishing environmental change, the indirect pathway should be used: IM has adopted the approach of Simons-Morton and colleagues [36] and suggests that environmental change is most often eventually the result of changes in behavior of ' decision makers' or 'role actors' at the different levels of the environment: interpersonal, organizational, community or societal. For example parents, school management, local, state and national policy makers all determine part of the nutrition and PA environment of school children. The choices and practices (i.e. behavior) of these decision makers shape the environment to a large extent; it is their health-promotion-facilitating behavior that should induce changes in the environment so that the health-related behavior of the people at risk will change. IM therefore argues that environmental change interventions can best be regarded as behavior change interventions aimed at these decision makers. Consequently, planned environmental change intervention development should first explore the important and changeable mediators of the required behavior change, formulate learning and performance objectives to be accomplished, and identify evidence or theory-based intervention strategies to induce the required behavior change among these decision makers. Environment change is thus translated into behavior change among the agents that have decisional power to modify the environment. This may only be a first step toward bringing about changes at different environmental levels, but it at least opens up a way to systematically think about the important issue of translating our revived attention for and growing insight in the importance of environment as a mediator of nutrition and PA behaviors, into environmental change interventions.
Problem-driven and theory-driven research
In the last paragraph we argued that the shift in attention to social-ecological models is a much-needed development in nutrition and physical activity behavior-change research. However, what we really need are not studies that highlight the importance of individual factors, social factors or physical environmental factors in shaping nutrition and PA behaviors. We need more studies that integrate potential determinants at the environmental and the individual levels [11], to study the relative importance of motivation, abilities and opportunities [7] as mediators of nutrition and PA behaviors [37-39]. Without such integrative research, it will remain unclear which causal pathways determine behavior-change and which category of determinants is the preferred point of departure for behavior change interventions. In line with what was suggested by Kok and colleagues for applied social psychology research [40], we can distinguish two general directions in nutrition and PA behavior research: theory-driven and problem- or action-driven research [41]. Theory-driven research is conducted to test or improve the validity or applicability of a specific theory of nutrition or PA behavior (see for example [42-45]. Problem-driven or action-driven research is conducted to tackle a specific problem, to explain this problem to the fullest extent and to give direction to possible solutions. Theories are of most importance in problem-driven research, but the main focus is not on testing a theory, but on using insights from different relevant theories in order to solve a problem. Therefore, in such problem-driven research often concepts derived from different theories are used instead of the single-theory perspective of most theory-driven studies. IM has adopted this integrative (i.e. multi-level) problem-driven approach to explore mediators of behavior change and to identify potential behavior change strategies [10,16].
Conclusion
Theory is the only foothold we have in development of behavior nutrition and physical activity interventions since theories are (or should be) a generalized and careful interpreted systematic summary of empirical evidence. Thus, application of theory should improve the likelihood of effectiveness of interventions. However, most of the theories that are applied in behavior nutrition and PA interventions provide information on what needs to be changed to promote healthy behavior but not on how change can be induced. Furthermore, some of these theories lack a strong empirical foundation and do better in explaining behavior intentions or motivation than actual behavior or behavior change.
For more effective interventions, behavior change theory needs to be further developed with stronger research designs and such change theory should especially focus on how to promote action rather than mere motivation. Since voluntary behavior change requires motivation, ability as well as the opportunity to change, further development of behavior change theory should incorporate environmental change strategies. Intervention Mapping may provide a number of tools to further improve the development and application of theories in interventions to promote nutrition and PA behavior change.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
JB initiated this paper and wrote the first draft.
AO and IF discussed the draft paper with JB and provided written comments. AO wrote the paragraph on action-oriented theory.
Acknowledgements
None
==== Refs
French SA Worsley T What is the role of theory in health behavior change interventions? Int J Behav Nutr Phys Act 2004 1 9 15272937
Jeffery RW How can Health Behavior Theory be made more useful for intervention research? Int J Behav Nutr Phys Act 2004 1 10 15272938
Rothman AJ "Is there nothing more practical than a good theory?": Why innovations and advances in health behavior change will arise if interventions are used to test and refine theory Int J Behav Nutr Phys Act 2004 1 11 15279674
Green LW Kreuter MW Health promotion planning: An educational and ecological approach. 1999 3rd Mountain View, CA, Mayfield
McKenzie JF Smeltzer JL Planning, implementing, and evaluating health promotion programs: A primer. 2001 Third edition Boston, Allyn and Bacon
Baranowski T Weber Cullen K Baranowski J Psychosocial correlates of dietary intake: Advancing dietary intervention. Annu Rev Nutr 1999 19 17 40 10448515
Rothschild ML Carrots, Sticks, and Promises: A Conceptual Framework for the Management of Public Health and the Social Issue Behaviors J Mark 1999 63 24 37
Rootman I Goodstadt M Hyndman B McQueen DV Potvin L Springett J Ziglio E Evaluation in health promotion: Principles and perspectives. 2001 Copenhagen, WHO regional Office for Europe
Campbell M Fitzpatrick R Haines A Kinmonth AL Sandercock P Spiegelhalter D Tyrer P Framework for design and evaluation of complex interventions to imporve health. BMJ 2000 321 694 696 10987780
Bartholomew KL Parcel GS Kok G Gottlieb NH Intervention mapping: designing theory- and evidence-based health promotion programs 2001 California, Mayfield publishing company 515
Kok G Schaalma H Ruiter R Empelen P van Brug J Intervention mapping: A protocol for applying Health Psychol theory to prevention programmes J Health Psychol 2004 9 85 89 14683571
Baranowski T Cullen KW Nicklas T Thompson D Baranowski J Are current health behavioral change models helpful in guiding prevention of weight gain efforts? Obes Res 2003 11 Suppl 23S 43S 14569036
Ajzen I The theory of planned behavior Organizational Behavior and Human Decision Processes 1991 50 179 211
Conner M Norman P Predicting health behaviour 1996 Philadelphia, Open University Press 230
Brug J Kok GJ Misconceptie van consumenten over eigen vetconsumptie [misconceptions of consumers about their own fatconsumption] Voeding 1995 56 11 14
Brug J van Assema P Lenderink T Glanz K Kok GJ Self-rated dietary fat intake: association with objective assessment of fat, psychosocial factors and intention to change J Nutr Educ 1994 26 218 223
Adams JWM Why don't stage-based activity promotion interventions work? Health Educ Res 2004 20 237 243 15253998
Van Sluijs EMF Van Poppel MNM Van Mechelen W Stage-based lifestyle interventions in primary care: are they effective? Am J Prev Med 2004 26 330 343 15110061
Sutton S Interpreting cross-sectional data on stages of change Psychol Health 2000 15 163 171
Baranowski T Cullen KW Baranowski J Psychosocial correlates of dietary intake: advancing dietary intervention Annu Rev Nutr 1999 19 17 40 10448515
De Vet E De Nooijer J De Vries N Brug J Stages of change in fruit intake: A longitudinal examination of stability, stage transitions and transition profiles. Psychol Health
De Nooijer J VAPDVEBJ Van Assema P De Vet E Brug J How stable are stages of change for nutrition behaviors? Health Promot Int 2005 20 27 32 15668216
Brug J Conner M Harre N Kremers S McKellar S Whitelaw S The Transtheoretical Model and styles of change: a critique: Observations by five Commentators on the paper by Adams J and White M (2004). Why don't stage-based activity promotion interventions work? Health Educ Res 2004 20 244 258 15618324
Godin G Kok G The theory of planned behavior: a review of its applications to health-related behaviors Am J Health Promot 1996 11 87 98 10163601
Sheeran P Hewstone WSM Intention-Behavior Relations: A Conceptual and Empirical Review European Review of Social Psychology 2002 12 , John Wiley & Son 1 36
Gollwitzer PM Implementation intentions: Strong effects of simple plans Am Psychol 1999 54 493 503
Latham GP Locke EA Self-regulation through goal setting Organizational Behavior and Human Decision Processes 1991 50 212 247
Schwarzer R Renner B Social-cognitive predictors of health behavior: Action self-efficacy and coping self-efficacy. Health Psychol 2000 19 487 495 11007157
Prestwich A Lawton R Connor M The use of implementation intentions and the decision balance sheet in promoting exercise behaviour Psychol Health 2003 18 707 721
Armitage CJ Evidence That Implementation Intentions Reduce Dietary Fat Intake: A Randomized Trial Health Psychol 2004 23 319 323 15099174
Wammes BM Kremers S Breedveld B Brug J Correlates of motivation to prevent weight gain: A cross-sectional study Int J Behav Nutr Phys Act 2005 16 1 15771774
Swinburn BA Caterson I Seidell JC James WP Diet, nutrition and the prevention of excess weight gain and obesity Public Health Nutr 2004 7 123 146 14972057
Mayor S Government task force needed to tackle obesity BMJ 2004 328 363 14962856
Schwartz MB Puhl R Childhood obesity: a societal problem to solve Obes Rev 2003 4 57 71 12608527
Simons-Morton DGSMBGPGSBJF Influencing personal and environmental conditions for community health: A multilevel intervention model. Family and Community Health 1988 11 25 35 10288523
Crawford D Ball K Behavioural determinants of the obesity epidemic Asia Pac J Clin Nutr 2002 11 S718 S721 12534697
Ball K Crawford D The obesity epidemic: contextual influences on physical activity and body weight J Sci Med Sport 2003 6 377 378 14723387
Flay BR Petraitis J The theory of triadic influence: a new theory of health behavior with implications for preventive interventions Advances in Medical Sociology 1994 4 4 19
Kok G Stroebe W and Hewstone M Social psychology and health education European Review of Social Psychology 1996
Brug J Van Assema P Frewer LJ, Risvik E and Schifferstein Beliefs about fat. Why do we hold beliefs about fat and how do we study these beliefs? Food, People and Society 2001 Berlin, Springer 39 54
Armitage CJ Conner M Distinguishing perceptions of control from self-efficacy: Predicting consumption of a low-fat diet using the theory of planned behavior. J Appl Soc Psychol 1999 29 72 90
Lechner L Brug J de Vries H van Assema P Mudde A Stages of change for fruit, vegetable and fat intake: consequences of misconception Health Educ Res 1998 13 1 11
Rodgers WM Courneya KS Bayduza AL Examination of the transtheoretical model and exercise in 3 populations Am J Health Behav 2001 25 33 41 11289727
Sarkin JA Johnson SS Prochaska JO Prochaska JM Applying the transtheoretical model to regular moderate exercise in an overweight population: validation of a stages of change measure. Prev Med 2001 33 462 469 11676588
| 15807898 | PMC1087867 | CC BY | 2021-01-04 16:38:04 | no | Int J Behav Nutr Phys Act. 2005 Apr 4; 2:2 | utf-8 | Int J Behav Nutr Phys Act | 2,005 | 10.1186/1479-5868-2-2 | oa_comm |
==== Front
Int J Behav Nutr Phys ActThe International Journal of Behavioral Nutrition and Physical Activity1479-5868BioMed Central London 1479-5868-2-31581119010.1186/1479-5868-2-3ResearchParental exercise is associated with Australian children's extracurricular sports participation and cardiorespiratory fitness: A cross-sectional study Cleland Verity [email protected] Alison [email protected] Jayne [email protected] Terence [email protected] Leigh [email protected] Menzies Research Institute, University of Tasmania, Hobart, Australia2005 6 4 2005 2 3 3 27 9 2004 6 4 2005 Copyright © 2005 Cleland et al; licensee BioMed Central Ltd.2005Cleland 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 relationship between parental physical activity and children's physical activity and cardiorespiratory fitness has not been well studied in the Australian context. Given the increasing focus on physical activity and childhood obesity, it is important to understand correlates of children's physical activity. This study aimed to investigate whether parental exercise was associated with children's extracurricular sports participation and cardiorespiratory fitness.
Methods
The data were drawn from a nationally representative sample (n = 8,484) of 7–15 year old Australian schoolchildren, surveyed as part of the Australian Schools Health and Fitness Survey in 1985. A subset of 5,929 children aged 9–15 years reported their participation in extracurricular sports and their parents' exercise. Cardiorespiratory fitness was measured using the 1.6 km (1-mile) run/walk and in addition for children aged 9, 12 or 15 years, using a physical work capacity test (PWC170).
Results
While the magnitude of the differences were small, parental exercise was positively associated with children's extracurricular sports participation (p < 0.001), 1.6 km run/walk time (p < 0.001) and, in girls only, PWC170 (p = 0.013). In most instances, when only one parent was active, the sex of that parent was not an independent predictor of the child's extracurricular sports participation and cardiorespiratory fitness.
Conclusion
Parental exercise may influence their children's participation in extracurricular sports and their cardiorespiratory fitness levels. Understanding the correlates of children's extracurricular sport participation is important for the targeting of health promotion and public health interventions, and may influence children's future health status.
==== Body
Background
An important component of children's total physical activity comes from extracurricular sports participation. Extracurricular sports are those organised and non-organised sports played outside of the school curriculum, and form an element of children's discretionary physical activity. Sports participation in childhood and adolescence is important because it is thought to be associated with lower levels of antisocial behaviour [1,2], higher levels of positive emotional wellbeing [3] and greater participation in sport in adulthood [4].
Extracurricular sports participation has substantially decreased in Australian children over the last 15 years. In 1985, 91% of boys and 90% of girls participated in at least one extracurricular sport [5], whereas in 2000 this had decreased to 71% of boys and 58% of girls participating in at least one extracurricular sport [5,6]. Over the same period, the prevalence of childhood overweight and obesity has increased, with Australian children in 1995 being nearly twice as likely to be overweight and more than three times as likely to be obese as children in 1985 [7].
With the increasing rates of childhood overweight and obesity [7], and suspected declines in levels of overall physical activity, it is important to explore correlates of children's physical activity. A recent review of these correlates recommended more research examining the inconsistent associations seen in previous studies between children's physical activity and parental physical activity [8]. There have also been inconsistent findings in studies that have assessed whether the sex of the physically active parent is associated with children's physical activity [9-12]. These inconsistencies are possibly due to methodological differences, which make it difficult to compare results directly. For instance, studies that have used accelerometers to objectively measure physical activity in children and their parents have found significant associations between children's and parental physical activity [9,13,14], as have studies that have used children's reports of their parent's physical activity [10,15,16]. Studies that have used self-reported parental physical activity have generally found no associations[11,17-20]. A number of these studies, however, had limited sample sizes [9,13,14,17], reducing the statistical power and limiting the ability to draw firm conclusions.
Few studies have explored the relationship between parental exercise and children's extracurricular sports participation and cardiorespiratory fitness levels. In the two studies that have explored this relationship [11,18], no association was found between parental physical activity and children's physical fitness, as measured by maximal oxygen uptake [11] and 1-mile run time [18]. However, it may be expected that parental exercise plays a role in children's physical fitness through a variety of mechanisms, including role modeling and genetic associations. No research to date has used a population-based sample, nor has this relationship been assessed in the Australian setting, which may have unique sociocultural features.
This study aimed to examine the relationship between children's extracurricular sports participation, cardiorespiratory fitness and parental exercise in a large, nationally representative sample of Australian children aged 9–15 years. We hypothesise that children who report having more active parents are involved in a greater number of extracurricular sports and have higher levels of cardiorespiratory fitness than children who report having less active parents. We also hypothesise that where only one parent is reported as active, that the sex of that parent influences children's extracurricular sport participation and cardiorespiratory fitness.
The Australian Council for Health and Physical Education Research's Australian Schools Health and Fitness Survey (ASHFS), conducted in 1985, remains the largest population-based study of children in Australia with extensive health and fitness measures and a high response rate. Due to this high response rate and population-based approach, results from this study are less likely to be subject to selection bias. Data from this study provide information on children and adolescents from a wide age range (9–15 years) and provide the opportunity to compare with current and future trends.
Methods
Participants
In 1985, a nationally representative sample of 8,484 children aged 7–15 years participated in the ASHFS, which gathered extensive measures of health and fitness through various field and technical tests, questionnaires and blood samples. The sampling design and sampling techniques are described in detail elsewhere [21]. Briefly, the first stage of sampling involved selecting schools with probability proportional to enrolment, and the second stage involved simple random sampling within each age/sex category to select children from within those schools. This sampling plan was designed to yield a self-weighted sample. Ninety per cent of schools approached agreed to participate (n = 109), and informed consent was obtained from the parents of 77.5% of the students selected in the initial sample. Participants aged 9–15 years (n = 6,659) were surveyed about their extracurricular sports participation and their parents' exercise. Children aged 7 and 8 years were not included in this aspect of the survey, because they were considered too young to reliably complete the questionnaire. Trained data collectors read questionnaire instructions to children in groups of four and supervised the completion of the questionnaire. Approval was granted to contact schools by the State Directors General of Education, and parental and child consent was required for inclusion in the study.
Predictor Variable
Children's report of parental exercise
In the student questionnaire, children were asked to report their parents' involvement in exercise. Specifically, children were asked, "Does your father exercise regularly (2 or more times a week)?" and "Does your mother exercise regularly (2 or more times a week)?". The examples of jogging, playing sport, doing exercises, going to a gym and doing aerobics were provided. Children's responses were classified as 'both parents active', 'mother active only', 'father active only' and 'both parents inactive'. Children who responded 'yes' for both their mother and their father were categorised as 'both parents active' (n = 1,183). If the child responded 'yes' for mother but 'no', 'don't know' or had missing data for their father, they were categorised as 'mother active only' (n = 1,055); 'father active only' was defined analogously (n = 1,192). If the child reported 'no' for both parents, or 'no' for one parent and 'don't know' or had missing data for the other parent, this resulted in the category 'both parents inactive' (n = 2,499). Children who responded 'don't know' or had missing data for both parents were categorised as 'incomplete' and excluded from the analyses (n = 485).
Outcome Variables
Children's extracurricular sports participation
The children were asked to list all the sports they had played regularly in the previous year, including sports played with an organised team, group, club or school, but not including sports played in physical education class at school. Children were provided with the opportunity to list up to six sports.
Cardiorespiratory fitness
Two measures of children's cardiorespiratory fitness were used. All children completed a timed 1.6 km (1 mile) run/walk. Children aged 9, 12 or 15 years also participated in a physical work capacity (PWC170) test on a Monark bicycle ergometer as a continuous test (n = 2,619). The workload corresponding to a heart rate of 170 beats per minute was predicted by linear regression from the heart rates recorded at the three submaximal loads [22]. This score was then divided by the child's weight in kilograms to take body size into consideration.
Potential Confounders
Area-Level Socioeconomic status
Numerous studies have found that indicators of socioeconomic status (SES), such as education, occupation and income, are associated with adult physical activity levels [23], although findings are not so clear in children [8,24]. Area-level socioeconomic status (SES) was therefore estimated in this study based on the Australian Bureau of Statistics (ABS) socioeconomic index for areas (SEIFA). The SEIFA is a summary of five indices designed to measure different aspects of SES by geographical area, based on questions asked in the Australian population census [25]. The ABS classified all Australian postcodes into one of four categories (low, medium-low, medium-high, high) based on one of these indices, the index of relative socioeconomic disadvantage (IRSD). The IRSD is constructed from items in the population census, including income, educational attainment, employment and occupation type. Using the 1981 census data, approximately 25% of Australia's population falls into each of the four IRSD categories. In this analysis, the IRSD was assigned to the child's residential postcode. In some cases, information on the child's residential postcode was unavailable and these children were therefore excluded from the analyses (n = 115).
School type
Children attended government (public or non-fee paying), independent (private or fee-paying) or Catholic (private or fee-paying) schools. The type of school attended was considered because educational policies and ethos may vary between school sectors, which may influence children's participation in extracurricular sport.
Analysis
Two-sample t-tests were used to compare the sample characteristics of boys and girls. Linear regression methods were used to determine whether parental exercise was associated with children's extracurricular sports participation and cardiorespiratory fitness (1.6 km run & PWC170). To adjust for the effects of clustering of children within schools, the estimates of standard errors were computed using the Taylor-series approximation. For these analyses, three indicator (dummy) variables were included for parental exercise (both parents active, mother active only, father active only), with the reference group being 'both parents inactive'. P-values corresponding to the test of association (multiple-partial F-test) are reported. The analyses were restricted to the 5,929 children who provided complete information on extracurricular sports participation (representing 96% of those who completed the questionnaire), and the number of sports reported was treated as a continuous variable.
Similar methods were used to examine the relationship between children's extracurricular sports participation and each of the cardiorespiratory fitness measures. As a test for trend, we categorised the number of sports into four levels and report the test of significance of a single predictor with values ranging from one (at most one sport) to four (four or more sports).
To assess whether associations between parental exercise and children's extracurricular sports participation differed by the sex of the active parent, 'mother active only' was used as the reference category and compared to the category 'father active only'. Additionally, 'both parents active' was compared to 'both parents inactive'. These analyses are presented in the text.
All analyses were stratified by sex to assess whether associations varied between boys and girls. All analyses were routinely adjusted for age, SES and school type because these variables confounded associations in a number of the analyses. We tested for effect modification by age by including a (child's age × parents' exercise) product term in each final model, and found evidence of it only in one isolated instance (the model relating girls' participation in extracurricular sport to mothers' exercise, data not reported). Stata software (Version 8.2) was used for all analyses (28).
Results
Sample characteristics
The number and percentage of boys and girls in each age group, school type, SES group, proportion overweight and obese, and parental exercise category are presented in Table 1 (n = 5,929). The majority of children attended government schools, one fifth attended Catholic schools and the remaining children attended independent schools. 10.9% of boys and 12.0% of girls were considered overweight or obese. Boys and girls reported parents' exercise similarly, with 43.4% of boys and 40.9% of girls reporting that they had two physically inactive parents.
Table 1 Total number and percentage of children participating in the ASHFS by age, school type, socioeconomic status (SES), weight status, parental exercise and number of sports played
Boys (N = 3,256) Girls (N = 3,158)
n % n %
Age
9 476 14.6 482 15.3
10 479 14.7 489 15.5
11 473 14.5 475 15.0
12 473 14.5 479 15.2
13 454 13.9 426 13.5
14 453 13.9 400 12.7
15 448 13.8 407 12.9
School Type
Government 2,403 73.8 2,335 73.9
Catholic 655 20.1 677 21.4
Independent 198 6.1 146 4.6
SES
Low 310 9.7 272 8.8
Medium-low 1,230 38.5 1,197 38.6
Medium-high 914 28.6 886 28.6
High 742 23.2 748 24.1
Weight status*
Under/acceptable weight 2,899 89.0 2,778 88.0
Overweight 301 9.2 335 10.6
Obese 56 1.7 45 1.4
Parental Exercise
Both parents active 553 18.4 630 21.5
Mother active only 518 17.3 537 18.3
Father active only 627 20.9 565 19.3
Both parents inactive 1,303 43.4 1,196 40.9
Incomplete data 255 7.8 230 7.3
Number of sports
0 sport 184 5.7 230 7.3
1 sport 586 18.0 691 21.9
2 sports 881 27.1 857 27.1
3 sports 740 22.7 658 20.8
4 sports 484 14.9 404 12.8
5 sports 235 7.2 187 5.9
6 sports 146 4.5 131 4.2
* Cut-off points for BMI in childhood based on international data linked to the widely accepted adult cut-off points for overweight (BMI 25.0–29.9 kg/m2) and obesity (≥ 30.0 kg/m2) [42]
The mean (standard deviation) weight, height, number of sports played, time taken to complete a 1.6 km run/walk and PWC170 (kgm.kg-1.min-1) are presented in Table 2. Compared to girls, boys in this sample were heavier (p = 0.008), taller (p < 0.001), played more sport (p < 0.001), completed the 1.6 km run/walk in less time (p < 0.001), and had higher PWC170 scores (p < 0.001).
Table 2 Weight, height, number of extracurricular sports played, time to complete a 1.6 km run/walk, and PWC170 (kgm.kg-1.min-1) of children aged 9–15 years in the ASHFS 1985
Mean (standard deviation) Boys Girls
Weight (kg) 44.2 (12.8) 43.4 (11.4)
Height (m) 152.3 (13.9) 150.3 (11.6)
Extracurricular sports* (n) 2.6 (1.5) 2.4 (1.5)
1.6 km run/walk time (mins & secs) 8 m 6 s (1 m 30 s) 9 m 48 s (1 m 42 s)
PWC170 (kgm.kg-1.min-1) 14.7 (3.5) 11.2 (3.0)
*Number of extracurricular sports played
Parental exercise and children's extracurricular sports participation
Table 3 shows that parental exercise was associated with children's extracurricular sports participation in both boys (p < 0.001) and girls (p < 0.001). Children who had two active parents participated in, on average, 0.6 more sports than those with two inactive parents. Where only one parent was active, the number of sports boys and girls participated in was significantly higher than where neither parent was active (p < 0.001 for all associations). Interestingly, the sex of that parent made no significant difference to the number of extracurricular sports played by either boys (p = 0.66) or girls (p = 0.21).
Table 3 Number of extracurricular sports played by children by parental exercise
Parental exercise Boys Girls
Both active 3.0 (0.07) 2.8 (0.08)
Mother active only 2.7 (0.07) 2.5 (0.09)
Father active only 2.7 (0.07) 2.6 (0.08)
Both inactive 2.4 (0.06) 2.2 (0.06)
Test of association p < 0.001 p < 0.001
Values are least squares means (standard error) adjusted for age, SES & school type
Parental exercise and children's 1.6 km run/walk time
Table 4 shows the association between parental exercise and the number of minutes taken by boys and girls to complete a 1.6 km run/walk. On average, boys who had two physically active parents completed the 1.6 km run/walk 18 seconds faster than boys of two physically inactive parents (p < 0.001), and girls of two physically active parents completed the 1.6 km run/walk on average 24 seconds faster than girls of two physically inactive parents (p < 0.001). Compared to children with two inactive parents, girls whose mother only was active, and boys whose father only was active, completed the 1.6 km run/walk faster (p = 0.05 and p = 0.08 respectively). There was no significant difference in the 1.6 km run/walk time between children with no active parents compared to boys whose mother only was active (p = 0.41) and girls whose father only was active (p = 0.10). When only one parent was active, the sex of that parent made no significant difference to the time taken to complete the 1.6 km run/walk by boys (p = 0.53) and girls (p = 0.85).
Table 4 Average minutes and seconds (standard deviation) taken by children to complete a 1.6 km run/walk by parental exercise
Parental exercise Boys Girls
Both active 7 m 48 s (4.2 s) 9 m 30 s (4.8 s)
Mother active only 8 m 6 s (4.2 s) 9 m 42 s (5.4 s)
Father active only 8 m 0 s (3.6 s) 9 m 48 s (6.0 s)
Both inactive 8 m 6 s (3.0 s) 9 m 54 s (5.4 s)
Test of association p < 0.001 p < 0.001
Values are least squares means (standard error) adjusted for age, SES & school type
Parental exercise and children's PWC170
A subset of the sample, which included those children aged 9, 12 and 15 years, participated in a PWC170 test. Table 5 shows that parental exercise was significantly associated with girls' PWC170 (p = 0.002), but not with boys' (p = 0.55). Girls who had two physically active parents had PWC170 scores on average 0.7 kgm.kg-1.min-1 greater than girls of two physically inactive parents (p = 0.005). When only one parent was active, the sex of that parent did make a difference to PWC170. Girls achieved significantly higher scores if their mother only was active than if their father only was active (p = 0.02). While having an active mother only was associated with higher PWC170 scores in boys than having an active father only, this difference was not statistically significant (p = 0.63). Similarly, while girls with an active mother had higher scores than girls with two active parents, this difference was not statistically significant (p = 0.56).
Table 5 PWC170 (kgm.kg-1.min-1) in children by parental exercise
Parental exercise Boys Girls
Both active 14.9 (0.25) 11.6 (0.21)
Mother active only 14.7 (0.28) 11.7 (0.24)
Father active only 14.5 (0.22) 11.0 (0.27)
Both inactive 14.7 (0.23) 10.9 (0.18)
Test of association p = 0.55 p = 0.002
Values are least squares means (standard error) adjusted for age, SES & school type
Children's extracurricular sports participation and cardiorespiratory fitness
In general, participation in a greater number of extracurricular sports was reflected by a higher level of cardiorespiratory fitness. Table 6 shows the average minutes taken to complete a 1.6 km run/walk and the PWC170 score according to the number of extracurricular sports played by children. A dose-response relationship existed between extracurricular sports participation and minutes taken to complete a 1.6 km run/walk in all children (p < 0.001), between extracurricular sports participation and PWC170 in girls (p < 0.001), and a borderline association between extracurricular sport participation and PWC170 in boys (p = 0.09). Boys who played four or more sports completed the 1.6 km run/walk on average 30 seconds faster than boys who played at the most one sport, and girls who played four or more sports completed the 1.6 km run/walk on average 42 seconds faster than girls who played at the most one sport. Boys who played four or more sports had PWC170 scores on average 0.4 kgm.kg-1.min-1 greater than boys who played at the most one sport, and girls who played four or more sports had PWC170 scores on average 1.3 kgm.kg-1.min-1 greater than girls who played at most one sport.
Table 6 Children's extracurricular sports participation by cardiorespiratory fitness, measured by a 1.6 km run/walk time (minutes) and PWC170 (kgm.kg-1.min-1)
Number of sports Cardiorespiratory Fitness
1.6 km run/walk (mins and secs) PWC170 (kgm.kg-1.min-1)
Boys Girls Boys Girls
0–1 sport 8 m 18 s (3.6 s)
(n = 727) 10 m 6 s (4.8)
(n = 838) 14.5 (0.21)
(n = 339) 10.6 (0.17)
(n = 382)
2 sports 8 m 12 s (3.0 s)
(n = 837) 9 m 54 s (4.2)
(n = 788) 14.6 (0.17)
(n = 358) 11.0 (0.15)
(n = 358)
3 sports 8 m 0 s (3.0 s)
(n = 696) 9 m 42 s (4. 2)
(n = 607) 14.7 (0.16)
(n = 314) 11.5 (0.16)
(n = 268)
4+ sports 7 m 48 s (3.6 s)
(n = 814) 9 m 24 s (4.8)
(n = 652) 14.9 (0.19)
(n = 322) 11.9 (0.21)
(n = 278)
Test for trend p < 0.001 p < 0.001 p = 0.09 p < 0.001
Values are least squares means (standard error) adjusted for age, SES & school type
Discussion
The findings from this study suggest that parental exercise is positively associated with children's extracurricular sports participation and cardiorespiratory fitness. Having two physically active parents was associated with participation in a significantly higher number of extracurricular sports and with significantly greater cardiorespiratory fitness, compared to having two physically inactive parents. When only one parent was active, the sex of that parent was not an independent predictor of the child's extracurricular sports participation and cardiorespiratory fitness in most instances. These associations remained after adjusting for age, SES and school type.
While the magnitude of the differences observed in the current study were small, they may still be of public health significance. In adults, clear associations have been demonstrated between physical activity and cardiorespiratory fitness and various health outcomes, including all-cause mortality [26,27] and cardiovascular disease [28-30]. Other research has suggested that physical activity behaviours in childhood may predict physical activity behaviours in adulthood [4,31,32]. The contribution that childhood behaviours make to adult health is unclear, but it is possible that small associations such as those seen in the present study may play a role.
There are a number of possible mechanisms, also identified in other studies, through which parental exercise may influence children's extracurricular sports participation. Firstly, it is plausible that parents act as role models for children's extracurricular sports participation. When children observe that their parents are actively involved in and value sport and exercise, they may adopt these values and behaviours themselves. Secondly, a recent review of correlates of children and adolescent's physical activity found parental encouragement and support to be more strongly associated with children and adolescents' participation in physical activity than parental role modeling [8]. It is possible that physically active parents provide more support and encouragement for their children to be active and influence their children's activity levels in this way. Thirdly, it may be that a genetic predisposition to physical activity exists. In a recent review, Beunen and Thomis reported that the heritability coefficients for sports participation ranged from 0.35–0.83, and that children who had a parent active in sport had 1.2–5.8 times the odds of participating in sport than children whose parents were not active in sport [33].
In the current study, an association was also seen between parental exercise and children's cardiorespiratory fitness. It is possible that by influencing children's physical activity, parental exercise may indirectly influence children's fitness. However, a recent analysis found that while more active children tended to be fitter, children's physical activity only had a modest correlation with fitness (r = 0.17) [34]. It is also possible that parental fitness may influence children's fitness through genetic mechanisms. Results from studies assessing familial resemblance of cardiorespiratory fitness have found varying results, with the contribution of genetic factors ranging from 20–50% [35,36].
Interestingly, our findings are consistent with the results of other studies that have used similar measures of physical activity. For instance, both Anderssen and Wold [15] and Gregson and Colley [10] measured children's physical activity using self-reported questionnaires and measured parental physical activity using a questionnaire with children. Our findings also parallel the results of studies that have used accelerometers to objectively measure physical activity in children and adults [9,13]. However, studies that have used a parental self-report of physical activity have generally found no association between children's physical activity and parental physical activity [11,16-18,20].
Little research has explored the reliability and validity of children's reports of their parents' physical activity. One study of 198 Grade 7–9 students found that children's report of their parents' physical activity correlated with parental self-reported physical activity at low to moderate levels (r = 0.16–0.47), except in Grade 9 girls (r = -0.01) [16]. Another study of 755 Norwegian families demonstrated a moderate correlation between parents' self-reports of physical activity and their 13 year-old children's reports of parental physical activity (r 0.42–0.54, p < 0.001) [37]. This suggests that the children in the current study were likely to report reasonable estimates of their parents' exercise. It is possible that children's perceptions of their parents' exercise behaviours may have a greater influence on children's physical activity than parents' exercise behaviours themselves. However, studies that have assessed parental physical activity objectively have found similar associations as the current study.
The current study found very little difference between the association of mothers' exercise compared to fathers' exercise with children's extracurricular sports participation and cardiorespiratory fitness, with the exception of a stronger association of mothers' exercise and girls' PWC170. These findings generally suggest that as long as one parent is active, the sex of the active parent makes little difference to extracurricular sports participation and cardiorespiratory fitness, with the exception of PWC170 in girls, where having an active mother only was associated with greater fitness than having an active father only.
These results contrast with previous research that found fathers' activity to be more influential than mothers' activity [9,12], mothers' activity to be more influential on girls' physical activity [10,11], mothers' activity to be more influential than fathers' activity on boys' physical activity [38] and parents of the same sex as the child to have the most influence on their child's physical activity [39]. These inconsistencies may be due to different definitions and assessments of children's physical activity. In the current study, the extracurricular sports component of physical activity was the main outcome factor, whereas previous studies have examined children's leisure time physical activity. Previous research has found that parents play a strong 'gate keeping' role in their children's extracurricular sports participation [18]. Parents are the usual providers of transport, equipment and funding typically not required for general and school-based physical activity. It is possible that as long as one parent provides support for extracurricular activity, the sex of that parent is not important. The sex of the active parent may be more important in influencing other domains of children's physical activity, such as active play or household activity. The differences in findings between this and other studies may also be related to the differing measures of physical activity. Some measures may "capture" some activities – such as high intensity, structured activities – better than others, and it is plausible that men and women participate in different types of activities that are "captured" differently.
A limitation of this study was that regular parental exercise was defined in the ASHFS 1985 survey as two or more times per week, while guidelines dating back to the 1960s recommend physical activity on three to five days per week for health benefits [40]. Although the ASHFS definition of regular physical activity may not meet current guidelines, this data still provides an estimate of those parents who participated in regular exercise and those who did not. A second limitation is that this study did not collect information on the frequency, duration or intensity of the extracurricular sports. It is possible that children who listed six sports may only participate in these occasionally, whereas children who listed one sport may participate every day for a number of hours. Additionally, children were asked to lists sports they had played "regularly" in the past year, yet a definition of "regularly" was not provided. The interpretation of this question by the children may therefore be variable. However, it was interesting to note the strong and consistent relationship seen between the number of extracurricular sports listed and both cardiorespiratory fitness measures in boys and in girls.
As evidenced in the current study, parental exercise may be important for children's extracurricular sports participation and cardiorespiratory fitness, which may influence children's health later in life. In the nationwide Active Australia survey, parents were deemed one of six population groups at risk of physical inactivity [41]. Those parents with one or more children under the age of 18 living at home were 20% less likely to be considered sufficiently active than those adults without children [41]. The low national levels of parental physical activity in Australia are consequently of concern. While these data are historical, we believe it would be useful for other researchers to use the data reported here to assess whether cultural changes, technological advances and increases in overweight and obesity have impacted on the apparent association between children's extracurricular sports participation, cardiorespiratory fitness and parental exercise.
Conclusion
The results of this cross-sectional study suggest that parental exercise may influence children's participation in extracurricular sport and their cardiorespiratory fitness. While the magnitude of the differences seen was small, these differences may be important for future health. Because Australian parents are also a group at risk of physical inactivity, targeting parents may provide health benefits for the whole family.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
VC conceptualised and drafted the paper. AV contributed to the conceptualisation of the paper, interpretation of the data and critical revision of the paper. JF provided statistical support and critical revision of the paper. TD was involved in the design, management and data collection in the ASHFS, contributed to the interpretation of the data and provided critical revision of the paper. LB provided statistical guidance and critical revision of the paper. All authors read and approved the final manuscript.
Acknowledgements
The authors wish to acknowledge all those who contributed to the scientific development of this project, and to those who were involved in the data collection process. We also wish to thank all of the children who participated in this project. The Australian Council for Health and Physical Education Research's Australian Schools Health and Fitness Survey was jointly funded by the Commonwealth Department of Sport, Recreation and Tourism, the Commonwealth Schools Commission, the Commonwealth Department of Health and the National Heart Foundation of Australia.
==== Refs
Morris L Sallybanks J Willis K Sport, physical activity and antisocial behaviour in youth 2003 Canberra, Australian Institute of Criminology
Harrison PA Narayan G Differences in behavior, psychological factors, and environmental factors associated with participation in school sports and other activities in adolescence J Sch Health 2003 73 113 120 12677730
Steptoe A Butler N Sports participation and emotional wellbeing in adolescents Lancet 1996 347 1789 1792 8667922 10.1016/S0140-6736(96)91616-5
Tammelin T Nayha S Hills AP Jarvelin MR Adolescent participation in sports and adult physical activity Am J Prev Med 2003 24 22 28 12554020 10.1016/S0749-3797(02)00575-5
Pyke JE Australian Health and Fitness Survey 1985 1985 South Australia, The Australian Council for Health, Physical Education and Recreation
Australian Bureau of Statistics Children's participation in cultural and leisure activities 2000 Canberra, Australian Bureau of Statistics
Magarey AM Daniels LA Boulton TJ Prevalence of overweight and obesity in Australian children and adolescents: reassessment of 1985 and 1995 data against new standard international definitions Med J Aust 2001 174 561 564 11453327
Sallis JF Prochaska JJ Taylor WC A review of correlates of physical activity of children and adolescents Med Sci Sports Exerc 2000 32 963 975 10795788 10.1097/00005768-200005000-00014
Moore LL Lombardi DA White MJ Campbell JL Oliveria SA Ellison RC Influence of parents' physical activity levels on activity levels of young children J Pediatr 1991 118 215 219 1993947
Gregson JF Colley A Concomitants of sport participation in male and female adolescents Int J Sp Psy 1986 17 10 22
McMurray RG Bradley CB Harrell JS Bernthal PR Frauman AC Bangdiwala SI Parental influences on childhood fitness and activity patterns Res Q Exerc Sport 1993 64 249 255 8235046
Greendorfer SL Lewko JH Role of family members in sport socialization of children Res Q 1978 49 146 152 725281
Kalakanis LE Goldfield GS Paluch RA Epstein LH Parental activity as a determinant of activity level and patterns of activity in obese children Res Q Exerc Sport 2001 72 202 209 11561385
Freedson PS Evenson S Familial aggregation in physical activity Res Q Exerc Sport 1991 62 384 389 1780560
Anderssen N Wold B Parental and peer influences on leisure-time physical activity in young adolescents Res Q Exerc Sport 1992 63 341 348 1439157
Godin G Shephard RJ Colantonio A Children's perception of parental exercise: influence of sex and age Percept Mot Skills 1986 62 511 516 3503259
Dempsey JM Kimiecik JC Horn TS Parental influence on children's moderate to vigorous physical activity participation: an expectancy-value approach Ped Exerc Sci 1993 5 151 167
Sallis JF Alcaraz JE McKenzie TL Hovell MF Kolody B Nader PR Parental behavior in relation to physical activity and fitness in 9-year-old children Am J Dis Child 1992 146 1383 1388 1415081
Trost SG Sallis JF Pate RR Freedson PS Taylor WC Dowda M Evaluating a model of parental influence on youth physical activity Am J Prev Med 2003 25 277 282 14580627 10.1016/S0749-3797(03)00217-4
Welk GJ Wood K Morss G Parental influences on physical activity in children: an exploration of potential mechanisms Ped Exerc Sci 2003 15 19 33
Dwyer T Sallis J Blizzard CL Lazarus R Dean K Relation of academic performance to physical activity and fitness in children Ped Exerc Sci 2001 13 225 237
Dwyer T Gibbons LE The Australian Schools Health and Fitness Survey: physical fitness related to blood pressure but not lipoproteins Circulation 1994 89 1539 1544 8149519
King AC Blair SN Bild DE Dishman RK Dubbert PM Marcus BH Oldridge NB Paffenbarger RSJ Powell KE Yeager KK Determinants of physical activity and interventions in adults Med Sci Sports Exerc 1992 24 S221 36 1625548
Gordon-Larsen P McMurray RG Popkin BM Determinants of adolescent physical activity and inactivity patterns Pediatrics 2000 105 E83 10835096 10.1542/peds.105.6.e83
McLennan W 1996 Census of population and housing: socioeconomic index for areas 1998 Canberra, Australian Bureau of Statistics
Lee IM Skerrett PJ Physical activity and all-cause mortality: what is the dose-response relation? Med Sci Sports Exerc 2001 33 S459 71; discussion S493-4 11427772
Paffenbarger RSJ Hyde RT Wing AL Hsieh CC Physical activity, all-cause mortality, and longevity of college alumni N Engl J Med 1986 314 605 613 3945246
Kannel WB Belanger A D'Agostino R Israel I Physical activity and physical demand on the job and risk of cardiovascular disease and death: the Framingham Study Am Heart J 1986 112 820 825 3766383 10.1016/0002-8703(86)90480-1
Paffenbarger RSJ Hyde RT Hsieh CC Wing AL Physical activity, other life-style patterns, cardiovascular disease and longevity Acta Med Scand Suppl 1986 711 85 91 3535417
Wannamethee SG Shaper AG Physical activity in the prevention of cardiovascular disease: an epidemiological perspective Sports Med 2001 31 101 114 11227978
Kuh DJ Cooper C Physical activity at 36 years: patterns and childhood predictors in a longitudinal study J Epidemiol Community Health 1992 46 114 119 1583424
Telama R Leskinen E Yang X Stability of habitual physical activity and sport participation: a longitudinal tracking study Scand J Med Sci Sports 1996 6 371 378 9046549
Beunen G Thomis M Genetic determinants of sports participation and daily physical activity Int J Obes Relat Metab Disord 1999 23 Suppl 3 S55 63 10368004 10.1038/sj.ijo.0800885
Trost SG The association between physical activity and cardiorespiratory fitness in children and
adolescents: A meta-analytic review. Report prepared for the University of South Carolina School
of Public Health 1998 South Carolina, University of South Carolina
Perusse L Lortie G Leblanc C Tremblay A Theriault G Bouchard C Genetic and environmental sources of variation in physical fitness Ann Hum Biol 1987 14 425 434 3688830
Bouchard C Daw EW Rice T Perusse L Gagnon J Province MA Leon AS Rao DC Skinner JS Wilmore JH Familial resemblance for VO2max in the sedentary state: the HERITAGE family study Med Sci Sports Exerc 1998 30 252 258 9502354
Anderssen N Jacobs DRJ Aas H Jakobsen R Do adolescents and parents report each other's physical activity accurately? Scand J Med Sci Sports 1995 5 302 307 8581574
Fogelholm M Nuutinen O Pasanen M Myohanen E Saatela T Parent-child relationship of physical activity patterns and obesity Int J Obes Relat Metab Disord 1999 23 1262 1268 10643682 10.1038/sj.ijo.0801061
Snyder E Spreitzer D Family influences and involvement in sports Res Q 1973 44 249 255
US Department of Health and Human Services Physical activity and health: a report of the Surgeon General 1996 Atlanta, GA, United States Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion
Armstrong T Bauman A Davies J Physical activity patterns of Australian adults: results of the 1999 National Physical Activity Survey 2000 Canberra, Australian Institute of Health and Welfare
Cole TJ Bellizzi MC Flegal KM Dietz WH Establishing a standard definition for child overweight and obesity worldwide: international survey BMJ 2000 320 1240 1243 10797032 10.1136/bmj.320.7244.1240
| 15811190 | PMC1087868 | CC BY | 2021-01-04 16:38:04 | no | Int J Behav Nutr Phys Act. 2005 Apr 6; 2:3 | utf-8 | Int J Behav Nutr Phys Act | 2,005 | 10.1186/1479-5868-2-3 | oa_comm |
==== Front
Int Semin Surg OncolInternational seminars in surgical oncology : ISSO1477-7800BioMed Central London 1477-7800-2-91581999410.1186/1477-7800-2-9ResearchResearch: Is resection of tumours involving the pelvic ring justified? : A review of 49 consecutive cases Yuen Alex [email protected] Eugene T [email protected] Peter FM [email protected] Department of Orthopaedics, University of Melbourne, St. Vincent's Hospital, 41 Victoria Parade, Fitzroy, 3065, Victoria, Australia2 Sarcoma Unit, Division of Surgical Oncology, Peter MacCallum Cancer Institute, Melbourne, Australia3 Department of Surgery, St. Vincent's Hospital, 41 Victoria Parade, Fitzroy, 3065, Victoria, Australia2005 9 4 2005 2 9 9 6 12 2004 9 4 2005 Copyright © 2005 Yuen et al; licensee BioMed Central Ltd.2005Yuen 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
Pelvic surgery is challenging and impacts significantly on limb and visceral function, thus, raising the question "is heroic surgery justifiable". This study assessed the functional, oncologic and surgical outcomes following pelvis tumour resections.
Methods
Between 1996–2003, 49 patients (mean age 43 years) underwent pelvic tumour resections- 38 primary malignant tumours, 5 secondary tumours and 6 benign tumours. Bone tumours comprised 5 osteosarcomas, 5 Ewings sarcomas, and 12 chondrosarcomas. Of the soft tumours, 9 were of neural origin. Tumours involved the ilium, acetabulum, pubic bones, sacrum or a combination of these. Functional assessment was performed and no patient had metastases at presentation.
Results
There were 41 limb sparing resections and 8 hindquarter amputations. Surgical margins were intralesional (1), marginal (13), wide (26), and radical (3). Of limb sparing surgery, prosthetic reconstructions were performed in 10 patients, biologic reconstructions in 6, a combination of these in 3 and no reconstruction in others. There was 1 intraoperative death, 7 local recurrences and 19 metastases. Death from disease occurred at a mean of 14.2 months with a mean followup of 27 (1–96) months. Amputation and periacetabular resections had worse functional outcomes. Emotional acceptance was surprisingly high.
Conclusion
Pelvic resections are complex. Functional outcome is significantly affected by surgery. Disease control is similar to limb tumours. Emotional acceptance of surgery in survivors was surprisingly high. Major pelvic resection for malignancy appears justified.
==== Body
Background
Pelvic resection is challenging and can have significant functional, social and psychological impact on the patient. It is also associated great morbidity especially when extensive resection is required to attain adequate surgical margins. With the introduction of adjuvant chemotherapy and radiotherapy, and improved reconstruction techniques, limb- sparing surgery has largely replaced external hemipelvectomy. However, given the often poor prognosis of patients with pelvic tumours, the question of whether this type of surgery is justified remains unanswered. This study assessed the functional, oncologic and surgical outcomes following pelvic tumour resection.
Methods
Patients
From 1996 to 2003, 49 patients at St. Vincent's Hospital underwent pelvic tumour resections. Clinical and functional information on all 49 patients were available for review. There were 25 males and 24 females. The mean age was 43.2 (15 – 72) years. 38 patients underwent surgery for primary malignancy, 6 patients for benign tumour, while 5 patients had surgery for metastasis.
Tumour characteristics
Of the 38 primary malignancy cases, there were 5 osteosarcomas, 5 Ewing sarcoma, and 12 chondrosarcoma. Of the 5 metastatic tumours, there was 1 colorectal cancer, 1 vulvar cancer, 1 renal cell cancer, 1 squamous cell cancer and one Lung cancer. The histological subgroup is summarised in Table 1.
Table 1 Tumour Types
Primary Tumours Number
Central osteosarcoma 1
Chondroblastic osteosarcoma 4
Paget's Osteosarcoma 1
Undifferentiated sarcoma 1
Ewing Sarcoma 5
Chondrosarcoma 12
Chordoma 4
MFH 4
Desmoid tumour 1
Leiomyosarcoma 1
Malignant periphereal nerve sheath tumour 1
Spindle cell carcinoma 1
Neurofibrosarcoma 1
Rhabdomyosarcoma 1
Neurofibroma 1
Osteochondroma 2
Schwannoma 2
Chondromyxoid fibroma 1
Secondary Tumours
Colorectal 1
Vulvar 1
Renal cell cancer 1
Squamous cell carcinoma 1
neuroendocrine cancer from the lung 1
According to the Enneking's classification, one patient had stage 1A tumour, one patient had stage 1B tumour, 6 patients had stage 2A tumours and 30 patients had 2B tumours.
Tumours involved the ilium (P1), the periacetabular region (P2) with or without involvement of the proximal femur (H1), the pubis (P3), the sacrum (P4) or a combination of these. The location and the corresponding histological subgroup is summarised in Table 2.
Table 2 Location in relation to type of tumour
P1 P1/2 P2 P2/H1 P2/3 P3 P4 P1/4
BENIGN 3 0 0 0 0 0 3 0
SARCOMA 5 4 4 4 9 2 5 5
CARCINOMA 1 0 1 0 1 2 0 0
Operations
Pelvic resections were classified according to Enneking and Dunham into 4 types; namely, iliac (T1), acetabular (T2), pubis or ischium (T3) and sacral (T4). Combinations of these resections with or without high femoral resection (H1) were also performed (Figure 1).
Figure 1 An illustration of the types of pelvic resection.
The types of resection and the methods of reconstruction are summarised in Table 3.
Table 3 Types of resection and Methods of reconstruction
T1 T1/2 T2 T2/H1 T2/3 T3 T4 T1/4
Limb Sparing Surgery (41) 7 3 0 3 9 3 7 7
Amputation (8) 8
Reconstruction (19) 1 FVFG 1 pelvic reconstruction with bone cement 3 ischiofemoral pseudoarthrodesis 2 pelvic allograft with hip arthroplasty 1 saddle prosthesis with reconstruction plate 8 saddle prosthesis 1 pelvic allograft with hip arthroplasty 1 FVFG 1 ischiofemoral pseudoarthrodesis
The surgical margins achieved were classified according to the criteria established by the Musculoskeletal Tumour society; namely intralesional, marginal, wide, and radical.
Adjuvant Treatment
9 patients received chemotherapy alone, 6 received radiotherapy alone, and 6 patients received both chemotherapy and radiotherapy.
Follow-up
Follow – up was calculated from the time of surgery to the last date of review or death. The series was updated by reviewing the clinical charts of the patients. In addition, a standardized questionnaire regarding clinical outcome and function was completed for every patient by way of phone interview. Evaluation of disease status and functional result at last follow-up review was performed for all patients. The mean length of follow-up was 27 months (1 to 96 months). The distribution of the follow-up period is summarised in Figure 2.
Figure 2 A bar graph showing the distribution of the follow-up periods.
Functional assessment
Function was assessed by the modified function evaluation system recommended by Enneking et al [1] With this system, functional assessment is based on an analysis of factors (pain, functional activities, emotional acceptance) pertinent to the patient as a whole and factors specific to the lower limb (use of external supports, walking ability, and gait). For each of the six factors, values of 0 to 5 are assigned on the basis of established criteria. Descriptive terms like excellent, good, fair or poor are assigned to a specific numerical range. (26–30 Excellent, 21–25 Good, 16–20 Satisfactory, 11–15 Fair, and 10 or less Poor)
Results
Duration of surgery
The mean operative time was 5.2 (1.5 – 10) hours.
Blood loss
The mean number of packed cell units for intra-operative transfusion was 10 (2 – 26) units.
Length of stay
The mean length of in – hospital stay was 23 (2 – 110) days.
Oncologic Outcome
Of the 5 patients who underwent surgery for metastatic carcinoma, marginal margins were achieved in 3 patients, and wide margins were achieved in the other 2 patients.
Of the 38 patients who underwent surgery for sarcoma, intralesional margin was achieved in 1 patient, marginal margins were achieved in 10 patients, wide margins were achieved in 24 patients and radical margins were achieved in 3 patients.
The oncologic outcomes in relation to tumour type, previous biopsy and surgical margin are shown in Table 4. 7 patients had local recurrence. 19 patients developed metastasis. 19 patients died of disease. The mean survival of these patients was 14.2 (1 – 51) months.
Table 4 Oncologic outcome in relation to surgical margin, tumour type and previous biopsy.
MARGINS
CARCINOMA n L.R. Prev. Biopsy Mets
INTRALESIONAL 0 0 0 0
MARGINAL 3 1 1 1
WIDE 2 0 0 2
SARCOMA
INTRALESIONAL 1 1 1 1
MARGINAL 10 3 1 3
WIDE 24 2 1 9
RADICAL 3 0 0 3
Complications
22 of the 49 patients had complications.
There was one intraoperative death.
3 patients had common peroneal nerve palsy and 1 patient had sciatic nerve palsy. 2 patients had urinary incontinence and 1 patient had erectile dysfunction after Type 4 pelvic resection.
14 patients had wound infections: 2 of these were superficial infections and 12 were deep infections.
1 patient had a wound hematoma.
3 patients had dislocation/disarticulation of the saddle prosthesis.
3 patients after external hemipelvectomy had phantom limb pain.
13 patients required additional surgeries. 6 of these patients required open drainage or debridement of infected wound. 1 patient required removal of prosthesis due to infection. 2 patients required subsequent split-skin graft. 2 patients required subsequent rectus flaps. 1 patient required a rhomboid flap. 1 patient required a Latissimus Dorsi flap but due to recurrent wound infections subsequently underwent external hemipelvectomy.
The complications in relation to the type of pelvic resection are shown in Table 5.
Table 5 Types of resection and complications
T1 T1/2 T2 T2/H1 T2/3 T3 T4 T1/4 AMP Total
Infection 2 3 1 1 2 5 14
Hematoma 1 1
skin/flap problem 2 2
Venous thrombus 1 1 2
nerve injury 1 1 2 1 5
Phantom limb pain 3 3
Urinary incontinence 2 2
Erectile dysfunction 1 1
Prosthetic dislocation 3 3
Require additional surgery 1 3 1 2 2 4 13
Functional Outcome
The outcomes for 44 patients were assessed. According to the criteria of Enneking et al [1], 11 patients had excellent functional results, 7 patients had good functional results, 9 patients had satisfactory functional results, 5 patients had fair functional results, and 12 patients had poor results.
The functional results in relation to the performed surgical procedure are shown in Table 6.
Table 6 Functional Result in relation to surgical of procedure
T1 T1/2 T2 T2/H1 T2/3 T3 T4 T1/4 AMP
Mean Functional Score 22.8 19.3 18 12 27 25 19 5
Range (12 – 29) (19 – 20) (15 – 21) (3 – 18) (15 – 30) (8 – 28) (0 – 10)
Emotional Acceptance 4.5 4.3 4.5 2.7 4.5 3.9 3.3 2
Functional score excluding the emotional acceptance component 18.3 (73.2%) 15 (60%) 13.5 (54%) 9.3 (37.2%) 22.5 (90%) 21.1 (84.4%) 15.7 (62.8%) 3 (12%)
Of the internal hemipelvectomy, Type 3 and Type 4 pelvic resections have the best mean overall functional scores and are both in the range of excellent functional result. On the other hand, combined type 2 and 3 resection has the worst mean overall functional score and falls in the range of fair functional result.
The mean overall function score for external hemipelvectomy falls in the range of poor functional result.
When the emotional acceptance score is excluded from the overall functional score, Type 3 and Type 4 pelvic resections still have the best total score. Resections involving the hip joint (i.e. T1/2, T2/H1, and T2/3) all scored below 60% of the total.
All internal hemipelvectomy have mean emotional acceptance score of greater than 4 except combined type 2 and 3 pelvic resection, combined type 1 and 4 resection, and type 4 resection. (They scored 2.7, 3.3 and 3.9 respectively)
External hemipelvectomy has the lowest mean emotional acceptance score.
Discussion
Pelvic resections are complex. They are technically difficult due to the usually large size of pelvic tumours, and the close proximity of pelvic viscera and neurovasculature of the pelvis and lower limb. The technical and human resources required are intensive. Patients often require large amount of blood products. They require post-operative recovery in ICU and in most cases long periods of in-patient rehabilitation. Most importantly, the diagnosis and management of pelvic tumour require a multidisciplinary approach, which includes radiologist, oncologist, physiotherapist, Occupation therapist, and Orthotist. Hence, patients must be treated in specialised centres where such resources are available.
As shown in our series, the morbidity is great with a high risk of wound infection and nerve injury. The overall complication rate has been reported as high as 50–60% [2]. In the current series, 22 out of 49 patients (45%) had complications. Regardless of the type of resection, there is a significant functional, social and psychological impact on the patient. Moreover, there is a significant risk of peri -operative mortality.
Because of these issues, there remains the question as to whether pelvic surgery performed with curative intent or for palliation is justified particularly for patients with high grade malignancy where the prognosis is poor. However, the results of the current series show that this type operation may be worthwhile.
In regards to oncologic outcome, forty out of the forty-nine patients survived longer than six months. The mean survival was greater than one year. The local failure rate was 16%, and 40% of patients developed metastasis. These figures are supported by Pring et al [3] who reported a similar local failure rate of 19% and an overall survival rate of 69%. Ozaki et al [4] reported a higher local recurrence rate of 60% and a poorer 5-year survival rate of 27%. However, in that study only patients with osteosarcoma were studied and 70% of patients had inadequate surgical margin. Disease control by pelvic resection is comparable to that of surgery for limb tumors. Bacci et al [5] reported, in his series of 526 patients with osteosarcoma of the extremities, a local recurrence rate of 6% and an overall survival rate of 70%. Sluga M et al [6] in his series of 130 patients also found a comparable LR rate of 2.3% and OS of 71% for limb-sparing surgery. The comparable oncologic result of pelvic resection to surgery for limb tumours is surprising because the response of pelvic tumors to adjuvant therapy is generally poor and the adequacy of surgical margin difficult to achieve.
The functional outcome varied with the level of resection but more than 50% of patients had satisfactory or better overall functional results in the current series. This is supported by Wirbel et al [7] who also found more than 60% of patients had good or excellent functional results in his series of 93 patients.
In the current study, we observed, not surprisingly, a significantly worse functional score in patients who had hindquarter amputation compared to those who had limb sparing surgery. Emotional acceptance was likewise poor in the hemipelvectomy group. Other series [7,8] have observed similar results. We found that the reasons for the poor emotional acceptance are largely due to loss of mobility and the common complication of phantom limb pain.
In addition, patients who underwent external hemipelvectomy in the current series had the highest rate of wound infection at 56%. They also had the worst survival rate with only 11% of patients alive at the time of follow-up. These figures are supported by Masterson et al [9] who reported a 79% incidence of wound infection and 8 deaths within a year among 22 patients. These results further support the view that careful patient selection is required for hindquarter amputation. In recent years, hemipelvectomy is only performed for patients in whom extensive bone or soft tissue resection makes reconstruction difficult or leaves the leg with poor function. Involvement of the sciatic nerve necessitating resection of the nerve has not been considered a contraindication to limb-sparing surgery [10].
With Limb-Sparing Surgery, Type 3 and Type 4 resections had the best results when the function is examined separately from the emotional acceptance component. On the other hand, resections that involve the hip joint confer the worst results. The reason could be the many problems associated the methods of reconstruction.
Ischiofemoral arthrodesis and pseudoarthrosis are associated with shortening of the leg and lack of mobility [11]. The restricted range of flexion and extension permitted at the pubic symphysis may also result in aching symptoms. They also have long consolidation times, which means patients require longer periods of rehabilitation and use of gait support. Enneking [10] and Menendez [12] found that the maximum possible activity was achieved after an average rehabilitation period of 14.2 months.
Saddle prosthesis is suited for bridging large area of defect in cases where part of the iliac crest could be preserved. This form of endoprosthetic replacement provides good cosmetic result [11]. However, the eccentric position of the new hip center reduces the range of movement. Moreover, if major parts of the ilium were resected, loosening of the prosthesis with lateral shift of the prosthesis could be a long-term problem [7]. In this regard, Dacron ties have been advocated for use to secure the saddle while a pseudo – capsule develops [13]. In the current series, 3 out of 9 patients had dislocation of the prosthesis.
There are other alternatives, which include reconstruction with allografts with or without hip arthroplasty. This was performed for 3 patients in the current series. Their overall functional result was only fair to satisfactory but all had high emotional acceptance score. Langlais et al [14] found, in a small series of 12 patients, 8 patients had good to excellent overall functional results. Bell at al [15] in a larger series of 17 patients similarly found a high functional result among his patients. However, there are major complications associated with this procedure in particular deep infection and graft disintegration. The risks of these complications are relatively high [16]. Moreover, this procedure is restricted by the availability of allograft bone [17].
Regardless of the functional outcome, it is interesting that most patients except those who had amputation found the procedure acceptable. This is supported by Hillmann et al [2] who also found high acceptance scores for all types of resection except amputation and acetabular resection with pelvic prosthetic reconstruction. The high emotional acceptance of patients is probably the most important factor that advocates pelvic resections: particularly limb sparing surgery.
It should be noted that the recent introduction of transxemic acid has had a significant impact in the reduction of operative bleeding and need for blood transfusion. Transxemic acid has been shown to be effective in many types of surgery including major cardiac surgery, liver transplantation, and hip and knee arthroplasty [18-21]. Although not examined in the current study, its use in pelvic surgery will likely to have the same beneficial effect.
Conclusion
Pelvic surgery is challenging. The morbidity may be great and requirements on technical and human resources are high. Disease control is similar to limb tumours. Functional result and emotional acceptance of patients are generally high. The relief of pain and improvement in function in both curative and palliative setting is extremely rewarding. Major pelvic resection for malignancy appears justified.
Competing Interests
The author(s) declare that they have no competing interests.
Authors' Contributions
Dr. Alex Yuen collected the clinical data, wrote and prepared manuscript.
Dr. Eugene Ek participated in data collection.
Prof Peter Choong carried out the surgery, and participated in data collection and preparation of manuscript.
==== Refs
Enneking WF Dunham W Gebhardt MC Malawar M Pritchard DJ A system for the functional Evaluation of reconstructive procedures after surgical treatment of tumours of the musculoskeletal system Clin Orthop 1993 286 241 246 8425352
Hillmann A Hoffmann C Gosheger G Rodl R Winkelmann W Ozaki T Tumours of the pelvis: complications after reconstruction Arch Ortho Trauma Surg 2003 123 340 4 10.1007/s00402-003-0543-7
Pring ME Weber KL Unni KK Sim FH Chondrosarcoma of the pelvis. A review of sixty – four cases J Bone Joint Surg Am 2001 83-A 1630 42 11701784
Ozaki T Hillmann A Bettin D Wuisman P Winkelmann W High complication rates with pelvic allograft. Experience of 22 sarcoma resections Acta Orthop Scand 1996 67 333 338 8792734
Bacci G Ruggieri P Bertoni F Ferrari S Longhi A Biagini R Zavatta M Versari M Forni C Local and systemic control for osteosarcoma of the extremity treated with neoadjuvant chemotherapy and limb salvage surgery: the Rizzoli experience Oncol Rep 2000 7 1129 33 10948351
Sluga M Windhager R Lang S Heinzl H Bielack S Kotz R Local and systemic control after ablative and limb sparing surgery in patients with osteosarcoma Clin Orthop 1999 120 7 9973983 10.1097/00003086-199901000-00015
Wirbel RJ Schulte M Mutschler WE Surgical treatment of pelvic sarcomas: oncologic and functional outcome Clin Orthop 2001 190 205 11550866
Donati D Giacomini S Gozzi E Ferrari S Sangiorgi L Tienghi A DeGroot H Bertoni F Bacchini P Bacci G Mercuri ME Osteosarcoma of the pelvis Eur J Surg Oncol 2004 30 332 40 15028318 10.1016/j.ejso.2003.12.004
Materson EL Davis AM Wunder JS Bell RS Hindquarter amputation for pelvic tumours: the importance of patient selection Clin Orthop 1998 1 187 194
Enneking WF Dunham WK Resection and reconstruction for primary neoplasms involving the innominate bone J Bone Joint Surg 1978 60A 731 746 701308
Windhager R Karner J Kutschera HP Polterauer P Salzer-Kuntschik M Kotz R Limb Salvage in periacetabular sarcomas: review of 21 consecutive cases Clin Orthop 1996 1 265 276 8895649 10.1097/00003086-199610000-00038
Menendez LR Nichter LS Reconstructive considerations for limb salvage surgery Orthop Clin North Am 1993 24 511 21 8341523
Choong PF Review article: Reconstructive surgery following resection of primary and secondary tumours of the hip J Orthop Surg (Hong Kong) 2000 8 83 94 12468866
Langlais F Lambotte J Thomazeau H Long term results of hemipelvis reconstruction with allografts Clin Orthop 2001 1 178 186 11451117
Bell RS Davis AM Wunder JS Buconjic T McGoveran B Gross AE Allograft reconstruction of the acetabulum after resection of stage-IIB sarcoma. Intermediate-term results J Bone Joint Surg Am 1997 79 1663 74 9384426
Ozaki T Flege S Kevric M Lindner N Maas R Delling G Schwarz R von Hochstetter AR Salzer-Kuntschik M Berdel WE Jurgens H Exner GU Reichardt P Mayer-Steinacker R Ewerbeck V Kotz R Winkelmann W Bielack SS Osteosarcoma of the pelvis: experience of the cooperative osteosracoma study group Journal of clinical oncology 2003 21 334 341 12525527 10.1200/JCO.2003.01.142
Asada N Tsuchiya H Kitaoka K Mori Y Tomita K Massive autoclaved allografts and autografts for limb salvage surgery. A 1–8 year follow-up of 23 patients Acta Orthop Scand 1997 68 392 5 9310047
Husted H Blond L Sonne-Holm S Holm G Jacobsen TW Gebuhr P Tranexamic acid reduces blood loss and blood transfusions in primary total hip arthroplasty: a prospective randomized double-blind study in 40 patients Acta Orthop Scand 2003 74 665 9 14763696 10.1080/00016470310018171
Dalmau A Sabate A Acosta F Garcia-Huete L Koo M Sansano T Rafecas A Figueras J Jaurrieta E Parrilla P Tranexamic acid reduces red cell transfusion better than epsilon-aminocaproic acid or placebo in liver transplantation Anesth Analg 2000 91 29 34 10866882 10.1097/00000539-200007000-00006
Dryden PJ O'Connor JP Jamieson WR Reid I Ansley D Sadeghi H Burr LH Munro AI Merrick PM Tranexamic acid reduces blood loss and transfusion in reoperative cardiac surgery Can J Anaesth 1997 44 934 41 9305556
Benoni G Fredin H Fibrinolytic inhibition with tranexamic acid reduces blood loss and blood transfusion after knee arthroplasty: a prospective, randomised, double-blind study of 86 patients J Bone Joint Surg Br 1996 78 434 40 8636182
| 15819994 | PMC1087869 | CC BY | 2021-01-04 16:38:37 | no | Int Semin Surg Oncol. 2005 Apr 9; 2:9 | utf-8 | Int Semin Surg Oncol | 2,005 | 10.1186/1477-7800-2-9 | oa_comm |
==== Front
J Autoimmune DisJournal of Autoimmune Diseases1740-2557BioMed Central London 1740-2557-2-31585422810.1186/1740-2557-2-3EditorialThe importance of APC Ablamunits Vitaly [email protected] Department of Medicine, St. Luke's-Roosevelt Hospital Center, New York, NY, USA2 Institute for Human Nutrition, Columbia University, New York, NY, USA2005 26 4 2005 2 3 3 12 4 2005 26 4 2005 Copyright © 2005 Ablamunits; licensee BioMed Central Ltd.2005Ablamunits; 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.
Readers in immunology are familiar with the importance of antigen presenting cells in mounting immune responses. For the purpose of this particular editorial article, however, the abbreviation APC will stand for article processing charges. The publisher will introduce APCs for this Journal in May, 2005. Here we explain why article-processing charges are important to maintain our Open Access journal.
==== Body
Journal of Autoimmune Diseases is published by BioMed Central, an independent publisher committed to ensuring peer-reviewed biomedical research is Open Access – universally and freely available online to everyone, with authors retaining copyright, and the full text being archived in numerous internationally recognised free repositories [1]. Journal of Autoimmune Diseases however, has taken this further, by making all its content Open Access. To fund this, authors of articles accepted for publication in the Journal will be asked to pay an article-processing charge (APC). In 2005, that charge will be 330 GBP.
Traditionally, readers pay to access articles, either through subscriptions or by paying a fee each time they download an article. Escalating journal subscription prices have resulted in libraries subscribing to fewer journals [2], and the range of articles available to readers is therefore limited. Although traditional journals publish authors' work for free (unless there are page or colour charges), having to pay to access articles limits how many can read, use and cite them.
Journal of Autoimmune Diseases' Open Access policy changes the way in which articles are published. First, all articles become freely and universally accessible online, and so an author's work can be read by anyone at no cost. Second, the authors hold copyright for their work and grant anyone the right to reproduce and disseminate the article, provided that it is correctly cited and no errors are introduced [1]. Third, a copy of the full text of each article is permanently archived in a number of online repositories separate from the journal. Articles published in Journal of Autoimmune Diseases are archived in PubMed Central [3], the US National Library of Medicine's full-text repository of life science literature, and also in repositories at the University of Potsdam [4] in Germany, at INIST [5] in France and in e-Depot [6], the National Library of the Netherlands' digital archive of all electronic publications.
Who benefits from the Open Access?
Open Access has four broad benefits for science and the general public. First, authors are assured that their work is disseminated to the widest possible audience, given that there are no barriers to access their work. It has been shown that free online articles are more highly cited because of their easier availability [7]. Second, the information available to researchers will not be limited by what their library can afford, and the widespread availability of articles will enhance literature searching [8]. Third, the results of publicly funded research will be accessible to all taxpayers and not just those with access to a library with a subscription. Note that this public accessibility may become a legal requirement in the USA if the proposed Public Access to Science Act is made law [9]. Fourth, a country's economy will not influence its scientists' ability to access articles because resource-poor countries (and institutions) will be able to read the same material as wealthier ones (although creating access to the internet is another matter [10]). APC will allow continued Open Access to all of the articles published in Journal of Autoimmune Diseases.
What does APC cover?
The APC pays for the article to be freely and universally accessible in various formats online, and for the processes required for inclusion in PubMed and archiving in PubMed Central, e-Depot, Potsdam and INIST. Although some authors may consider 330 GBP expensive, it must be remembered that Journal of Autoimmune Diseases does not levy additional page or colour charges on top of this fee, which can easily exceed 330 GBP. With the article being online only, any number of colour figures and photographs can be included, at no extra cost.
How will it work?
Authors will be asked to pay 330 GBP if their article is accepted for publication. Waiver requests will be considered on a case-by-case basis, by the Editor-in-Chief. Authors can circumvent the charge by getting their institution to become a 'member' of BioMed Central, whereby the annual membership fee covers the APC for all authors at that institution for that year. Current members include NHS England, the World Health Organization, the US National Institutes of Health, Harvard, Princeton and Yale universities, and all UK universities [11]. No charge will be made for articles that are rejected after peer review.
By providing a forum for Open Access, APC will enable Journal of Autoimmune Diseases to better serve the scientific community. We believe this change will benefit our readers and aid research in the field of autoimmunity.
==== Refs
BioMed Central Open Access Charter
Tamber PS Is scholarly publishing becoming a monopoly? BMC News and Views 2000 1 1
PubMed Central
Potsdam
INIST
e-Depot
Lawrence S Free online availability substantially increases a paper's impact Nature 2001 411 521 11385534 10.1038/35079151
Velterop J Should scholarly societies embrace Open Access (or is it the kiss of death)? Learned Publishing 2003 16 167 169 10.1087/095315103322110932
Open Access law introduced
Tan-Torres Edejer T Disseminating health information in developing countries: the role of the internet BMJ 2000 321 797 800 11009519 10.1136/bmj.321.7264.797
BioMed Central Institutional Members
| 15854228 | PMC1087870 | CC BY | 2021-01-04 16:38:03 | no | J Autoimmune Dis. 2005 Apr 26; 2:3 | utf-8 | J Autoimmune Dis | 2,005 | 10.1186/1740-2557-2-3 | oa_comm |
==== Front
Lipids Health DisLipids in Health and Disease1476-511XBioMed Central London 1476-511X-4-91583679810.1186/1476-511X-4-9ReviewPotential cellular receptors involved in hepatitis C virus entry into cells Favre Daniel [email protected] Beat [email protected] Division of Gastroenterology and Hepatology, Department of Internal Medicine, University Hospital Zürich, Rämistrasse 100, CH-8091 Zürich, Switzerland2005 19 4 2005 4 9 9 8 4 2005 19 4 2005 Copyright © 2005 Favre and Muellhaupt; licensee BioMed Central Ltd.2005Favre and Muellhaupt; 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.
Hepatitis C virus (HCV) infects hepatocytes and leads to permanent, severe liver damage. Since the genomic sequence of HCV was determined, progress has been made towards understanding the functions of the HCV-encoded proteins and identifying the cellular receptor(s) responsible for adsorption and penetration of the virus particle into the target cells. Several cellular receptors for HCV have been proposed, all of which are associated with lipid and lipoprotein metabolism. This article reviews the cellular receptors for HCV and suggests a general model for HCV entry into cells, in which lipoproteins play a crucial role.
hepatitis C virusHCVlow density lipoproteinsLDLCD81SR-B1exosomesreceptordextran sulfateinfectionlipidsstatins
==== Body
Review
Hepatitis C virus (HCV) and cellular receptors for HCV
Hepatitis C virus (HCV) is a major cause of chronic liver disease, with approximately 170 million people infected worldwide [1]. Infection with HCV can lead to hepatocellular carcinoma [2]. To study the adsorption, penetration and replication of the virus, a major obstacle has been the lack of an efficient and reproducible in vitro infection system. Thus, the identification of the HCV receptor on the surface of susceptible cells, especially hepatocytes, remains a major challenge for the development of both in vitro cell culture systems, and for the design of successful therapies [3,4].
Several cellular receptors have been proposed to mediate the entry of HCV into cells, namely the CD81 receptor [5,6], the scavenger receptor class B type I (SR-BI) receptor [7], and the low density lipoprotein (LDL) receptor [8,9].
CD81 as a receptor for HCV
The tetraspanin CD81 (also named TAPA-1) is a widely-expressed cell surface protein of 26 kDa that is involved in pleiotropic activities such as cell adhesion, motility, metastasis, cell activation and signal transduction [10]. It physically associates with a variety of other membrane proteins such as integrins, lineage-specific molecules and other tetraspanins. It is expressed in most human tissues with the exception of the red blood cells and platelets. Association of CD81 with other molecules has been extensively studied with B and T cells.
It was shown that the expression of CD81 on nonpermissive human, but not murine, hepatic cells enabled the entry of HCV pseudoviruses. The inhibition of viral entry, achieved by application of anti-CD81 monoclonal antibodies, occurred at a step following viral attachment to target cells [11]. When the HCV envelope glycoproteins E1 and E2 were expressed in a baculovirus system, the purified E1-E2 heterodimer interacted with CD81, as well as with the LDL receptor [12]. The human CD81 protein was expressed in bacteria, and the critical amino acids in CD81 involved in the interaction with the viral envelope protein E2 were identified [13]. HIV-HCV pseudotypes bearing native HCV glycoproteins were infectious to the human hepatoma cell lines Huh7 and PLC/PR5. The infectivity was inhibited by recombinant soluble CD81, suggesting that CD81 was a component of the receptor complex [14]. CD81 chimeras, but not wild-type CD81, could internalize recombinant E2 protein and E2-enveloped viral particles from the serum of HCV-infected patients into Huh7 hepatoma cells [15]. Moreover, the expression of CD81 in human liver-derived cells, HepG2 and HH29, that were previously resistant to infection conferred permissiveness to HCV pseudotype infection [16].
In contrast, in several studies, no correlation between CD81 expression and HCV binding has been observed. It has been suggested that HCV binding to hepatocytes might not entirely depend on CD81 [17]. Instead, these authors proposed that CD81 may be an attachment receptor with poor capacity to mediate the viral entry, and that reducing environments may not not favor CD81-HCV interaction. Indeed, it was shown that the binding of E2 to CD81 was not predictive of an infection-producing interaction between HCV and host cells [18]. Moreover, the binding of HCV-like particles was CD81-independent and did not correlate with the expression of the LDL receptor [19]. Finally, human CD81 transgenic mice that lacked expression of the endogenous mouse CD81 were resistant to HCV infection [20]. These authors concluded that the expression of human CD81 alone was not sufficient to confer susceptibility to HCV infection in the mouse.
SR-BI as a receptor for HCV
Scavenger receptors are cell membrane proteins that bind chemically modified lipoproteins, such as acetylated and oxidized LDLs. These receptors have been categorized into broad classes (A, B, C, D, etc), according to their structures [21]. The SR-BI receptor is involved in bidirectional cholesterol transport at the cell membrane and can bind both high density lipoproteins (HDL) and low density lipoproteins [22]. The cholesterol uptake is different from the classic LDL receptor-mediated endocytosis pathway, since it appears to involve initial transfer to the plasma membrane [23]. SR-BI is highly expressed in hepatocytes [24], and is located in the cholesterol-rich lipid raft membrane compartment [25]. The HCV E2 protein could bind to hepatoma cell lines independently of the CD81 receptor. SR-BI was identified as a mediator of this binding. This interaction was selective, since neither the mouse SR-BI nor the closely related human scavenger receptor CD36, were able to bind E2. The E2 recognition by SR-BI was competitively inhibited, in an isolate-specific manner, by a monoclonal antibody raised against the hypervariable region 1 (HVR1, a 27 amino acid segment located at the N-terminus of the E2 poplypeptide) [7].
LDL receptor as a receptor for HCV
The LDL receptor is an endocytic receptor that transports lipoproteins, mainly the cholesterol-rich lipoprotein LDL, into cells through receptor-mediated endocytosis [26,27]. This process involves the cell surface receptor recognizing an LDL particle, followed by its internalization through clathrin-coated pits [28,29]. It has been suggested that HCV might enter the cells via the LDL receptor [8,9]. The binding of low density HCV particles correlated with the extent of the LDL receptor at the cell surface, but not soluble CD81 [30]. In contrast, the binding of HCV-like particles did not correlate with the LDL receptor expression, but was CD81-independent. These hepatoma and lymphoma cells were directly incubated with the virus-like particles [19], without previous removal of the cell-bound lipoproteins. Moreover, free beta-lipoproteins in human serum may influence the rate of infection of hepatocytes by competing with the virus. In support of this, it has been shown that the LDL receptor can function as a HCV receptor and that beta-lipoproteins competitively inhibit the infection of hepatocytes with HCV through the LDL receptor [31]. Indeed, it has been suggested that the removal of the cell-bound lipoproteins is a crucial prerequisite for the infection of hepatocytes with HCV [32]. In the latter study, the viral inoculum that was employed was composed of a virus-lipoprotein complex, as described elsewhere [33,34].
HCV and exosomes
Exosomes are small membrane vesicles secreted by cells upon fusion of multivesicular endosomes with the cell surface [35,36]. They are 60 to 100 nm in diameter and originate from late endosomes. Exosomes are secreted from most hematopoietic and epithelial cells in vitro. Intracellularly, they are formed by inward budding of the endosomal membrane in a process that sequesters particular proteins and lipids [37]. The unique composition of exosomes may confer specific functions to them upon secretion.
Recently, it has been shown that the HCV envelope proteins were associated with exosomes [38]. In the absence of the human CD81, HCV envelope proteins were almost completely retained in the endoplasmic reticulum of hamster CHO cells. Instead, when the human CD81 was present, a fraction of the HCV envelope proteins passed through the Golgi apparatus, matured acquiring complex sugars and was found extracellularly associated with exosomes. It was proposed that the HCV-CD81 complex exits the cells in the form of exosomes, circulates in the blood as a complex and exploits the fusogenic capabilities of these vesicles to infect cells even in the presence of neutralizing antibodies. Therefore, the human CD81 may in fact act as an exit receptor for HCV. The authors concluded that a fraction of HCV RNA was bound to CD81 in patients infected with HCV, because it was difficult to estimate the exact fraction of HCV RNA in human plasma that was associated with exosomes. Differential centrifugation was employed for the purification of the exosomes, so that the buoyant density of these vesicles was not measured. This measure would be important to discriminate between free and bound HCV [38].
Lipid raft-associated protein sorting has been involved with exosomes. Some molecules are released in the extracellular medium via their association with lipid raft domains of the exosomal membrane. The presence of lipid microdomains in exosomal membranes is suggesting their participation in vesicle formation and structure, as well as the direct implication of exosomes in regulatory mechanisms [39]. Buoyant density is the quality for a compound to rise or float in a liquid. The measure of this density can be employed for the discrimination of exosomes, for example. Exosomes float to a density close to 1.13 g/ml (as revealed by ultracentrifugation), but this may vary from cell to cell depending of the exosome protein content [37,40]. In several studies, the buoyant density of exosomes originating from B lymphocytes has been determined to be in the range 1.08–1,22 g/ml with a peak at 1.13–1.15 g/ml [41,42].
Interestingly, hepatitis C virus is structurally migrating with buoyant densities lying in the same range as those determined for low density lipoproteins and exosomes. It was shown that HCV was associated with beta-lipoproteins having buoyant densities between 1.03 g/ml and 1.20 g/ml in the human serum [33,34]. Moreover, action of lipoprotein lipase on hepatitis C virus in human sera was shown to be virolytic [43]. In order to analyze the potential HCV-lipoprotein complex, the binding of sucrose gradient-purified low-density particles (1.03 to 1.07 g/cm3), intermediate-density particles (1.12 to 1.18 g/cm3), recombinant E2 protein, or control proteins to MOLT-4 cells, foreskin fibroblasts, or LDL receptor-deficient foreskin fibroblasts, was assessed. This revealed that the low-density HCV particles, but not intermediate-density HCV or controls, bound to MOLT-4 cells and fibroblasts expressing the LDL receptor. Binding correlated with the extent of cellular LDL receptor expression and was inhibited by LDL but not by soluble CD81. In contrast, E2 binding was independent of LDL receptor expression and was inhibited by human soluble CD81 but not mouse soluble CD81 or LDL [30].
The complete complementary DNA of an isolate of the hepatitis C virus was cloned into a tetracycline-inducible expression vector and stably transfected into the human hepatoma cell lines Huh7 and HepG2. Viral RNA levels peaked at two separate ranges, one with a buoyant density of 1.08 g/ml and another from 1.17 to 1.39 g/ml. Anti-E2 antibodies strongly labelled cytoplasmic vesicular structures and some viral-like particles. Complete viral particles of about 50 nm which reacted with anti-E2 antibodies were observed in the culture media of tet-induced clonal HuH-7 cells following negative staining [44].
A general scheme for the adsorption and penetration of HCV onto cells is emerging
It is tempting to speculate that HCV bound to, or contained within, low density lipoproteins [33,34], viro-lipo-particles [45], or exosomes [38] would be constituted as one similar structure that would allow the virus to adsorb and penetrate into the target cells. This structure would in general migrate in sucrose gradients at a density similar to low density lipoproteins. A common feature of these three different structures is that they all contain lipids such as sphingomyelin, cholesterol, glycolipids and lipids that are critical for the maintenance of lipid rafts [37,40]. Consistent with the initial buoyant density measures [33,34], the lipoproteins bound to HCV might therefore be similar to those present in exosomes [38]. Since hepatocytes are from hematopoïetic origin [46], they are producing exosomes [47]. Moreover, since CD81 is enriched in exosomes [48], it may not be a receptor mediating the HCV entry into hepatocytes, but rather an exit receptor. Although SR-BI is highly expressed in hepatocytes [22,25] and is a receptor for HDL and LDL [22,49], hepatocyte cell lines such as HepG2 do not express CD81 [7]. Therefore, the afforementioned compilation of observations might argue against the role played by CD81 as the cellular receptor for HCV. Therefore, since low density lipoproteins and exosomes do bind the LDL receptor [50], HCV might enter the cells via the LDL receptor solely. Work is currently under progress in order to analyze the uptake of the hepatitis C virus-lipoprotein in HepG2 and HuH-7 cells in absence or presence of N4-octadecyl-1-β-D-arabinofuranosylcytosine, as described elsewhere [51].
Conclusion
In vitro and in vivo implications
It might be envisioned that one of the crucial parameter for viral entry would therefore be the LDL receptor at the cell surface of the hepatocytes. If this is true, the CD81 tetraspanin and/or SR-B1 would rather regulate virus adhesion and/or fusion with target cells than playing the role of a cellular receptor (Fig. 1). Indeed, it has been proposed that in order to infect hepatocyte cell lines with HCV in vitro, the cell-bound lipoproteins have to be removed with dextran sulfate prior to the addition of the viral inoculum onto the cells [32]. According to this model, the binding of the HCV-lipoprotein complex to the LDL receptor might be hampered in vitro by the cell-bound lipoproteins, or by the vast excess of free lipoproteins over virus-bound lipoproteins in the human blood. The similar procedure of infection would therefore also apply for viro-lipo-particles and exosomes.
Figure 1 A model for HCV infection. A. HCV is present as a virus-lipoprotein complex in the human blood. HCV can be surrounded by several low density lipoproteins, so that the viral envelope proteins E1 and E2 might be masked. B. HCV is present within exosomes. It was recently shown that these exosomes are containing, apart from the genomic ribonucleic acid of HCV, the tetraspanin CD81 associated with E1 and E2. It is likely that the binding of the HCV-lipoprotein and/or the exosome-HCV complexes to the low density lipoprotein receptor (LDLr) might therefore be hampered in vitro by the cell-bound lipoproteins, or by the vast excess of free lipoproteins in the human blood. Therefore, it may be beneficial to remove the cell-bound lipoproteins with dextran sulfate (thus generating free LDL receptors) prior to the addition of the viral inoculum onto the target hepatocytes for the generation of an in vitro infection. The same may be true for the scavenger receptor SR-BI, since it does also bind LDL. In vivo, the use of statins may enhance the rate of HCV infection in HCV-infected patients, because of the increase of the LDL receptors at the surface of the hepatocytes.
The lipidemic status in HCV-infected individuals might also play a critical role for the onset and the maintainance of a robust immune response directed against HCV, especially in patients suffering from hypercholesterolemia, coronary artery disease, diabetes mellitus or obesity. Cholesterol lowering drugs with statins are abundantly employed for the lipid management in hyperlipidemic patients [52,53]. The primary effect of statins is the induction of the expression of LDL receptors on the surface of the hepatocytes [54,55]. Safety and tolerability profiles are available and statins have become the drugs of choice when diet alone has failed [56]. However, there are no controlled trials published that might reveal the link between the cholesterol management with statins and the efficiency of the hepatitis C virus replication, yet. It remains obviously to be shown, whether the recovery from hyperlipidemia to normo- or hypolipidemia is indeed not dramatically allowing the infection or the re-infection of hepatocytes with HCV in the human liver.
Authors' contributions
DF wrote the initial draft of the article and created the figure 1. BM critically revised the article until its final version. Both authors read and approved the final manuscript.
Acknowledgements
This work was supported in part by the Naef Foundation for In Vitro Research (Geneva, Switzerland), the Kurt und Senta Hermann Stiftung (Zürich) and the Rentenanstalt-Swiss Life Research Fund (Zürich). We thank Prof. Reto Schwendener and Dr. Jyrki Eloranta for helpful comments and suggestions.
==== Refs
Lauer GM Walker BD Hepatitis C virus infection N Engl J Med 2001 345 41 52 11439948 10.1056/NEJM200107053450107
Koike K Tsutsumi T Fujie H Shintani Y Kyoji M Molecular mechanism of viral hepatocarcinogenesis Oncology 2002 62 29 37 11868782 10.1159/000048273
Moriishi K Matsuura Y Mechanisms of hepatitis C virus infection Antivir Chem Chemother 2003 14 285 297 14968935
Nishioka T Chayama K [Mechanism of HCV cell entry mediated by envelope and receptor proteins] Nippon Rinsho 2004 62 Suppl 7 195 198 15359792
Flint M Maidens C Loomis-Price LD Shotton C Dubuisson J Monk P Higginbottom A Levy S McKeating JA Characterization of hepatitis C virus E2 glycoprotein interaction with a putative cellular receptor, CD81 J Virol 1999 73 6235 6244 10400713
Pileri P Uematsu Y Campagnoli S Galli G Falugi F Petracca R Weiner AJ Houghton M Rosa D Grandi G Binding of hepatitis C virus to CD81 Science 1998 282 938 941 9794763 10.1126/science.282.5390.938
Scarselli E Ansuini H Cerino R Roccasecca RM Acali S Filocamo G Traboni C Nicosia A Cortese R Vitelli A The human scavenger receptor class B type I is a novel candidate receptor for the hepatitis C virus Embo J 2002 21 5017 5025 12356718 10.1093/emboj/cdf529
Monazahian M Bohme I Bonk S Koch A Scholz C Grethe S Thomssen R Low density lipoprotein receptor as a candidate receptor for hepatitis C virus J Med Virol 1999 57 223 229 10022791 10.1002/(SICI)1096-9071(199903)57:3<223::AID-JMV2>3.0.CO;2-4
Agnello V Abel G Elfahal M Knight GB Zhang QX Hepatitis C virus and other flaviviridae viruses enter cells via low density lipoprotein receptor Proc Natl Acad Sci U S A 1999 96 12766 12771 10535997 10.1073/pnas.96.22.12766
Levy S Todd SC Maecker HT CD81 (TAPA-1): a molecule involved in signal transduction and cell adhesion in the immune system Annu Rev Immunol 1998 16 89 109 9597125 10.1146/annurev.immunol.16.1.89
Cormier EG Tsamis F Kajumo F Durso RJ Gardner JP Dragic T CD81 is an entry coreceptor for hepatitis C virus Proc Natl Acad Sci U S A 2004 101 7270 7274 15123813 10.1073/pnas.0402253101
Lambot M Fretier S Op De Beeck A Quatannens B Lestavel S Clavey V Dubuisson J Reconstitution of hepatitis C virus envelope glycoproteins into liposomes as a surrogate model to study virus attachment J Biol Chem 2002 277 20625 20630 11937498 10.1074/jbc.M111020200
Higginbottom A Quinn ER Kuo CC Flint M Wilson LH Bianchi E Nicosia A Monk PN McKeating JA Levy S Identification of amino acid residues in CD81 critical for interaction with hepatitis C virus envelope glycoprotein E2 J Virol 2000 74 3642 3649 10729140 10.1128/JVI.74.8.3642-3649.2000
Hsu M Zhang J Flint M Logvinoff C Cheng-Mayer C Rice CM McKeating JA Hepatitis C virus glycoproteins mediate pH-dependent cell entry of pseudotyped retroviral particles Proc Natl Acad Sci U S A 2003 100 7271 7276 12761383 10.1073/pnas.0832180100
Tan YJ Lim SP Ng P Goh PY Lim SG Tan YH Hong W CD81 engineered with endocytotic signals mediates HCV cell entry: implications for receptor usage by HCV in vivo Virology 2003 308 250 269 12706076 10.1016/S0042-6822(02)00136-8
Zhang J Randall G Higginbottom A Monk P Rice CM McKeating JA CD81 is required for hepatitis C virus glycoprotein-mediated viral infection J Virol 2004 78 1448 1455 14722300 10.1128/JVI.78.3.1448-1455.2004
Petracca R Falugi F Galli G Norais N Rosa D Campagnoli S Burgio V Di Stasio E Giardina B Houghton M Structure-function analysis of hepatitis C virus envelope-CD81 binding J Virol 2000 74 4824 4830 10775621 10.1128/JVI.74.10.4824-4830.2000
Meola A Sbardellati A Bruni Ercole B Cerretani M Pezzanera M Ceccacci A Vitelli A Levy S Nicosia A Traboni C Binding of hepatitis C virus E2 glycoprotein to CD81 does not correlate with species permissiveness to infection J Virol 2000 74 5933 5938 10846074 10.1128/JVI.74.13.5933-5938.2000
Wellnitz S Klumpp B Barth H Ito S Depla E Dubuisson J Blum HE Baumert TF Binding of hepatitis C virus-like particles derived from infectious clone H77C to defined human cell lines J Virol 2002 76 1181 1193 11773394
Masciopinto F Freer G Burgio VL Levy S Galli-Stampino L Bendinelli M Houghton M Abrignani S Uematsu Y Expression of human CD81 in transgenic mice does not confer susceptibility to hepatitis C virus infection Virology 2002 304 187 196 12504561 10.1006/viro.2002.1631
Krieger M Stern DM Series introduction: multiligand receptors and human disease J Clin Invest 2001 108 645 647 11544266 10.1172/JCI200113932
Krieger M Scavenger receptor class B type I is a multiligand HDL receptor that influences diverse physiologic systems J Clin Invest 2001 108 793 797 11560945 10.1172/JCI200114011
Acton S Rigotti A Landschulz KT Xu S Hobbs HH Krieger M Identification of scavenger receptor SR-BI as a high density lipoprotein receptor Science 1996 271 518 520 8560269
Calvo D Gomez-Coronado D Lasuncion MA Vega MA CLA-1 is an 85-kD plasma membrane glycoprotein that acts as a high-affinity receptor for both native (HDL, LDL, and VLDL) and modified (OxLDL and AcLDL) lipoproteins Arterioscler Thromb Vasc Biol 1997 17 2341 2349 9409200
Calvo M Enrich C Biochemical analysis of a caveolae-enriched plasma membrane fraction from rat liver Electrophoresis 2000 21 3386 3395 11079559 10.1002/1522-2683(20001001)21:16<3386::AID-ELPS3386>3.3.CO;2-C
Chung NS Wasan KM Potential role of the low-density lipoprotein receptor family as mediators of cellular drug uptake Adv Drug Deliv Rev 2004 56 1315 1334 15109771 10.1016/j.addr.2003.12.003
Nykjaer A Willnow TE The low-density lipoprotein receptor gene family: a cellular Swiss army knife? Trends Cell Biol 2002 12 273 280 12074887 10.1016/S0962-8924(02)02282-1
Steer CJ Zakim D, Boyer TD Receptor-mediated endocytosis: mechanisms, biologic function, and molecular properties Hepatology: a textbook of liver disease 1996 1 W.B. Saunders Company 149 214
Cooper AD Ellsworth JL Zakim D, Boyer TD Lipoprotein metabolism Hepatology: a textbook of liver disease 1996 1 W. B. Saunders Company 92 130
Wunschmann S Medh JD Klinzmann D Schmidt WN Stapleton JT Characterization of hepatitis C virus (HCV) and HCV E2 interactions with CD81 and the low-density lipoprotein receptor J Virol 2000 74 10055 10062 11024134 10.1128/JVI.74.21.10055-10062.2000
Enjoji M Nakamuta M Kinukawa N Sugimoto R Noguchi K Tsuruta S Iwao M Kotoh K Iwamoto H Nawata H Beta-lipoproteins influence the serum level of hepatitis C virus Medical Science Monitor 2000 6 841 844 11208419
Favre D Berthillon P Trepo C Removal of cell-bound lipoproteins: a crucial step for the efficient infection of liver cells with hepatitis C virus in vitro C R Acad Sci III 2001 324 1141 1148 11803815
Thomssen R Bonk S Propfe C Heermann KH Kochel HG Uy A Association of hepatitis C virus in human sera with beta-lipoprotein Med Microbiol Immunol (Berl) 1992 181 293 300 1335546 10.1007/BF00198849
Thomssen R Bonk S Thiele A Density heterogeneities of hepatitis C virus in human sera due to the binding of beta-lipoproteins and immunoglobulins Med Microbiol Immunol (Berl) 1993 182 329 334 8121333
Farsad K Exosomes: novel organelles implicated in immunomodulation and apoptosis Yale J Biol Med 2002 75 95 101 12230314
Stahl PD Barbieri MA Multivesicular bodies and multivesicular endosomes: the "ins and outs" of endosomal traffic Sci STKE 2002 2002 PE32 12122203
Stoorvogel W Kleijmeer MJ Geuze HJ Raposo G The biogenesis and functions of exosomes Traffic 2002 3 321 330 11967126 10.1034/j.1600-0854.2002.30502.x
Masciopinto F Giovani C Campagnoli S Galli-Stampino L Colombatto P Brunetto M Yen TS Houghton M Pileri P Abrignani S Association of hepatitis C virus envelope proteins with exosomes Eur J Immunol 2004 34 2834 15368299 10.1002/eji.200424887
de Gassart A Geminard C Fevrier B Raposo G Vidal M Lipid raft-associated protein sorting in exosomes Blood 2003 102 4336 4344 12881314 10.1182/blood-2003-03-0871
Fevrier B Raposo G Exosomes: endosomal-derived vesicles shipping extracellular messages Curr Opin Cell Biol 2004 16 415 421 15261674 10.1016/j.ceb.2004.06.003
Raposo G Nijman HW Stoorvogel W Liejendekker R Harding CV Melief CJ Geuze HJ B lymphocytes secrete antigen-presenting vesicles J Exp Med 1996 183 1161 1172 8642258 10.1084/jem.183.3.1161
Wubbolts R Leckie RS Veenhuizen PT Schwarzmann G Mobius W Hoernschemeyer J Slot JW Geuze HJ Stoorvogel W Proteomic and biochemical analyses of human B cell-derived exosomes. Potential implications for their function and multivesicular body formation J Biol Chem 2003 278 10963 10972 12519789 10.1074/jbc.M207550200
Thomssen R Bonk S Virolytic action of lipoprotein lipase on hepatitis C virus in human sera Med Microbiol Immunol (Berl) 2002 191 17 24 12137195 10.1007/s00430-001-0106-x
Lim SP Soo HM Tan YH Brenner S Horstmann H MacKenzie JM Ng ML Lim SG Hong W Inducible system in human hepatoma cell lines for hepatitis C virus production Virology 2002 303 79 99 12482660 10.1006/viro.2002.1687
Andre P Komurian-Pradel F Deforges S Perret M Berland JL Sodoyer M Pol S Brechot C Paranhos-Baccala G Lotteau V Characterization of low- and very-low-density hepatitis C virus RNA-containing particles J Virol 2002 76 6919 6928 12072493 10.1128/JVI.76.14.6919-6928.2002
Lagasse E Connors H Al-Dhalimy M Reitsma M Dohse M Osborne L Wang X Finegold M Weissman IL Grompe M Purified hematopoietic stem cells can differentiate into hepatocytes in vivo Nat Med 2000 6 1229 1234 11062533 10.1038/81326
Denzer K Kleijmeer MJ Heijnen HF Stoorvogel W Geuze HJ Exosome: from internal vesicle of the multivesicular body to intercellular signaling device J Cell Sci 2000 113 3365 3374 10984428
Escola JM Kleijmeer MJ Stoorvogel W Griffith JM Yoshie O Geuze HJ Selective enrichment of tetraspan proteins on the internal vesicles of multivesicular endosomes and on exosomes secreted by human B-lymphocytes J Biol Chem 1998 273 20121 20127 9685355 10.1074/jbc.273.32.20121
Krieger M Charting the fate of the "good cholesterol": identification and characterization of the high-density lipoprotein receptor SR-BI Annu Rev Biochem 1999 68 523 558 10872459 10.1146/annurev.biochem.68.1.523
Laulagnier K Motta C Hamdi S Roy S Fauvelle F Pageaux JF Kobayashi T Salles JP Perret B Bonnerot C Mast cell- and dendritic cell-derived exosomes display a specific lipid composition and an unusual membrane organization Biochem J 2004 380 161 171 14965343 10.1042/BJ20031594
Koller-Lucae SK Schott H Schwendener RA Low density lipoprotein and liposome mediated uptake and cytotoxic effect of N4-octadecyl-1-beta-D-arabinofuranosylcytosine in Daudi lymphoma cells Br J Cancer 1999 80 1542 1549 10408395 10.1038/sj.bjc.6690558
McKenney JM Lipid management: tools for getting to the goal Am J Manag Care 2001 7 S299 306 11517816
Stein EA Management of hypercholesterolemia. Approach to diet and drug therapy Am J Med 1989 87 20S 27S 2679081 10.1016/S0002-9343(89)80595-9
Ginsberg HN Effects of statins on triglyceride metabolism Am J Cardiol 1998 81 32B 35B 9462602 10.1016/S0002-9149(98)00035-6
Duriez P [Mechanisms of actions of statins and fibrates] Therapie 2003 58 5 14 12822195
Jenkins DJ Kendall CW Marchie A Faulkner D Vidgen E Lapsley KG Trautwein EA Parker TL Josse RG Leiter LA The effect of combining plant sterols, soy protein, viscous fibers, and almonds in treating hypercholesterolemia Metabolism 2003 52 1478 1483 14624410 10.1016/S0026-0495(03)00260-9
| 15836798 | PMC1087871 | CC BY | 2021-01-04 16:39:19 | no | Lipids Health Dis. 2005 Apr 19; 4:9 | utf-8 | Lipids Health Dis | 2,005 | 10.1186/1476-511X-4-9 | oa_comm |
==== Front
Malar JMalaria Journal1475-2875BioMed Central London 1475-2875-4-181582900810.1186/1475-2875-4-18ResearchPrevalence and risk factors for Plasmodium falciparum malaria in pregnant women of eastern Sudan Adam Ishag [email protected] Amar H [email protected] Mustafa I [email protected] New Haifa Teaching Hospital, P. O. Box 61, New Haifa, Sudan2 Albayan College for Science, Sudan University for Science and Technology, Sudan3 Department of Biochemistry, Faculty of Medicine, University of Khartoum, Sudan2005 13 4 2005 4 18 18 15 1 2005 13 4 2005 Copyright © 2005 Adam et al; licensee BioMed Central Ltd.2005Adam 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
Pregnant women are more susceptible to malaria, which is associated with serious adverse effects on pregnancy. The presentation of malaria during pregnancy varies according to the level of transmission in the area. Our study aimed to demonstrate the prevalence and risk factors for malaria (age, parity and gestational age) among pregnant women of eastern Sudan, which is characterized by unstable malaria transmission.
Methods
The prevalence and possible risk factors for Plasmodium falciparum malaria were investigated in 744 pregnant Sudanese women attending the antenatal clinic of New Haifa Teaching Hospital, eastern Sudan, during October 2003-April 2004.
Results
A total 102 (13.7%) had P. falciparum malaria, 18(17.6%) of these were severe cases (jaundice and severe anaemia). Univariate and multivariate analysis showed that, age and parity were not associated with malaria. Women who attended the antenatal clinic in the third trimester were at highest risk for malaria (OR = 1.58, 95% CI = 1.02–2.4; P < 0.05).
Women with malaria had significantly lower mean haemoglobin (9.4 g/dl, 95% CI 9.1–9.7 versus 10.7, CI 10.6–10.8, P < 0.05). A significantly lower haemoglobin was observed in those with severe falciparum malaria compared to non-severe form (8.3 g/dl, 95% CI 7.6–9.1 versus 9.4, 95% CI 9.1–9.7, P = < 0.05).
Conclusion
The results suggest that P. falciparum malaria is common in pregnant women attending antenatal care and that anaemia is an important complication. Preventive measures (chemoprophylaxis and insecticide-treated bednets) may be beneficial in this area for all women irrespective of age or parity.
==== Body
Background
Pregnant women are more susceptible to malaria, which causes serious adverse effects including abortion, low birth weight and maternal anaemia. It is the leading cause maternal mortality in Sudan [1-7].
The presentation of malaria during pregnancy varies according to the pre-existing immunity of the mother. Women living in areas of low transmission have little immunity to malaria which can cause severe syndromes, such as cerebral malaria and pulmonary oedema. In contrast, those who live in areas of stable malaria transmission enjoy greater immunity and experience fewer symptoms during episodes of malaria, although they commonly develop severe anaemia as consequence of the infection [1,2,5,8,9].
Understanding the epidemiology of malaria during pregnancy provides important insight into relevant immunological processes and facilitates decision on control strategies. Although there are extensive studies in highly endemic African countries [1,2,5,6,10-12], there is little published data available for Sudan, an African country where P. falciparum malaria transmission is unstable in the eastern region [13]. The study was conducted to investigate the prevalence and associated risk factors for P. falciparum malaria in pregnant women from eastern Sudan.
Patients and methods
Data collection
Pregnant women attending antenatal clinic (booking visit) of the New Haifa Teaching Hospital, eastern Sudan, during the period October 2003-May 2004 were approached to participate in the study. After verbal consent, questionnaires were administered requesting demographic informations on age, parity, gestational age and history of maternal illness. Gestational age was calculated from the last menstrual period and confirmed by ultrasound, when clinically indicated. Physical examination was completed to identify signs of severe malaria [14] and also obstetrical examination (blood pressure, pallor, fundal level and foetal heart sound).
Laboratory methods
Thick and thin blood films were prepared from capillary blood, stained with Giemsa and 100 oil immersion fields were examined. Parasite density was determined by counting parasites and 200 leucocytes, assuming each woman has 6,000 leucocytes/μl. All the slides were double-checked blindly.
The haemoglobin concentration was estimated by the haematic acid method [15] in the first two months and, subsequently, using Haemocure haemoglobinometer (HemCue AB, Angelhom, Sweden). Ferrous sulfate (200 mg/day) and folic acid tablets (0.25 mg/day) were supplied.
Ethical clearance
The study received ethical clearance from the Research Board of the Faculty of Medicine, University of Khartoum.
Statistical analysis
Data were entered in a computer using SPSS for windows. Comparisons between means and percentages were done by Students' t-test, ANOVA, X2 and Fisher's exact tests as appropriate. P < 0.05 was regarded as significant. Multivariate logistic regression was performed with malaria as the dependent variable, using age, parity and gestation as independent variables. These variables were categorized and used as follows: median age, ≤ 25 years versus >25 years); gravidae as primigravidae, secundigravidae, multigravidae (3–5)or grandmultigravidae >5; gestation as first (< 14 weeks), second (14–28 weeks) and third (> 28 weeks) trimester.
Results
Malaria and pregnancy
A total of 744 pregnant women attended the antenatal clinic of New Haifa Teaching Hospital during the study, 29.5% were primigravidae.
Table 1 summarizes patient characteristics for all sub-groups. 102 (13.7%) of women were infected with P. falciparum, 18(17.6%) of these were severe cases (jaundice and severe anaemia). Malaria prevalence and intensity (parasite count) were not significantly different amongst the different gravidity sub-groups (P > 0.05). The highest prevalence (18.3%) occurred in grandmultigravidae and the highest intensity (11,511 parasites/μl) was observed in primigravidae (Table 1).
Table 1 Characteristics of pregnant women in the study subjects *
Characteristics Total Primi-gravidae Secundi-gravidae Multi-gravidae Grand-multigravidae P Value
N (744) N (220) N (131) N (322) N (71)
Age (years) 26.6(9.9) 21.2(2.9) 25.7(2.3) 28.2(3.8) 34.9 (3.2) < 0.001
Gestation (weeks) 27.7 (7.7) 27.3(8.09) 28.1(6.9) 27.6(7.6) 28.3(28.3) 0.18
Haemoglobin (g/dl) 10.5(1.4) 10.5(1.3) 10.3(1.6) 10.6(1.3) 10.3(1.2) 0.2
Malaria (total) 102(13.7) 32(14.5) 14(10.7) 43(13.4) 13(18.3) 0.4
Non severe 84(11.3) 28(12.7) 9(6.9) 34(10.6) 13 (18.3) 0.08
Severe 18(2.4) 4(1.8) 5(3.8) 9 (2.8) 0 0.3
Parasite count (parasites/μl) 6,586 (19,640.2) 11,511 (3,222) 6,817 (14,383) 3,504 (7,729) 4,497 (7,748) 0.4
*Data as mean (SD) or number (%) as appropriate
Mean (SD) age (26.2 ± 5.7 years versus 25.9 ± 5.3 years, P > 0.05) and parity (2.4 ± 2.4 versus 2.1 ± 2.1, P > 0.05) were not significantly different between infected and non-infected women. Two of 18 (11.1%), 29 of 290 (10%) and 71 of 436 (16.3%) women were infected in the first, second and third trimester respectively (P = 0.05).
Factors associated with malaria infection
Univariate and multivariate analysis indicated that, age and parity were not significantly associated with malaria infection. The third trimester was significantly associated with malaria infection (OR = 1.58, 95% CI = 1.02–2.4; P < 0.05)(Table 2).
Table 2 Risk factor analysis
Univariate Multivariate
Variable OR 95% CI P OR 95% CI P
Age ≤ 25 1.03 0.68–1.5 0.40 0.90 0.52–1.6 0.74
Parity
Primigravidae 1.10 0.70–1.7 0.60 0.76 0.32–1.8 0.55
Secundigravidae 0.72 0.39–1.3 0.20 0.50 0.21–1.4 0.22
Multigravidae 0.96 0.62–1.4 0.80 0.74 0.37–1.4 0.41
Grandmultigravidae 1.40 0.77–2.7 0.20 3.20 0.27–46.2 0.30
Gestational age
First trimester 0.61 0.27–1.3 0.15 0.51 0.22–1.1 0.11
Second Trimester 0.71 0.45–1.3 0.09 0.67 0.41–1.08 0.10
Third trimester 1.50 1.02–2.4 0.02 1.58 1.02–2.4 0.03
OR: odds ratio
Malaria and haemoglobin
Mean haemoglobin was significantly lower in infected women (9.4 g/dl, 95% CI 9.1–9.7versus 10.7, CI 10.6–10.8, P < 0.05). A lower haemoglobin also occurred in women with severe malaria (9.4 g/dl, 95% CI 9.1–9.7versus 8.3, 95% CI 7.6–9.1 P < 0.05).
Discussion
This study investigated the morbidity pattern of P. falciparum malaria during pregnancy in an area of New Haifa, which is characterized by unstable transmission [13]. The malaria prevalence was 13.7% and 18 women (17.6%) were severe cases. With the exception of the neighbouring country Ethiopia [6], malaria prevalence was much lower than in other African countries characterized by intense malaria transmission [10-12]. Several severe cases occured. In areas of unstable transmission, pregnant women are more susceptible to severe falciparum malaria than their non-pregnant peers [16] and different manifestations of severe falciparum malaria including cerebral malaria have been reported among pregnant women of central Sudan [17].
The study showed no significant association between malaria and parity, which is also reported from other areas and locations with intense malaria transmission [10,12,18]. In areas where transmission is high and the level of acquired pregnancy immunity against malaria is expected to be significant, primigravidae will be more affected [1,2,11].
In contrast to the previous observations [10-12], age was not significantly associated with malaria in the present study. Lander et al have also reported no significant association between malaria infection and maternal age [18].
Women who attended the antenatal clinic in the third trimester had about a 1.5-fold higher risk of malaria parasitaemia. This is in line with some observations, although several studies report the second and early third trimesters as the time of peak prevalence [1,3,4,10-12]. Dicko et al reported the first trimester as the main risk period [19].
The mean haemoglobin was significantly lower in women with malaria infection and significantly lower in severe cases. Reduction of haemoglobin has been reported in areas of unstable transmission in Thailand and in Ethiopia, as well as in areas with stable transmission [3,6]. Regardless of transmission level and pre-pregnancy level of malaria immunity, maternal anaemia remains the most frequent consequence of malaria during pregnancy [4].
Conclusion
The results suggest that prevalence of P. falciparum malaria is considerable in pregnant women in this part of Sudan and severe cases do occur. Preventive measures (chemoprophylaxis and insecticide-treated bednets) may be beneficial in this area for all women irrespective of their age or parity.
Authors' contributions
IA carried out the study and participated in the statistical analysis and procedures, AHK participated in the statistical analysis, MIE coordinated and participated in the design of the study, statistical analysis and the drafting of the manuscript. All the authors read and approved the final version.
Acknowledgements
We wish to thank all the patients for their excellent cooperation and we are very grateful to the local health authority in Kassala State and to the entire staff of New Haifa Teaching Hospital. Thanks are also extended to Mr. A. A. Hufazalla for his excellent technical assistance.
==== Refs
Brabin BJ An analysis of malaria in pregnancy in Africa Bull World Health Organ 1983 61 1005 1016 6370484
McGregor IA Epidemiology, malaria and pregnancy Am J Trop Med Hyg 1984 33 517 525 6383091
Nosten F Ter Kuile FO Maelankirri L Decludt B White NJ Malaria during pregnancy in an area of unstable endemicity Trans R Soc Trop Med Hyg 1991 85 424 429 1836685 10.1016/0035-9203(91)90205-D
Menendez C Malaria during pregnancy: A priority area of malaria research and control Parasitol Today 1995 11 178 183 15275350 10.1016/0169-4758(95)80151-0
Dafallah SE EL-Agib FH Bushra GO Maternal mortality in a teaching hospital in Sudan Saudi Med J 2003 24 369 373 12754536
Newman RD Hailmariam A Jimma D Degefie A Kebede D Rietveld AC Nahlen BL Barnwell JW Steketee RW Parise ME Burden of malaria during pregnancy in areas of stable and unstable transmission in Ethiopia during a nonepidemic year J Infect Dis 2003 187 1765 1772 12751034 10.1086/374878
Adam I Elbashir MI Maternal death due to servere pulmonary edema caused by falciparum malaria: case report East Mediterr Health J 2004
Okoko BJ Enwere G Ota MOC The epidemiology and consequences of maternal malaria: a review of immunological basis Acta Trop 2003 87 193 205 12826295 10.1016/S0001-706X(03)00097-4
Shulman CE Dorman EK Importance and prevention for malaria in pregnancy Trans R Soc Trop Med Hyg 2003 97 30 35 12886801 10.1016/S0035-9203(03)90012-5
Rogerson SJ Van den Broek NR Chaluluka E Qongwane C Mhango CG Molyneux ME Malaria and anaemia in antenatal women in Blantyre, Malawi: a twelve- months survey Am J Trop Med Hyg 2000 62 335 340 11037774
Bouyou-Akotet MK Ionete-Collard DE Mabika-Manfoumbi M Kendjo E Matsiegui P-B Mavoungou E Kombila M Prevalence of Plasmodium falciparum infection in pregnant women in Gabon Malar J 2003 2 18 12919637 10.1186/1475-2875-2-18
Dicko A Mantel C Aly Thera M Doumbia S Diallo M Diakete M Sagara I Doumbo O Risk factors for malaria infection and anemia for pregnantwomen in the Sahel area of Bandiagara, Mali Acta Trop 2003 89 17 23 14636978 10.1016/j.actatropica.2003.07.001
Al Gadal AA Buck AA Malaria in the Sudan Proceedings of the Conference on Malaria in Africa 1986 Washington, DC: American Institute of Biological Sciences 156 159
World Health Organization Severe falciparum malaria Trans R Soc Trop Med Hyg 2000 94
Sahli H Lehrbuch der Klinischen unter Suchungsmethoden 1984 Leipzig: Deuticke
Luxemburger C Ricci F Nosten F Raimond D Bathet S White NJ The epidemiology of severe malaria in an area of unstable and low transmission in Thailand Trans R Soc Trop Med Hyg 1997 91 256 262 9231189 10.1016/S0035-9203(97)90066-3
Adam I Mirghani OM Saed OK Ahmed SM Mohamadani AA Ahmed HM Mackenzie CD Homeida MM Elbashir MI Quinine therapy in severe Plasmodium falciparum malaria during pregnancy in Sudan East Mediterr Health J 2004
Lander J Leroy V Simonon A Karita E Bogaerats J Clercq AD Van de Perre P Dabis F HIV infection, malaria, and pregnancy: a prospective cohort study in Kigali, Rwanda Am J Trop Med Hyg 2002 66 56 60 12135269
Adam I Elbashir MI Comments on the article: Risk factors for malaria infection and anemia for pregnant women in the Sahel area of Bandiagara, Mali Acta Trop 2005
| 15829008 | PMC1087872 | CC BY | 2021-01-04 16:37:30 | no | Malar J. 2005 Apr 13; 4:18 | utf-8 | Malar J | 2,005 | 10.1186/1475-2875-4-18 | oa_comm |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.